jun18 medpacreporttocongress sec

Transcript

1 JUNE 2018 Report to the Congress: Medicare and the Health Care Delivery System T TO THE CON REPOR G RESS Medicare and the Health Care Delivery System | June 2018 Washington, DC 20001 Suite 701 • 425 I Street, NW • (202) 220-3700 • Fax: (202) 220-3759 • www.medpac.gov

2 The Medicare Payment Advisory Commission (MedPAC) is an independent congressional agency established by the Balanced Budget Act of 1997 (P.L. 105–33) to advise the U.S. Congress on issues affecting the Medicare program. In addition to advising the Congress on payments to health plans participating in the Medicare Advantage program and providers in Medicare’s traditional fee-for-service program, MedPAC is also tasked with analyzing access to care, quality of care, and other issues affecting Medicare. The Commission’s 17 members bring diverse expertise in the financing and delivery of health care services. Commissioners are appointed to three-year terms (subject to renewal) by the Comptroller General and serve part time. Appointments are staggered; the terms of five or six Commissioners expire each year. The Commission is supported by an executive director and a staff of analysts, who typically have backgrounds in economics, health policy, and public health. MedPAC meets publicly to discuss policy issues and formulate its recommendations to the Congress. In the course of these meetings, Commissioners consider the results of staff research, presentations by policy experts, and comments from interested parties. (Meeting transcripts are available at www.medpac.gov.) Commission members and staff also seek input on Medicare issues through frequent meetings with individuals interested in the program, including staff from congressional committees and the Centers for Medicare & Medicaid Services (CMS), health care researchers, health care providers, and beneficiary advocates. Two reports—issued in March and June each year—are the primary outlets for Commission recommendations. In addition to annual reports and occasional reports on subjects requested by the Congress, MedPAC advises the Congress through other avenues, including comments on reports and proposed regulations issued by the Secretary of the Department of Health and Human Services, testimony, and briefings for congressional staff.

3 JUNE 2018 REPORT TO THE CONGRESS Medicare and the Health Care Delivery System 425 I Street, NW • Suite 701 • Washington, DC 20001 (202) 220-3700 • Fax: (202) 220-3759 • www.medpac.gov

4

5 601 New Jersey A venue, NW • Suite 9000 425 I Street, NW • Suite 701 Wa shington, DC 20001 Washington, DC 20001 - - - - 202-220-3700 • Fax: 202-220-3759 220 202 3700 • Fax: 202 3759 220 www.medpac.gov medpac.gov www. Glenn M. Hackbar n th, J.D., Chairma Francis J. Crosson, M.D., Chairman Rober t A. Berenson, M.D., F. A.C.P ., Vice Chairma n Jon Christianson, Ph.D., Vice Chairman Mark E. Miller , Ph.D., Executive Director James E. Mathews, Ph.D., Executive Director June 15, 2018 The Honorable Michael R. Pence President of the Senate U.S. Capitol Washington, DC 20510 The Honorable Paul D. Ryan Speaker of the House U.S. House of Representatives U.S. Capitol Room H-232 Washington, DC 20515 Dear Mr. President and Mr. Speaker: I am pleased to submit the Medicare Payment Advisory Commission’s June 2018 Report to the Congress: Medicare and the Health Care Delivery System . This report fulfills the Commission’s legislative mandate to evaluate Medicare payment issues and to make recommendations to the Congress. In the 10 chapters of this report, we consider: the effects of the Hospital Readmissions Reduction Program. • • using payment to ensure appropriate access to and use of hospital emergency department services. rebalancing Medicare’s physician fee schedule toward ambulatory evaluation and management services. • paying for sequential stays in a unified prospective payment system for post-acute care. • encouraging Medicare beneficiaries to use higher quality post-acute care providers. • issues in Medicare’s medical device payment policies. • • applying the Commission’s principles for measuring quality to population-based measures and hospital quality incentives. • recent performance of and long-term issues confronting Medicare accountable care organizations. managed care plans for dual-eligible beneficiaries. • Medicare coverage policy and use of low-value care. •

6 Page 2 I hope you find this report useful as the Congress continues to grapple with the difficult task of controlling the growth of Medicare spending while preserving beneficiaries’ access to high-quality care and providing suf- ficient payment for efficient providers. Sincerely, Francis J. Crosson, M.D. Enclosure

7 Acknowledgments Lynch, Don May, Sharon McIlrath, Michael McWilliams, This report was prepared with the assistance of many Farzad Mostashari, Ann O’Hare, Andrew Ryan, Thomas people. Their support was key as the Commission Ryan, Sherry Smith, Steve Speil, Aaron Tripp, and Carolyn considered policy issues and worked toward consensus. Zollar. Despite a heavy workload, staff members of the Centers for Medicare & Medicaid Services and the Department Once again, the programmers at Social and Scientific of Health and Human Services were particularly helpful Systems provided highly capable assistance to Commission during preparation of the report. We thank Susan Bogasky, staff. In particular, we appreciate the hard work of Tim Engelhardt, Kate Goodrich, Ryan Howe, Joel Michael Brown, Po-Lun Chou, Daksha Damera, Darya Kaiser, Pauline Lapin, Paul Moore, John Pilotte, Suzanne Leyzarovich, Sravani Mallela, Sanee Maphungphong, Seagrave, Tiffany Swygert, Tamara Syrek-Jenson, Donald Shelley Mullins, Lorena Ortiz, Cindy Saiontz-Martinez, Thompson, and Marie E. Vasbinder. and Susan Tian. In particular, this year, the Commission wishes to acknowledge Mary Beth Spittel, who in 2018 The Commission also received valuable insights and left Social and Scientific Systems after nearly 15 years assistance from others in government, industry, and the of service—the full duration of which she managed the research community who generously offered their time Commission’s analytic and data support portfolio. She has and knowledge. They include Peter Bach, Kirstin Blom, been an essential partner in the conduct of our work; she Cristina Boccuti, David Certner, Mike Cheek, Mike will be missed tremendously, and we wish her well in her Chernew, James Cosgrove, Akin Demehin, Al Dobson, new endeavors. Bill Dombi, Laura Dummit, Tom Fise, Greg Fonarow, Erin Giovannetti, Charles Haley, Vivian Ho, Gretchen Finally, the Commission wishes to thank Hannah Fein, Jacobson, Joanna Kim, Kathy King, Jennifer Kowalski, Mary Gawlik, and Melissa Lux for their help in editing Harlan Krumholz, Miriam Laugesen, Christine Aguiar and producing this report. ■ June 2018 | Report to the Congress: Medicare and the Health Care Delivery System v

8

9 Table of contents ... v Acknowledgments Executive summary ... xi Chapters 1 Mandated report: The effects of the Hospital Readmissions Reduction Program ... 3 Background 7 ... Prior research on the effects of the HRRP ... 9 Our methodology for evaluating the HRRP effects ... 12 Results ... 14 Policy implications ... 26 2 Using payment to ensure appropriate access to and use of hospital emergency department services ... 35 ... 39 Background Rural areas: Maintaining access to emergency department services ... 46 Urban areas: Incentives have led to an abundance of urban stand-alone EDs ... 53 Future analyses ... 58 3 Rebalancing Medicare’s physician fee schedule toward ambulatory evaluation 65 and management services ... ... 69 Background on the fee schedule for physician and other health professional services Ambulatory E&M services are underpriced relative to other services ... 70 79 An approach to rebalance the fee schedule toward ambulatory E&M services ... Conclusion ... 82 4 Paying for sequential stays in a unified prospective payment system for post-acute care ... 87 Background ... 91 Challenges with paying for sequential post-acute care stays ... 91 Summary of the proposed PAC PPS design ... 92 Definition of sequential PAC stays ... 93 93 Characteristics of sequential PAC stays ... 98 PAC PPS payments need to align with the cost of stays throughout a sequence of post-acute care ... Defining the beginning and end of stays when treating in place 101 ... Conclusion ... 105 Encouraging Medicare beneficiaries to use higher quality post-acute care providers ... 111 5 ... 115 Introduction Beneficiaries seeking PAC often have many PAC options that vary substantially in quality ... 115 Patients referred to PAC need assistance to identify better quality providers ... 117 Principles for improving hospital discharge planning ... 121 Approaches for identifying higher quality PAC providers ... 123 126 Conclusion ... | June 2018 Report to the Congress: Medicare and the Health Care Delivery System vii

10 Issues in Medicare’s medical device payment policies 6 133 ... ... 137 Introduction 137 DMEPOS background ... ... 145 Non-CBP DMEPOS products Policy options to improve the accuracy of Medicare’s payment rates for non-CBP DMEPOS products and 153 protect beneficiaries ... 158 Physician-owned distributors ... Conclusion ... 165 7 Applying the Commission’s principles for measuring quality: Population-based measures and hospital quality incentives ... 175 Introduction 179 ... 179 Applying the Commission’s principles for measuring quality to population-based measures ... ... 189 Applying the Commission’s principles for measuring quality to hospital quality incentives 8 Medicare accountable care organization models: Recent performance and long-term issues ... 211 ... Introduction 215 Background on ACOs ... 215 ACO quality and financial performance relative to CMS-designed benchmarks ... 218 ACO quality and financial performance results according to other researchers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226 New tools to allow ACOs to manage care ... 229 Long-term issues for Medicare ACOs ... 230 Conclusion ... 237 Managed care plans for dual-eligible beneficiaries 243 9 ... Introduction ... 245 Background on dual-eligible beneficiaries ... 245 Update on the financial alignment demonstration ... 246 More states are using Medicaid managed care for dual eligibles ... 264 Medicare plans that serve dual eligibles differ in key respects ... 266 Potential policies to encourage the development of integrated plans ... 276 Conclusion ... 283 10 Medicare coverage policy and use of low-value care ... 293 Primer on Medicare coverage policy ... 297 ... Evidence of low-value care 310 Tools for addressing low-value care ... 334 Conclusion 349 ... Appendix A Commissioners’ voting on recommendations ... 367 Acronyms ... 371 More about MedPAC Commission members ... 377 Commissioners’ biographies ... 379 383 ... Commission staff Table of contents viii

11 Executive summary

12

13 Executive summary them more transparent to beneficiaries, enforcement As part of its mandate from the Congress, each June the agencies, and others. Commission reports on refinements to Medicare payment systems and issues affecting the Medicare program, • Applying the Commission’s principles for measuring including broader changes in health care delivery and the quality: Population-based measures and hospital market for health care services. In the 10 chapters of this We formalize the Commission’s quality incentives. report we consider: quality principles and apply them to two population- • The effects of the Hospital Readmissions Reduction based outcome measures that may be used to evaluate In this mandated report, we conclude Program. quality of care for different populations. We also apply that the Hospital Readmissions Reduction Program the principles to the design of a new hospital quality contributed to a significant decline in readmission incentive program that combines measures of hospital rates without causing a material increase in emergency outcomes, patient experience, and Medicare spending department (ED) visits or observation stays or an per beneficiary. adverse effect on mortality rates. • Medicare accountable care organization models: Using payment to ensure appropriate access to and • We review Recent performance and long-term issues. use of hospital emergency department services. To the current Medicare accountable care organization reduce the risk of ED services being undersupplied (ACO) models and look at ACO performance on cost in rural areas and oversupplied in urban areas, we and quality thus far. Based on this review, we raise six recommend two changes to Medicare payment for ED issues that are important for two-sided-risk ACOs in services. the long term. Rebalancing Medicare’s physician fee schedule • Managed care plans for dual-eligible beneficiaries. • toward ambulatory evaluation and management We consider three potential policies to encourage services. We describe a budget-neutral approach the development of plans that integrate care for to rebalance the fee schedule that would increase individuals who receive both Medicare and Medicaid payment rates for ambulatory evaluation and (known as dual-eligible beneficiaries). management services while reducing payment rates Medicare coverage policy and use of low-value care. • for other services. We find that the fee-for-service coverage process Paying for sequential stays in a unified prospective • does not prevent the use of low-value services and We consider payment system for post-acute care. that the use of such services is prevalent in Medicare. refinements to a unified post-acute care (PAC) We describe six tools that Medicare could consider to prospective payment system, focusing on increasing reduce the use of low-value care. the accuracy of payment for cases that involve a Mandated report: The effects of the Hospital course of PAC care—that is, sequential stays. Readmissions Reduction Program • Encouraging Medicare beneficiaries to use higher To encourage hospitals to reduce preventable quality post-acute care providers. We discuss readmissions, CMS began to publicly report hospitals’ increasing the use of higher quality PAC providers. readmission rates for three conditions in 2009. In 2010, At discharge from an inpatient stay, the selection of a the Congress added a financial incentive to reduce provider within a PAC category can be crucial because readmission rates when it enacted legislation providing for the quality of care varies widely among providers. the Hospital Readmissions Reduction Program (HRRP). At the same time, the Congress funded programs to help • Issues in Medicare’s medical device payment hospitals improve care transitions and reduce preventable We explore ways to improve Medicare’s policies. readmissions. The end goal of reducing hospital payment policies for durable medical equipment, readmissions is to relieve Medicare beneficiaries of the prosthetic devices, prosthetics, orthotics, and burden of returning to the hospital and to relieve taxpayers supplies. We also address how to constrain the risks of the cost of unnecessary readmissions. posed by physician-owned distributors by making June 2018 | Report to the Congress: Medicare and the Health Care Delivery System xi

14 ED visits increased after introduction of the HRRP. • In the 21st Century Cures Act of 2016, Congress However, this increase appears to be due primarily to mandated that the Commission evaluate whether the reasons other than the HRRP. recent declines in readmission rates were associated with offsetting increases in observation stays and ED visits. Some researchers have raised the question of whether In Chapter 1, we first conclude that HRRP did indeed efforts to reduce avoidable readmissions have also reduced reduce readmission rates. We then consider the question necessary readmissions, resulting in higher mortality in the mandate and, finally, evaluate whether hospitals for heart failure patients. We examined readmission and that lowered their readmission rates saw an increase in mortality changes from 2010 to 2016. Our measure of mortality rates. mortality includes deaths that occurred during the hospital stay and within 30 days after discharge. We found no Hospitals’ response to the HRRP has contributed to evidence to suggest that the readmission policy on net had a large decline in readmissions since 2010, with the a negative effect on mortality. To the extent that there was greatest declines being in conditions initially covered by a small effect, our data as a whole suggest the HRRP may the program (acute myocardial infarction (AMI), heart have done more to improve than harm mortality rates. failure, and pneumonia). We measured the change in readmission rates from 2010 to 2016 and found that raw In summary, the HRRP gave hospitals an incentive to (not risk-adjusted) readmission rates fell by 3.6 percentage reduce inappropriate readmissions. After implementation points for AMI, 3.0 percentage points for heart failure, of the HRRP, readmission rates declined, and our analysis and 2.3 percentage points for pneumonia, compared suggests the decline was in part due to the HRRP. with 1.4 percentage points on average across conditions Beneficiaries endured fewer readmissions to the hospital, not covered by the program. Our analyses support the without an increase in risk-adjusted mortality. While the conclusion that the HRRP led to fewer readmissions. HRRP may have contributed slightly to the secular trend of increasing observation and ED use, the small increases readmission rates for heart raw The rate of decline in • in costs were far outweighed by reduced readmissions risk-adjusted failure, pneumonia, and AMI and in costs. (The decline in readmissions across all conditions readmission rates for heart failure were faster by resulted in net savings to the Medicare program of roughly a statistically significant amount after HRRP’s $2 billion per year.) enactment (2010 to 2016) than in prior years. • Raw and risk-adjusted readmission rates declined Using payment to ensure appropriate faster, on average, for conditions covered by the access to and use of hospital emergency program than for other conditions. The difference is department services statistically significant. Medicare’s payment policies should foster adequate access to care and encourage efficient delivery of services. After the reduction in readmission rates, some researchers Maintaining access to ED services can be a challenge in expressed concerns that the lower rates may have induced remote rural areas, where a single hospital may be the an increase in observation stays or ED use. Our analysis sole source of ED care. If that hospital closes, access to found the following: emergency care can be lost. In contrast, efficiency can be a challenge in urban areas, where EDs can be in oversupply. Observation stays increased at a slightly faster rate • New urban stand-alone EDs could result in patients being after introduction of the HRRP. However, the increase treated at higher cost EDs rather than lower cost urgent in observation stays was small and offset only a small care facilities and physician offices. These facilities also share of the reduction in readmissions. Therefore, could siphon off lower acuity patients from on-campus we conclude that the reduction in readmission hospital-based EDs. To reduce the risk of ED services rates reflects real changes in practice patterns and being undersupplied in rural areas and oversupplied in not simply a shifting of short-stay admissions into urban areas, in Chapter 2, we recommend two changes to observation stays to avoid readmission penalties. We Medicare payment for ED services. also found similar rates of increase in observation stays among patients without a recent admission. Maintaining access to ED services can be challenging in isolated rural areas with low population densities. Executive summary xii

15 Medicare overpays these facilities relative to what is paid Hospitals in many isolated rural areas have seen the to on-campus hospital EDs for more difficult cases. number of inpatient cases fall dramatically; many hospitals now average less than one inpatient admission per day. Medicare currently has two levels of payments for However, Medicare will pay a facility for emergency OCEDs. One is for EDs open 24 hours a day, 7 days a services only if it maintains inpatient services. Therefore, week (Type A payment rates), and the other is for EDs small isolated communities that want an ED must maintain open less than 24 hours a day, 7 days a week (Type B a low-occupancy inpatient department in the hospital. payment rates). In 2018, Type B payment rates are roughly 30 percent lower than Type A rates. The Commission As an alternative to maintaining empty inpatient beds, recommends that Medicare pay urban OCEDs the Type A the Commission recommends a new payment model that payment rates reduced by 30 percent—which would better would allow Medicare to pay for emergency services at align payments with costs and make off-campus ED rates outpatient-only hospitals in isolated rural areas (more similar to Type B rates. An exception would be needed than 35 miles from another ED). Isolated rural full- for the one-quarter of urban OCEDs located relatively service hospitals that choose to convert to outpatient-only far (more than six miles) from on-campus EDs and that hospitals would receive the same standard prospective are more likely to provide unique access to ED services payment rates for ED visits as a full-service hospital. for their local communities (other exceptions could be In addition, a set annual payment (common across all contemplated when an urban OCED is essential to retain outpatient-only hospitals) would be made to help cover the access—for example, if the OCED is the result of its facility’s fixed costs. parent hospital closing). Paying these more isolated urban The new payment option would allow rural communities OCEDs the full Type A payment rates would be justified that cannot support a full-service hospital to maintain to ensure continued appropriate access to emergency access to emergency care in their community while services. This recommendation also would reduce cost retaining the option to convert back to a full-service sharing for Medicare beneficiaries served at OCEDs close hospital if circumstances changed. The recommendation to on-campus EDs. Overall, this policy would reduce the would increase Medicare spending by less than $50 financial incentive to develop new OCEDs and would million per year. lower Medicare spending by between $50 million and $250 million annually. Conversely, an oversupply of EDs can be a problem in urban areas. Urban hospitals can set up stand-alone EDs Rebalancing Medicare’s physician fee that bill Medicare as if they are part of the hospital’s main schedule toward ambulatory evaluation and ED as long as those EDs are located within 35 miles of management services the main hospital campus. We refer to these facilities as The Commission is concerned that ambulatory evaluation off-campus EDs (OCEDs). The number of OCEDs has and management (E&M) services, such as clinician increased rapidly in recent years, particularly in areas with office and hospital outpatient visits, are underpriced in high household incomes. The number of ED visits and the the Medicare fee schedule for physicians and other health share of visits with high coded severity levels also have professionals (“the fee schedule”) relative to other services increased. Under Medicare’s current payment system, such as procedures. CMS has made incremental efforts to providers have an incentive to add new OCEDs rather review potentially mispriced services over the last several than urgent care centers, which are paid less than half the years, but there is evidence that certain types of services hospital ED rates. are still overpriced. CMS’s lack of current, accurate, and objective data on clinician work time and practice Patients who seek care at OCEDs appear to have less expenses is a key reason the review process has been complex care needs than those of patients served at on- inadequate. Under the fee schedule’s budget-neutrality campus hospital EDs. Ambulance operators typically take rules, the relative prices for ambulatory E&M services are trauma, stroke, and heart attack patients to on-campus too low because the prices for other services have become hospital EDs, which provide trauma services, operating artificially high. We call this process “passive devaluation.” rooms, and inpatient services. OCEDs do not incur the standby costs of these resource-intensive services. While In Chapter 3, we describe a budget-neutral approach for urban OCEDs may provide some services not available at rebalancing the fee schedule that would increase payment doctors’ offices and urgent care centers, we conclude that June 2018 | Report to the Congress: Medicare and the Health Care Delivery System xiii

16 by the Congress, in June 2016, the Commission evaluated rates for ambulatory E&M services while reducing a prototype design and concluded that it was feasible to payment rates for other services (e.g., procedures, design a unified PAC PPS that spans the four settings imaging, and tests). Under this approach, the increased and bases payments on patient characteristics. In June payment rates would apply to ambulatory E&M services 2017, the Commission recommended that a unified PAC provided by all clinicians. For illustration, we modeled PPS be implemented beginning in 2021 with a three-year the impact of a 10 percent increase in the payment rate transition and a corresponding alignment of setting- for ambulatory E&M services (higher or lower increases specific regulatory requirements. could be considered). A 10 percent increase would raise annual spending for ambulatory E&M services by $2.4 In Chapter 4, we consider a refinement to the unified PAC billion. To maintain budget neutrality, payment rates for PPS that would increase the accuracy of payment for cases all other fee schedule services would be reduced by 3.8 that involve a course of PAC care—that is, sequential percent. stays, which we define as PAC stays within seven days of each other. We evaluate two payment issues related Certain specialties would receive a large increase in to sequential stays. The first has to do with the way the their total fee schedule payments (on net) as a result of cost of a stay can vary, depending on where it falls in this change. The three specialties that would receive a sequence of PAC stays. The second involves how to the highest proportional increases in payments are identify, for payment purposes, distinct phases of care endocrinology, rheumatology, and family practice. Other for a PAC provider that treats a patient “in place” as care specialties—including diagnostic radiology, pathology, needs evolve rather than refers the patient to another PAC physical therapy, and occupational therapy—would provider. Under the unified PAC PPS, such providers experience reductions in their fee schedule payments would be financially disadvantaged unless the payment of about 3.8 percent because they provide very few system included a way to trigger payments for different ambulatory E&M services. phases of care. This change would be a one-time adjustment to the fee schedule to address several years of passive devaluation Our analysis of sequential PAC stays found different of ambulatory E&M services. Even if this approach is patterns of costs relative to estimated PAC PPS payments adopted, we urge CMS to accelerate its efforts to improve for home health stays and institutional PAC stays. For home the accuracy of the fee schedule by developing a better health stays, payments under the unified PAC PPS would mechanism to identify overpriced services and adjust decrease over the course of a sequence of stays, but the their payment rates. If successful, these efforts would cost of stays would decline more. These results suggest that improve the accuracy of prices for ambulatory E&M and payments for home health care need a separate downward other services going forward and could reduce the need adjustment for later stays, similar to the adjustment used in for future significant adjustments to the prices of E&M the current HHA PPS. By contrast, PAC PPS payments for services. Together, these actions will help reduce the risk institutional stays would remain reasonably well aligned of beneficiaries experiencing problems accessing these with the cost of stays throughout a sequence of care. services and will send a more positive signal to medical However, under its current design, the prototype PAC PPS students and residents contemplating careers in specialties would not be able to appropriately pay a PAC provider that that provide large shares of these services. offered a range of PAC services and was able to treat in place beneficiaries with evolving care needs. For payment Paying for sequential stays in a unified purposes, Medicare will need to define when one “stay” prospective payment system for post-acute or phase of care ends and the next one begins. Otherwise, care with only one admission and discharge date, providers Medicare uses separate prospective payment systems would receive only one payment, creating a financial (PPSs) to pay for stays in each of the four PAC settings— disincentive to treat in place. skilled nursing facilities (SNFs), home health agencies (HHAs), inpatient rehabilitation facilities (IRFs), and Of the approaches we examined, the most promising long-term care hospitals (LTCHs). As a result, Medicare’s involves episode-based payments; that is, Medicare would fee-for-service (FFS) payments can differ substantially for make a single payment for all post-acute care provided similar patients treated in different settings. As mandated Executive summary xiv

17 have concluded that these efforts have not significantly during an episode of PAC. Payment could be made to increased the use of higher quality PAC providers. a hospital, a health system, the PAC provider where the episode starts, an ACO, or a third-party convener Our analysis of referral patterns of Medicare beneficiaries that assumes financial risk for the episode. Under this who were sent to SNFs and HHAs indicates that many approach, Medicare would not need to define and set beneficiaries had another nearby provider that offered payments for subsequent stays because the entity would be better quality, though not all of the higher quality providers paid for the PAC provided during the episode, regardless may have had available capacity. For example, over 94 of how many stays were encompassed. percent of beneficiaries who used HHA or SNF services had at least one provider within a 15-mile radius that was The Commission will continue to explore episode-based of higher quality than the provider that served them. payments over the coming year. Shifting the unit of service from a stay to an episode would change certain incentives Helping beneficiaries to identify better quality PAC (most notably the incentive to initiate subsequent PAC providers should be a goal in a reformed discharge stays), but the most important features of a PAC PPS planning process, and authorizing hospital discharge would remain: correcting the biases of the current PPSs planners to recommend specific higher quality PAC and increasing the equity of payments across all types of providers would further this goal. However, several design stays so that providers have less incentive to selectively decisions would need to be resolved. First, a consistent admit certain beneficiaries over others. In the meantime, approach to identifying better quality PAC providers CMS should proceed with implementing a stay-based would be needed, and quality standards would need to be unified PAC PPS. transparent for PAC providers and beneficiaries. Second, policies would be needed to safeguard against potential Encouraging Medicare beneficiaries to use conflicts of interest that could ensue from the authority to higher quality post-acute care providers recommend specific providers. About 40 percent of Medicare acute inpatient hospital discharges result in use of PAC. Ensuring that the patient is Regardless of the approach selected to encourage the served by the appropriate type of PAC provider is critical, use of higher quality PAC providers, beneficiaries but the selection of a provider within a PAC category can should retain freedom of choice. Beneficiaries may have also be crucial because the quality of care varies widely important concerns that are not necessarily reflected in among providers. In Chapter 5, we discuss increasing the standard quality measures, such as language competency use of higher quality PAC providers. or proximity to family members. These preferences may lead them to select a PAC provider that has lower Medicare discharge planning regulations place the performance on some quality measures, but additional responsibility on hospitals for connecting acute hospital quality information would allow them to better understand inpatients with their options for PAC—including educating the nature of their options and any trade-offs. beneficiaries about their choices and facilitating access to PAC when necessary. But hospitals are limited in the Medicare’s options for expanding the authority of assistance they can provide. Although they are required to discharge planners to recommend higher quality PAC provide beneficiaries who need PAC with a list of nearby providers range from prescriptive approaches that provide SNFs and HHAs, Medicare regulations prohibit hospitals specific metrics or definitions that hospitals must use from recommending specific PAC providers. to more flexible approaches that leave key decisions to discharge planners. A hybrid approach could blend these Beneficiaries report that they value quality of care and two methods and specify certain selection criteria that that they prefer PAC providers that are close to their hospitals would need to use while granting hospitals home or family. The Improving Medicare Post-Acute discretion in the application of these criteria. Care Transformation Act of 2014 requires hospitals to include quality data when informing beneficiaries about Issues in Medicare’s medical device payment their options, but CMS has yet to finalize the regulations policies implementing this requirement. Medicare has developed In Chapter 6, we explore two distinct topics related consumer-oriented websites that provide information to medical devices. First, we look at ways to improve on the quality of SNFs and HHAs, but many studies Medicare’s payment policies for durable medical equipment, prosthetic devices, prosthetics, orthotics, June 2018 | Report to the Congress: Medicare and the Health Care Delivery System xv

18 and supplies (DMEPOS). Second, we examine ways owners perform on their own patients. PODs have the to constrain the risks posed by physician-owned ability to distort the supply chain for medical devices— distributors (PODs) and to make them more transparent to potentially resulting in an increase in the volume of beneficiaries, enforcement agencies, and others. surgeries performed on beneficiaries, higher costs for hospitals and the Medicare program, and inappropriate Medicare beneficiaries rely on DMEPOS products to care. treat their illness or injury and to allow them to remain in their homes, as opposed to seeking care in an institutional The Commission questions the value PODs produce for setting. DMEPOS comprises a large array of products that the Medicare program and beneficiaries. We suggest vary in cost and complexity, ranging from complex power several ways in which Medicare and policymakers wheelchairs to diabetes testing supplies to knee braces. can constrain the risks posed by PODs. We discuss two specific options to revise the Stark law (which is Pursuant to a statutory requirement, CMS implemented intended to prohibit physicians from referring Medicare the DMEPOS Competitive Bidding Program (CBP) to use beneficiaries to certain health care facilities in which market competition to set payment rates and limit fraud they have a financial interest) and several key topics and abuse, while ensuring beneficiaries retain access to for policymakers to consider if such changes are made. needed DMEPOS products. The CBP began in 2011 with While the options likely would limit the use of PODs, some of the highest cost and highest volume DMEPOS some PODs might continue to operate, even if the products in nine large urban areas. Over time, the CBP Stark law were modified. In addition, the Commission has added products and expanded geographically. The supports increasing the transparency of POD-physician CBP has successfully driven down the cost of DMEPOS relationships by requiring all PODs to report under the products for the Medicare program and beneficiaries. Open Payments program, a program designed to shed light Compared with payment rates in the year before the on financial ties between physicians and certain industries. CBP, Medicare’s payment rates for some of the highest expenditure DMEPOS products have fallen by an average Applying the Commission’s principles for of roughly 50 percent. measuring quality: Population-based measures and hospital quality incentives At the same time, Medicare expenditures for DMEPOS The Commission has recommended that Medicare link products excluded from the CBP have continued to grow. payment to the quality of care to reward accountable By 2015, nearly half of all Medicare expenditures on entities and providers for offering high-quality care to DMEPOS products were for products excluded from beneficiaries. In Chapter 7, the Commission formalizes the CBP. Medicare pays for these products using a fee a set of principles for measuring quality in the Medicare schedule that is largely based on supplier charges from program. Overall, quality measurement should be patient 1986 to 1987 (updated for inflation) and undiscounted oriented, encourage coordination, and promote delivery list prices. Medicare’s payment rates for the top 10 non- system change. Medicare quality incentive programs CBP DMEPOS products in 2015 were a third higher, on should use a small set of population-based measures (e.g., average, than private-payer rates for comparable products, outcomes, patient experience, value) to assess quality and some non-CBP DMEPOS products continue to of care for populations served by Medicare Advantage generate high rates of improper payments and utilization (MA) plans, ACOs, FFS in market areas, hospitals, groups growth and to exhibit patterns of potential fraud and abuse. of clinicians, and other providers. Medicare quality To address these issues, additional products that are not incentive programs should score these risk-adjusted, currently competitively bid could be moved into the CBP. population-based measure results against absolute We also observe that the participation and balance billing performance thresholds and then use peer grouping to rules for DMEPOS products and suppliers could be determine payment adjustments based on the provider’s strengthened to better protect beneficiaries and better align quality performance. In Chapter 7, we first apply the those policies with many other Part B services. Commission’s principles to two population-based outcome measures (potentially preventable admissions and home PODs are entities that derive revenue from selling, and community days) that may be used to evaluate quality or arranging for the sale of, devices ordered by their of care for different populations. Next, we apply the physician-owners for use in procedures the physician- principles to the design of a new hospital quality incentive Executive summary xvi

19 program that combines measures of hospital outcomes, modeled an HVIP in which quality-based payments are patient experience, and Medicare spending per beneficiary. distributed to hospitals organized into 10 peer groups, with awards funded by a payment withhold from all hospitals. Potentially preventable admissions (PPAs) constitute an important quality measure because hospitalizations for Under our HVIP model, relative to the withhold, about conditions such as diabetes and pneumonia can potentially half of hospitals would receive a negative payment be preventable if ambulatory care is provided in a timely adjustment, and about half would receive a positive and effective manner. We calculated the observed rate of adjustment. Our peer grouping of hospitals allowed us PPAs per 1,000 FFS beneficiaries for both chronic and to examine how hospitals serving large shares of low- acute conditions. We found that observed (that is, not income patients perform. We found that, compared with risk adjusted) PPA rates varied across population groups the existing quality payment programs, the HVIP approach and across market areas and hospital service areas. This makes more equitable payment adjustments among variation signals opportunities to improve the quality of hospitals that serve different populations. Over the next care within areas and the potential to use the measure year, the Commission plans to continue to design an HVIP to compare quality across local health care markets. that conforms with our principles for quality measurement. However, more development is needed to incorporate risk Some topics the Commission will further explore adjustment based on FFS data in the analysis. include weighting of measures, withhold values, patient experience measures, and patient safety measures. The Commission also tested a prototype home and community days (HCDs) measure to assess how Medicare accountable care organization well health care markets and organizations that take models: Recent performance and long-term responsibility for a population keep people alive and issues out of health care institutions. The HCD measure is Medicare ACOs were created to help moderate the defined as 365 days minus the sum of days a beneficiary growth in Medicare spending and improve quality spends in certain institutional and ambulatory health care of care for beneficiaries by giving providers greater settings coupled with mortality days. However, because responsibility for costs and quality. In Chapter 8, we first of the limited variation in HCDs over market areas and review the current Medicare ACO models and look at the challenges posed by the need to develop appropriate their performance on cost and quality. We find that some weights for constructing the composite measure, the models—predominantly two-sided models at risk for both Commission questions the immediate utility of the HCD savings and losses—are producing small savings relative measure in its current form to assess market-level FFS to the benchmarks set by CMS, and all are maintaining performance. or improving quality. Spending relative to benchmarks is important because it determines which ACOs will We also examined the potential to create a single quality- receive “shared savings” bonuses. However, some have based payment program for hospitals to replace the four observed that benchmarks are not necessarily the best current hospital payment incentive programs Medicare measure of what spending would have been in the absence uses: the Hospital Inpatient Quality Reporting Program, of the ACO and thus may not be a good measure of true Hospital Readmissions Reduction Program, Hospital- program savings. We review the literature on this question Acquired Condition Reduction Program, and Hospital and conclude that ACOs may have been saving Medicare Value-based Purchasing. The Commission is concerned 1 percent to 2 percent more than indicated by their that these overlapping hospital quality payment and performance relative to benchmarks, and that two-sided reporting programs create unneeded complexity in the ACO models appear to save more than one-sided ACO Medicare program. models. Ideally, the Congress could redesign the multiple hospital In light of evidence indicating that two-sided ACOs tend quality payment programs under a single hospital value to generate greater savings than one-sided ACOs, we incentive program (HVIP) that would be patient oriented, consider six issues that need to be resolved if two-sided encourage coordination across providers and time, and ACOs are going to be part of the Medicare program in the promote change in the delivery system. It also would long term: account for social risk factors by adjusting payment through peer grouping. Based on these principles, we June 2018 | Report to the Congress: Medicare and the Health Care Delivery System xvii

20 We Are hospitals viable participants in ACOs? • a group of patients and then transition toward taking find that, despite the apparent conflict in incentives, full accountability as an MA plan. We have found in hospitals may still want to participate in ACOs previous work that ACOs can be the low-cost option in because most savings for ACOs to date stem from some areas of the country, and their advantage of lower reduction in the use of post-acute care and not from administrative costs could make them a long-term reductions in inpatient care. option if benchmarks are set equitably. Should asymmetric models be continued? • Managed care plans for dual-eligible Asymmetric models—models with greater beneficiaries opportunities for savings than losses—could be one Individuals who receive both Medicare and Medicaid strategy to help ACOs transition to two-sided risk. (known as dual-eligible beneficiaries) often have complex The Commission will monitor the current asymmetric health needs but are at risk of receiving fragmented or ACO models to determine whether aspects of them low-quality care because of the challenges in obtaining should be extended. care from two distinct programs. Many observers have argued that the two programs could be better integrated • How should benchmarks be set initially and rebased by developing managed care plans that provide both for subsequent agreement periods? The basic ACO Medicare and Medicaid services. Supporters argue that model essentially sets benchmarks as a function of integrated plans would improve quality and reduce federal historical spending for beneficiaries who would have and state spending because they would have stronger been attributed to the ACO in the past. In subsequent incentives to coordinate care than either program has when agreement periods, ACOs must continuously improve acting on its own. However, these plans have been difficult over their own past performance to achieve savings, to develop, and only 8 percent of full-benefit dual-eligible which can create diminishing returns for consistently beneficiaries are now enrolled in a plan with a high level successful ACOs and potentially discourage long- of Medicare and Medicaid integration. In Chapter 9, we term participation. We discuss this issue and others examine the use of integrated plans and consider three related to benchmarking, and then highlight other potential policies that would encourage the development benchmarking approaches. of highly integrated plans. Should the 5 percent bonus for clinicians in • Since 2013, CMS and 10 states have tested the use of advanced alternative payment models (A–APMs) integrated Medicare–Medicaid Plans (MMPs) as part be distributed differently to encourage A–APM of the financial alignment demonstration. There are Under current law, clinicians receive a participation? limited data available on the demonstration’s effects on 5 percent bonus on all of their physician fee schedule quality, service use, and cost because the evaluations of (PFS) payments if they exceed a threshold level on the demonstration are taking longer to complete than payments or patients in A–APMs. Moving to a system expected. However, the information available is generally in which clinicians receive a 5 percent bonus with positive. Although the individual demonstrations often certainty on their share of PFS payments derived from have been difficult to implement, enrollment now an A–APM could make the incentive more equitable appears stable (although participation is lower than many and encourage participation in two-sided ACOs. expected), and quality appears to be improving. What will be the relationship between specialists and • The demonstration is part of a broader effort by many two-sided ACOs? We find that currently there are a states to use Medicaid managed care to provide long-term substantial number of specialists on the participant services and supports (LTSS), such as nursing home care lists of ACOs. ACOs may include specialists as a and personal care. Between 2004 and 2018, the number of way to more effectively coordinate the care of their states with managed LTSS programs grew rapidly from 8 beneficiaries, and specialists may join ACOs to receive to 24, and more states likely will develop similar programs referrals and potentially share in savings. in the future. The growing use of managed care to provide Are two-sided ACOs a long-term option in the • LTSS—which account for most of Medicaid’s spending Some maintain that ACOs are Medicare program? on dual eligibles—means that, in many states, the one way for providers to take greater accountability for development of health plans that provide both Medicare Executive summary xviii

21 and Medicaid services is probably the most feasible the use of low-value services. MA plans are permitted approach for pursuing closer integration. to use tools that are not widely used in FFS Medicare, such as requiring prior authorization to have a service Medicare now has four types of plans that serve dual covered and using variable levels of cost sharing. Part D eligibles: the demonstration’s MMPs, MA dual-eligible plan sponsors are responsible for creating and managing special needs plans (D–SNPs), fully integrated dual- formularies, which are lists of drugs their plans cover. eligible SNPs (FIDE SNPs), and the Program of All- By contrast, Medicare FFS lacks the flexibility to use Inclusive Care for the Elderly. There are significant formularies for drugs covered by Part B. differences among these plans in several key areas, such as their level of integration with Medicaid, ability to Our review of the literature on low-value care reveals that use passive enrollment, and payment methodology. In such care is prevalent across FFS Medicare, Medicaid, and addition, allowing MMPs and D–SNPs to operate in the commercial insurance plans. Evidence suggests that the same market has been problematic in some states because amount of low-value care within a geographic area appears competition between the plans has reduced enrollment to be more a function of local practice patterns than payer in the more highly integrated MMPs. Policy changes to type. We analyzed selected low-value services in FFS better define the respective roles of each type of plan or Medicare using 31 evidence-based measures developed consolidate plans in some fashion may be needed. by a team of researchers. In 2014, there were between 34 and 72 instances of low-value care per 100 beneficiaries— Three potential policies that would encourage the depending on whether we used a narrow or broad version development of integrated plans are (1) limiting how often of each measure—and annual Medicare spending for dual-eligible beneficiaries can change their coverage, these services ranged from $2.4 billion to $6.5 billion. (2) limiting enrollment in D–SNPs to dual eligibles who The spending estimates are conservative because they do receive full Medicaid benefits, and (3) expanding the use not reflect the downstream cost of low-value services. We of passive enrollment, particularly when beneficiaries first also conducted three case studies on care of potentially qualify for Medicare. Collectively, these policies would low value in FFS Medicare: the trend in starting dialysis improve care coordination and continuity of care, require earlier in the course of chronic kidney disease, proton D–SNPs to focus on the dual eligibles who stand to benefit ® beam therapy, and H.P. Acthar Gel (a drug covered under the most from integrated care, and encourage more dual Part D). eligibles to enroll in plans with higher levels of Medicare– Medicaid integration. Last, we identified six tools that Medicare could consider using to address the use of low-value care. Medicare coverage policy and use of low- Expanding prior authorization, which requires • value care providers to obtain approval from a plan or payer Some researchers contend that there is substantial use of before delivering a product or service, could help low-value care—care that has little or no clinical benefit or reduce certain types of low-value care. care in which the risk of harm from the service outweighs its potential benefit—in the Medicare program. Many new Implementing clinician decision support and provider • services disseminate quickly into routine medical care in education could decrease low-value care, and studies FFS Medicare with little or no basis for knowing whether show that these tools have reduced inappropriate they outperform existing treatments. prescribing of antibiotics. In Chapter 10, we review the coverage processes used Increasing cost sharing for low-value services has the • in FFS Medicare and MA plans and by Part D sponsors. potential to reduce their use. Although Medicare does Medicare covers many items and services without the need not currently do so, other health plans and payers have for an explicit coverage policy. When an explicit coverage raised cost sharing for targeted low-value services, and policy is required, some services do not show that they an evaluation of one program found that it reduced the are better than existing covered services. Coverage use of these services. policies often are based on little evidence and usually Establishing new payment models that hold providers • do not include an explicit consideration of a service’s accountable for the cost and quality of care—such as cost-effectiveness or value relative to existing treatment options. As a result, the coverage process does not prevent June 2018 | Report to the Congress: Medicare and the Health Care Delivery System xix

22 services to FFS coverage and payment policies has the ACOs—creates incentives for organizations to reduce low-value services. potential to improve the value of Medicare spending. Medicare’s coverage process considers, but does not Revisiting coverage determinations on an ongoing • require, comparative clinical effectiveness evidence, basis has the potential to both decrease use of low- and the program’s rate-setting processes generally value services and result in the development of more do not consider such evidence. For most items rigorous clinical evidence. and services, Medicare lacks statutory authority to consider evidence on cost-effectiveness in either the • Linking information about the comparative clinical coverage or payment processes. ■ effectiveness and cost-effectiveness of health care Executive summary xx

23 CHAPTER 1 Mandated report: The effects of the Hospital Readmissions Reduction Program

24

25 CHAPTER 1 Mandated report: The effects of the Hospital Readmissions Reduction Program In this chapter Chapter summary To encourage hospitals to reduce preventable readmissions, CMS began to • Background publicly report hospital-level readmission rates for acute myocardial infarction • Prior research on the effects (AMI), heart failure, and pneumonia in 2009. In 2010, the Congress added of the HRRP a financial incentive to reduce readmission rates when it enacted legislation providing for the Hospital Readmissions Reduction Program (HRRP). The Our methodology for • HRRP reduced Medicare payment rates by up to 3 percent for hospitals with evaluating the HRRP effects above-average readmission rates for these three conditions. At this same time, • Results the Congress also funded programs to help hospitals improve care transitions and reduce preventable readmissions. The end goal of preventing hospital • Policy implications readmissions is to relieve Medicare beneficiaries of the burden of returning to the hospital and to relieve taxpayers of the cost of unnecessary readmissions. In recent years, hospital administrators have reported that the HRRP has had a “great impact” on their efforts to reduce readmissions (Joynt et al. 2016). These efforts contributed to a large decline in readmissions since 2010, with the greatest declines in conditions initially covered by the policy (AMI, heart failure, and pneumonia). We measured the change in readmission rates from 2010 to 2016 and found that raw (not risk-adjusted) readmission rates fell by 3.6 percentage points for AMI, 3.0 percentage points for heart failure, 2.3 percentage points for pneumonia, and 1.4 percentage points on average across conditions not covered by the program. To evaluate whether the HRRP led to reduced readmission rates, we conducted a series of longitudinal and June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 3

26 cross-sectional analyses of both raw and risk-adjusted readmission rates. Taken as a whole, our analyses suggest that the HRRP did contribute to the decline in readmission rates. The evidence supporting the conclusion that the HRRP led to fewer readmissions includes the following: • The rate of decline in raw readmission rates for heart failure, pneumonia, and AMI was faster by a statistically significant amount after HRRP’s enactment (2010 to 2016) than in prior years. • The rate of decline in risk-adjusted readmission rates for heart failure was faster by a statistically significant amount after the HRRP’s enactment (2010 to 2016) than during prior years. Risk-adjusted pneumonia and AMI readmission rates also declined faster during the 2010 to 2016 period compared with prior years. However, the difference is not consistently statistically significant across different methods of testing. • Raw readmission rates declined faster, on average, for conditions covered by the program (combining all five conditions in effect in 2016) compared with other conditions. The difference is statistically significant. • Risk-adjusted readmission rates declined slightly faster for HRRP conditions than for non-HRRP conditions. The difference is also statistically significant. • In addition, a study found that readmission rates declined faster for hospitals covered by the policy than for critical access hospitals not covered by the policy (Ibrahim et al. 2017). After the reduction in readmission rates, some researchers expressed concerns that reduced readmission rates may have induced an increase in observation stays or emergency department (ED) use. In the 21st Century Cures Act of 2016, the Congress mandated that the Commission evaluate whether the recent declines in readmission rates were associated with offsetting increases in observation stays and emergency room visits. Our analysis found the following: • Observation stays increased at a slightly faster rate after introduction of the HRRP. However, the increase in observation stays was small and offset only a small share of the reduction in readmissions. Therefore, we conclude that the reduction in readmission rates reflects real changes in practice patterns and not simply a shifting of short-stay admissions into observation stays to avoid readmission penalties. We also found that patients without a recent admission had similar rates of increase in observation stays. The broad-based increase in observation use (including for those without a recent admission) could in part reflect the initiation of the recovery audit contractor reviews of admissions starting in 2010. Therefore, we could not determine conclusively whether the small increase in observation stays was due to the HRRP or to other factors. Mandated report: The effects of the Hospital Readmissions Reduction Program 4

27 • ED visits increased after introduction of the HRRP. However, this increase appears to be due primarily to reasons other than the HRRP. To investigate what share might have been driven by the HRRP, we first compared changes in postdischarge ED use for conditions covered by the HRRP and those not covered by the program. Change in postdischarge ED use was almost identical for HRRP-covered and noncovered conditions. We also compared ED-visit growth for beneficiaries with a recent discharge from a hospital with those growth rates for beneficiaries without a recent hospital discharge. The ED growth rates were approximately equal, and the share of all ED visits that were postdischarge visits was exactly the same (4.8 percent) in 2010 and 2016. Therefore, it appears that the growth in emergency room visits was a broad phenomenon and cannot be primarily attributed to growth in postdischarge ED visits. Some researchers have raised the question of whether efforts to reduce avoidable readmissions have also reduced necessary readmissions, resulting in higher mortality for heart failure patients. The literature is mixed on this question. One recent study reports a slight nationwide increase in 30-day postdischarge mortality rates for heart failure from 2010 to 2014. The study did not examine in-hospital mortality. Because this period of time coincided with the introduction of the HRRP and because readmission penalties are large relative to mortality penalties in the Medicare program, the study’s authors suggested the HRRP may have caused the increase in mortality (Gupta et al. 2017). However, it is not known whether the increase in heart failure mortality reported was caused by the HRRP or other factors, or whether it reflected an increase in patient severity that was not fully reflected by the measure’s risk adjustment model. A separate study used the hospital as the unit of analysis and found that reductions in heart failure readmissions were not correlated with increases in heart failure mortality. It concluded that the HRRP did not cause the increase in heart failure mortality from 2010 to 2014 (Dharmarajan et al. 2017). Using more recent data, we examined readmission and mortality changes from 2010 to 2016. Our measure used a combined inpatient and post-acute mortality. Our findings, which follow, suggest that the HRRP did not negatively affect mortality: Although raw rates of heart failure mortality increased (as has been reported), • raw rates of pneumonia and AMI mortality decreased rapidly after the HRRP was passed. On average, raw rates of mortality declined across HRRP-covered conditions. In contrast, on average, raw rates of mortality increased across non- HRRP conditions. On a risk-adjusted basis, mortality rates declined for all three HRRP-covered • conditions from 2010 to 2016. The combination of an increase in the raw rate of heart failure mortality per discharge and a decline in the risk-adjusted rate June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 5

28 may be explained by an increase in the severity of illness for those beneficiaries admitted for heart failure. While the reported increase in severity of illness may in part reflect greater coding intensity, we believe some of the increase in reported severity is real given the large decline in admissions per capita and the reduced share of cases that were one-day stays. During the 2010 to 2016 period, initial hospital admissions for heart failure per capita fell by 14 percent, which implies that practice patterns changed to treat the less severely ill patients on an outpatient basis. • Our hospital-level analysis also found a slight positive correlation between declining readmission rates and declining mortality across all three conditions, meaning that hospitals with larger than average improvements in readmissions also had larger than average improvements in mortality. Taken together, we find no compelling evidence to suggest that the readmission reduction policy has had a negative effect on mortality. To the extent that there is a small effect, our data as a whole suggest the HRRP may have done more to improve than harm mortality rates. In summary, the HRRP gave hospitals an incentive to reduce inappropriate readmissions. Readmission rates declined, and our analysis suggests the decline was in part due to the HRRP. Beneficiaries had to endure fewer readmissions to the hospital, and patient mortality did not increase because of the HRRP. While the HRRP may have contributed slightly to the secular trend of increasing observation use and ED use, the small increases in costs were far outweighed by reduced costs of readmissions. The decline in readmissions across all conditions resulted in net savings to the Medicare program of roughly $2 billion per year. We conclude that the HRRP contributed to a significant decline in readmission rates without causing a material increase in ED visits, a material increase in observation stays, or a net adverse effect on mortality rates. While the HRRP has largely been successful, that does not mean that hospitals’ financial incentives cannot be improved. In Chapter 7 of this report, we discuss redesigning Medicare’s quality improvement programs for hospitals into a single hospital value incentive program that would balance readmission reduction and mortality reduction incentives across conditions, account for patient experience, and adjust penalties to account for the fact that some hospitals serve larger shares of low-income Medicare beneficiaries. The Commission expects to continue to discuss potential changes to hospitals’ financial incentives in the Medicare program over the next year. ■ Mandated report: The effects of the Hospital Readmissions Reduction Program 6

29 Congressional mandate for this study whether such readmissions are related to any changes The Medicare Payment Advisory Commission shall in outpatient and emergency services furnished. The conduct a study to review overall hospital readmissions Commission shall submit to Congress a report on such described in Section 1886(q)(5)(E) of the Social study in its report to Congress in June 2018. Security Act (42 U.S.C. 1395ww(q)(5)(E)) and ■ (Birmingham and Oglesby 2018, Cary et al. 2018, Ibrahim Background et al. 2017, Medicare Payment Advisory Commission 2016, Zuckerman et al. 2016). While there is a general In 2008, the Commission reported on the need for consensus that readmission rates have declined, some have hospitals, physicians, and post-acute care providers to questioned whether the readmission reduction program improve care transitions and coordination across settings has led to increases in substitute modes of care, such (Medicare Payment Advisory Commission 2008). One as observation stays and emergency department (ED) goal of improved care transitions is to reduce preventable visits (Himmelstein and Woolhandler 2015). Others have readmissions. Unnecessary readmissions can pose risks suggested too many readmissions were avoided, resulting of iatrogenic infections, medication errors, muscle in increased mortality. weakening, and pressure injuries such as decubitus ulcers. In 2016, the Congress passed the 21st Century Cures Act, According to researchers at the Centers for Disease which mandated that the Commission examine how the Control and Prevention, health care–associated infections HRRP affected readmissions, observation stays, and ED in hospitals are a significant cause of morbidity and visits (see text box on the mandate). In response to the mortality in the United States (Klevens et al. 2007). In mandate, this chapter examines how observation stays and addition, the inpatient environment itself can lead to a ED use changed after the introduction of the HRRP. We reduction in elderly patients’ independence as they cope also investigate whether changes in readmission rates are with functional loss that can stem from extended bed rest. related to changes in mortality rates. To create an incentive for hospitals to improve care transitions, the Commission’s June 2008 report Enactment of and changes to the HRRP recommended publicly reporting readmission rates The HRRP was enacted in 2010 and required that and reducing payment rates to hospitals with relatively Medicare payments to hospitals with above-average risk- high readmission rates (Medicare Payment Advisory adjusted readmission rates be reduced starting in 2013. Commission 2008). In the following year, CMS started The 2013 reductions would depend on readmission rates to publicly report hospital-level readmission rates, and a during three previous years (July 2008 to June 2011) for series of articles documented high levels of readmissions three conditions (heart failure, acute myocardial infarction to U.S. hospitals and discussed programs to reduce (AMI), and pneumonia). CMS continues to set penalties readmission rates (Jack et al. 2009, Jencks et al. 2009, for a given year based on readmission performance during Kanaan 2009). In 2010, the Congress enacted the Patient the most recent three-year period of data available (e.g., Protection and Affordable Care Act (PPACA), which fiscal year 2018 penalties are based on discharges from provided for the Hospital Readmissions Reduction July 1, 2013, through June 30, 2016). The HRRP was Program (HRRP). Under this program, hospitals could be later expanded to include three more conditions (chronic penalized (starting in fiscal year 2013) if their readmission obstructive pulmonary disease (COPD), planned hip and rates for certain specified conditions were above the knee replacement surgery, and coronary artery bypass graft national average. (CABG) surgery). A time line of changes to the HRRP is shown in Figure 1-1 (p. 8). After enactment of the readmission reduction program, many studies found that readmission rates declined June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 7

30 FIGURE XXXX X-X FIGURE Timeline for the Hospital Readmissions Reduction Program 1–1 October 2018 2009: October 2013: : October 2015: 2011–2012: Proposed and final Pneumonia measure is CMS begins Maximum penalty Readmission penalties will be expanded to include public reporting rules indicate AMI, increases to 2 percent. heart failure, and CMS adopts the aspiration pneumonia of hospital adjusted using peer readmission rates. pneumonia will be and sepsis with “planned readmission” groups of hospitals pneumonia, more than algorithm, thus based on the share first three measures included in the excluding some doubling the number of of Medicare patients who are fully dually pneumonia cases readmissions from the program. Rules detail how eligible for Medicare measure. covered by the HRRP. program will work. and Medicaid. 2010 2011 2014 2009 2016 2017 2018 2013 2015 2012 October 2012: October 2014: October 2016: March 21, 2010: PPACA passes and CABG surgery is Readmission COPD and hip penalties are establishes the and knee surgery added to program. are added to the implemented for Hospital Readmissions program. three conditions, with a maximum Reduction Program. Maximum penalty is set at 3 percent. penalty of 1 percent of base payments. Note: AMI (acute myocardial infarction), PPACA (Patient Protection and Affordable Care Act of 2010), FY (fiscal year), COPD (chronic obstructive pulmonary disease), CABG (coronary artergy bypass graft). Note and Source in InDesign The HRRP caps the maximum penalty for an individual CMS computation of risk-adjusted readmission rates hospital at 3 percent of total base Medicare inpatient operating payments. In fiscal year 2018, 81 percent of The HRRP measures a hospital’s readmission performance hospitals will have payments reduced because of the using the National Quality Forum–endorsed risk-adjusted 1 HRRP. Most of the penalties are small, with 48 percent of 30-day readmission measures for six conditions. those hospitals receiving less than a 0.5 percent penalty. Measures are for all-cause readmissions for beneficiaries About 6 percent of the penalized hospitals receive the age 65 or older, with limited exclusions such as planned largest penalties (between 2 percent and 3 percent of base readmissions for patients with AMI. Risk adjustment is payments) for their relatively poor performance. The based on the use of hierarchical regression models using average penalty is $217,000 for those hospitals receiving selected hierarchical condition categories to adjust for a penalty in 2018. Total penalties are expected to be patient characteristics. Conditions are identified based on $556 million in 2018, or 0.3 percent of hospitals’ overall the principal discharge diagnosis, which is not necessarily Medicare payments. Mandated report: The effects of the Hospital Readmissions Reduction Program 8

31 the diagnosis related group (DRG) assigned to the case 2 Prior research on the effects of the HRRP for payment. A detailed discussion of how the penalty is computed is included in online Appendix 1-A, available at There is general agreement in the literature that http://www.medpac.gov. readmission rates declined after the passage of the HRRP and that conditions covered under the readmission penalty Commission discussions of potential changes saw the greatest reduction in readmissions (Medicare to the readmission reduction program Payment Advisory Commission 2016, Zuckerman et al. In its June 2013 report to the Congress, the Commission 2016). A survey of hospital administrators found that most suggested several improvements to the HRRP (Medicare believed the HRRP had a “great impact” on their efforts Payment Advisory Commission 2013). The first called for to reduce readmissions, suggesting that at least part of setting a fixed target for readmission rates, so aggregate the reduction in readmissions after the HRRP was due to penalties would go down when industry performance the program’s incentives (Joynt et al. 2016). Readmission improved. A second suggestion was to fix the current rates have also declined for Medicare Advantage and formula by removing the “multiplier,” which sets the privately insured patients, suggesting that factors in readmission penalty equal to a multiple of the price addition to the HRRP are acting to reduce readmissions or associated with the initial admission. As the policy that the effect of the HRRP may have “spilled over” to the currently stands, the penalty for each excess readmission Medicare Advantage and private insurer markets (Chen is disproportionately large relative to the cost of the and Grabowski 2017). As we have stated in the past, readmission. For example, under current law, the penalty reductions in readmissions generated more savings for the for one excess heart failure readmission equals almost 5 program than did the readmissions penalties (Medicare times the cost of the initial heart failure admission, and Payment Advisory Commission 2015). The more the penalty for one excess hip or knee readmission is controversial questions involve how the readmissions over 20 times the cost of an initial admission. Removing penalty affects hospitals serving large shares of low- the multiplier and setting the penalty equal to the cost income patients, whether the reduction in readmissions of excess readmission would reduce the penalty for simply reflects a shifting of patients to observation status a single excess heart failure readmission by about 70 or ED status, and whether reduced readmissions lead to percent and reduce the penalty for a single excess hip/ increased mortality. knee readmission by about 95 percent. A discussion of the penalty multiplier is in online Appendix 1-A, available at Social risk factors and readmission rates available at http://www.medpac.gov. In our initial examination of the readmissions policy, Third, the Commission suggested using an all-condition the Commission found that hospitals with larger shares readmission measure to increase the number of data points of low-income Medicare patients tended to have and reduce the random variation that single-condition systematically higher readmission rates because of readmission rates face under current policy. The extra individual effects, neighborhood effects, or both (Medicare savings from shifting to an all-condition measure would Payment Advisory Commission 2013). Numerous studies fund the cost of removing the multiplier, resulting in budget have similar findings (Gu et al. 2014, Hu et al. 2018, neutrality. A fourth improvement would be to evaluate Hu et al. 2014, Sheingold et al. 2016). The Office of the hospitals’ readmission rates against rates for peer hospitals Assistant Secretary for Planning and Evaluation, which with similar shares of low-income patients as a way to evaluated social risk factors under the hospital value-based adjust penalties for the effects of socioeconomic status. purchasing programs, found that lower income patients The Congress has acted on only one of these options. The did tend to have worse outcomes, but they also found that 21st Century Cures Act (Public Law 114–255) includes hospitals serving more lower income patients tended to a provision (Section 15002) requiring the Secretary of have worse outcomes even after controlling for patient the Department of Health and Human Services to adjust mix (Office of the Assistant Secretary for Planning and readmission penalties using peer groups of hospitals based Evaluation 2016). While hospitals serving the poor tend on the share of Medicare patients who are fully dual- to have higher levels of readmissions, they have also eligible for Medicare and Medicaid starting in fiscal year been able to improve readmission rates faster than other 2019. Descriptions of the problems each policy option aims hospitals (Salerno et al. 2017). to address are shown in the text box (p. 10). June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 9

32 Policy options previously discussed by the Commission to improve the program improve quality of care (Medicare Payment Advisory n its June 2013 report to the Congress, the Commission 2011b). Such a reform could increase the Commission published a chapter on the Hospital likelihood that providers and communities receive the Readmissions Reduction Program (HRRP) I technical assistance that hospitals deem relevant to their that discussed how the program was successful in quality improvement efforts. Other sources of federal motivating hospitals to reduce readmissions. But funding for readmission reduction efforts (such as the we also discussed several problems with the current Partnership for Patients and Community-Based Care program and how the HRRP could be revised to work Transitions Program) encourage hospitals to improve better. Table 1-1 summarizes some of the Commission’s care coordination with providers outside the hospital concerns and policy options to address those concerns. (and thus reduce readmissions) and make other quality In 2011, the Commission recommended redesigning improvements. These programs provide funds for the Quality Improvement Organization program so external organizations to help support hospitals’ efforts that the Secretary could fund time-limited technical to improve patient outcomes. ■ assistance directly to providers and communities to help TABLE Readmission policy issues and proposed solutions 1–1 Proposed solution Concern Description of the problem Small number of It is difficult to distinguish between random Use all-condition readmissions. • observations variation and true performance improvement • Continue to use 3 years of data. when examining a small number of cases for a • Allow hospitals to aggregate performance within small number of conditions. a system. The readmission rates hospital must achieve Create a prospective target. The target could be set Lack of a fixed target below current readmission rates to maintain budget to avoid penalties decrease as industry performance improves. neutrality. Computation of the The penalty is a multiple of the cost of each Drop the multiplier and set the penalty equal to penalty excess readmission. As national readmission the cost of excess readmissions. Use all-condition rates decline, the multiplier increases. Thus, readmissions to offset the cost of removing the penalties per readmission increase. multiplier. Correlation between Report all hospital risk-adjusted rates without an • Lower income patients have higher readmission socioeconomic status rates. SES adjustment. and readmission rates • Compute targets to determine the penalty for peer groups of hospitals with similar low-income shares The Congress enacted a (SSI beneficiaries). ( similar policy that will start in October 2018.) SES (socioeconomic status), SSI (Supplemental Security Income). Note: To protect hospitals serving the poor from experiencing of hospitals with a similar share of low-income patients. disproportionate penalties, the Commission has discussed Under this construct, the actual readmission scores measuring hospitals’ performance against a peer group (unadjusted for social risk factors) would continue to be Mandated report: The effects of the Hospital Readmissions Reduction Program 10

33 reported on Hospital Compare, but the thresholds hospitals Prior studies are inconclusive regarding the would have to meet to avoid readmission penalties would relationship between reduced readmissions and increased mortality be more lenient for hospitals serving more low-income patients (Medicare Payment Advisory Commission 2013). A goal of the HRRP is to improve care transitions and The Congress mandated that this type of peer grouping be coordination between the hospital, physicians, and incorporated in the HRRP beginning in October 2018. post-acute care providers. The benefits of reconciling medication before discharge, ensuring postdischarge Evidence suggests that increased appointments with primary care physicians, and observation care and ED visits are largely coordinating care plans with post-acute care providers due to factors other than the HRRP have been well documented (Coleman et al. 2006, Some researchers have contended that the decline in McHugh et al. 2017, Naylor et al. 2011, Zuckerman et readmissions can be largely attributed to the rapid al. 2017). To the extent that the HRRP improves care increase in use of observation, which means that the coordination—including reconciling medication before patient receives care in the hospital but is not formally discharge—it should reduce readmissions and reduce admitted (Himmelstein and Woolhandler 2015). Their (or at least not increase) mortality. However, some have concern is that clinicians are not truly taking steps to raised concerns that hospitals may avoid appropriate improve care and care transitions. The hypothesis is readmissions, possibly by encouraging ED physicians that the number of events requiring a readmission has to send patients home rather than readmit them, which not truly been reduced, but instead, ED clinicians opt to could lead to higher mortality and lower readmissions. treat these events by keeping the patients in an outpatient Another possibility is that the HRRP induces both positive observation status rather than readmitting them. If that and negative changes in practice patterns, such as better hypothesis were the case, the decline in readmissions care transitions and medication reconciliation but also might result in Medicare program savings but might the discouragement of readmissions that are medically not reflect any true gains in the quality of care for appropriate. beneficiaries. However, the Commission’s 2016 analysis Two 2017 studies examined changes in readmissions and of the increase in observation stays and decline in mortality from 2008 to 2014. Both studies found that readmissions from 2011 to 2013 found that readmission risk-adjusted mortality during the 30 days after hospital rates declined substantially, even after adjusting for discharge increased slightly from 2010 to 2014 (they the growth in observation stays (Medicare Payment 3 ignored in-hospital mortality changes). The question is Advisory Commission 2016). In general, only 20 whether that increase in postdischarge mortality is related percent to 25 percent of the readmissions decline could to the passage of the HRRP or to other factors. The first be accounted for by increased use of one-day or longer study, by Dharmarajan and colleagues, examined hospital- outpatient observation stays. Moreover, we found that level changes in mortality and readmission rates related in that 3-year period, growth in the use of 24-hour-plus to AMI, pneumonia, and heart failure (Dharmarajan et al. observation stays occurring within 30 days of discharge 2017). The researchers examined Medicare discharges at from a hospital (22.2 percent) was essentially the same approximately 3,500 hospitals, including 3 million heart as the overall per capita growth rate in 24-hour-plus failure discharges, and found a slight positive correlation observation stays (22.1 percent). Thus, the increased use between changes in mortality and changes in readmissions. of observation care was not systematically higher for This finding indicates that hospitals that reduced patients with a prior admission than for the Medicare readmissions more than average tended to reduce mortality population overall. Similarly, Zuckerman and colleagues more than average. The magnitude of the correlation for examined data through May 2015 and found “no heart failure is small (0.066), but statistically significant. significant association between changes in observation- Dharmarajan and colleagues concluded that the increasing service use and changes in readmission rates” after rate of postdischarge heart failure mortality was not related implementation of PPACA (Zuckerman et al. 2016). In to reductions in readmission rates. accordance with our mandate, we reexamined whether reduced readmissions were associated with increased The second study, by Gupta and colleagues, which observation stays or increased ED visits after a hospital examined a smaller data set and a narrower question, discharge. We are not aware of any literature that has focused only on heart failure mortality (Gupta et al. examined growth in ED visits after introduction of the 2017). The Gupta study looked only at national trends HRRP. June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 11

34 among heart failure patients using 115,245 discharges mortality for the period during the hospital admission and over 9 years from a sample of 416 hospitals. Gupta extending 30 days postdischarge. However, the Hospital and colleagues found that risk-adjusted mortality rates Compare risk adjustment method produces data that increased after the readmission reduction program was are not designed for longitudinal comparisons. The risk started. Relative to 2010, they found that one-year raw adjustment method is as follows: “The [risk-standardized rates of mortality increased by 3.6 percentage points (from mortality measure] is calculated as the ratio of the 34.5 to 38.1 percent) and risk-adjusted rates increased by number of ‘predicted’ deaths to the number of ‘expected’ 5 percentage points (from 31.3 percent to 36.3 percent) deaths at a given hospital, multiplied by the national 4 after the HRRP’s passage. observed mortality rate” (QualityNet 2017). Because Because the national trend the ratio of predicted to expected deaths is multiplied of increasing risk-adjusted mortality coincided with the by each year’s national raw rate of mortality for the national trend toward lower readmission rates, the authors year, when reviewed over time, the data are indicative of conclude that, “if further confirmed, these findings may trends in raw unadjusted mortality rates. Therefore, the require reconsideration of the HRRP in HF [heart failure].” Chatterjee study indicates that raw (not risk-adjusted) Unlike the Dharmarajan study, however, the Gupta study AMI mortality rates appear to have declined while raw did not conduct a hospital-level analysis to determine (not risk-adjusted) heart failure mortality rates increased whether hospitals with greater readmission reductions also from 2009 through 2015. On average (across baseline had greater mortality increases. poor performers and baseline good performers), the One concern with the Dharmarajan and Gupta studies is study suggests that raw rates of heart-failure mortality that the 2010 to 2014 time frames they used coincided increased slightly. This finding is consistent with the data with a large national drop in initial admissions and a we show in this chapter. However, as we discuss in the shift in the types of patients treated by the hospitals in chapter, raw rates of mortality are not fully illustrative the studies’ samples. This change in admission patterns of trends in risk-adjusted mortality due to increasing could result in a difference in the severity of patients severity of patients admitted for heart failure. that may not be fully picked up by the risk adjuster (as acknowledged by Dharmarajan and colleagues). The In an epidemiological study, Khera and colleagues changing patient mix and practice patterns were reflected reported that one-year mortality following an inpatient in the Gupta study by a doubling of hospice use from 2010 admission for heart failure increased slightly from 2010 to 2014. When Gupta and colleagues removed all hospice to 2012 among a 5 percent sample of Medicare patients patients from their model, the change in 30-day mortality (Khera et al. 2017). However, that article examined all rates after the HRRP’s introduction was no longer heart failure cases, including cases that are not subject to statistically significant. For one-year mortality, the excess the readmissions policy, such as those where heart failure risk of mortality was reduced, but was still slightly positive was a secondary diagnosis on admission. In addition, and remained statistically significant. It is not clear why the study ended before the implementation of the HRRP the HRRP would have a larger effect on one-year mortality penalties. Thus, the primary article contending that the than 30-day mortality. An alternative explanation for HRRP may have resulted in an increase in risk-adjusted increasing heart failure mortality is that patient severity mortality continues to be the article by Gupta and could have changed over the 2010 to 2014 period in ways colleagues. Later in this chapter, we also examine whether that were not fully accounted for by the risk adjuster lower readmission rates are associated with higher risk- (Dharmarajan and Krumholz 2017, Dharmarajan et al. adjusted mortality. 2017). A more recent study looked at Medicare’s Hospital Our methodology for evaluating the Compare data to examine changes in mortality for heart failure and AMI from 2009 through 2015 (Chatterjee and HRRP effects Joynt Maddox 2018). This study used Hospital Compare data to show that, on average, AMI mortality fell during To examine Medicare trends over time (in readmissions, the period, but heart failure mortality increased. However, observation stays, ED visits, and mortality), and heart failure mortality fell for the subset of hospitals hospital-specific correlations between readmission and that initially had high heart failure mortality. Unlike the mortality changes (as Dharmarajan and colleagues did), Gupta and Dharmarajan studies, this study examined we examined changes in readmissions and mortality Mandated report: The effects of the Hospital Readmissions Reduction Program 12

35 from 2010 to 2016. Our mortality analysis examined The population is further limited to beneficiaries at least changes in mortality during the admission and 30 days age 65 who were covered by fee-for-service Medicare postdischarge. As we explain in online Appendix 1-B, (both Part A and Part B) for 12 months before their available at http://www.medpac.gov, we believe looking at admission. the combination of inpatient and postdischarge mortality Risk adjustment is necessary because the severity of will reduce problems that can be caused by a shift in the patients admitted to the hospital has been increasing in site of mortality (for example, from the inpatient setting recent years. While some of the increase in patient risk to hospice, which may have the effect of increasing profiles over time could be because of coding, much of postdischarge mortality). We also put our findings in the increased severity of illness appears to be real. One context by discussing other Medicare program changes potential cause of a real increase in patient severity is the happening at that time. Changes include the Medicare large decline in admission rates since 2010; declining Recovery Audit Contractor (RAC) Program in 2010 admission rates may have raised the severity of illness (which started challenging whether hospital short stays of patients who were admitted. The decline in initial were medically necessary) and more intense coding admissions may have been partially caused by the RAC under the Medicare severity–diagnosis related group Program that started in 2010—the same year the HRRP (MS–DRG) system implemented in 2008. Because these was enacted. The RAC Program gave hospitals incentives factors coincided with the HRRP, we also conducted a to keep less severely ill patients who enter the emergency hospital-level analysis. We examined whether changes room as observation patients rather than admit those in readmission rates correlated with changes in mortality patients into the inpatient system. After introduction of rates, as did Dharmarajan and colleagues. If declines in the RAC Program, the share of patients discharged after risk-adjusted readmissions are correlated with increases in a one-day stay declined and the share of patients staying mortality, that would be of concern. In contrast, if declines longer than one day increased (see online Appendix 1-B, in readmission rates are associated with declines in available at http://www.medpac.gov). Because patients mortality, that would be reassuring. (Online Appendix 1-B, discharged after only one day tend to be less sick, the available at http://www.medpac.gov, provides more detail one-day stays probably had lower risk of readmission and on why we have chosen this methodology.) In this report, mortality. As expected, hospitals reported that the risk we show four types of analyses: (1) trends in raw rates profile of the admitted patients increased over this time of readmission and mortality; (2) trends in risk-adjusted frame. However, some of the increase over time may have readmissions and mortality (because we expect the been due to changes in coding practices. The changes in severity of admitted patients to be increasing, we expect coding pressure and RAC pressure differed over time. The risk-adjusted readmissions and mortality to fall faster than changes could be divided into three key periods: raw readmissions and mortality); (3) trends in observation stays and ED visits; and (4) a cross-sectional analysis 2008 to 2010— In 2008, Medicare introduced MS– • of hospital performance. We examine cross-sectional DRGs. The new DRGs created greater incentives for performance because the time trends for readmissions, complete coding. We and CMS have documented the observation, ED visits, and mortality may be affected by increased coding that occurred from 2008 to 2010 concurrent policy and coding changes, as discussed in (Centers for Medicare & Medicaid Services 2012, online Appendix 1-B. Medicare Payment Advisory Commission 2011a). For that reason, the more rapid decline in risk-adjusted Risk adjustment is necessary but imprecise compared with raw readmission rates during that To evaluate the HRRP’s effects, we started with a period may in part reflect coding changes. population of admissions that are subject to the HRRP From the end of 2010 to 2014, 2010 to 2014— • incentives. This population was identified using the list of hospitals were having the medical necessity of short International Classification of Diseases (ninth and tenth stays challenged by the RACs, resulting in denial revisions) codes that CMS uses to identify eligible cases. of some payments (see Appendix online Appendix We focused our analysis on the five conditions covered by 1-B, available at http://www.medpac.gov). During the HRRP through fiscal year 2016: AMI, heart failure, this period, Medicare admissions per capita declined pneumonia, COPD, and planned hip and knee replacement materially, with the largest declines being for one- surgery (the latter two conditions were added in 2014). day stays. This trend suggests that increased severity June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 13

36 of cases during this period was not simply a coding is sensitive to type of readmission measure used. All-cause phenomenon, but a real increase in patient complexity. readmissions include all returns (except transfers) to the hospital after a qualifying initial admission. CMS initially 2014 to 2016— • RAC pressure was reduced in 2014. used this approach to track readmissions; however, under While there continued to be a material decline in the legislation establishing the HRRP, CMS was not medical admissions per capita during this period, supposed to count readmissions that were “unrelated to the share of cases that were one-day stays actually the prior discharge (such as a planned readmission or increased slightly in 2016. The faster decline in risk- transfer to another applicable hospital).” As a result, CMS adjusted readmission rates in 2016 (compared with developed the planned readmission algorithm, which earlier years) could in part be because the share of was implemented in the second year of the HRRP. The admissions that needed to stay only one day was no planned readmission algorithm eliminates readmissions longer falling in 2016. for transplants, maintenance chemotherapy, rehabilitation, and a set of 59 surgical procedures that are generally Given the uncertainty about how much of the changes considered planned. However, if the surgical procedures in risk-adjusted readmissions was due to coding, we are accompanied by a selected set of medical diagnoses conducted cross-sectional analyses in addition to the as the principal discharge diagnosis, the readmission is time series analyses to determine whether hospital-level considered unplanned. We find that only about 5 percent differences in readmission rates over time were related of readmissions are removed with the planned-readmission to some combination of hospital-level differences in algorithm. rates of observation stays, ED use, and mortality after discharge. As a cross-check on the robustness of the unplanned- readmission methodology, we compared trends in Our categorical risk adjustment model unplanned readmissions with an alternative metric of We used a categorical risk adjustment model based on one potentially preventable readmissions (PPRs) developed developed by 3M and used by the Agency for Healthcare by 3M. The PPR methodology captures readmissions Research and Quality to risk adjust mortality rates. In that were clinically related to the prior admission and for our model, we calculated an expected rate of readmission which there is a reasonable expectation that it could have for a group based on the group’s three-year historical been prevented. average (e.g., 2010 to 2012) rate of readmission. Cases were grouped by base all-patient refined–diagnosis related group (APR–DRG), severity of illness, age, sex, and Results mental health diagnosis (if any). These classifications allowed us to examine the average rate of readmission for each category—for example, a male age 75 to 84 in base Raw all-cause, unplanned, and potentially preventable measures of readmissions all APR–DRG 194 (heart failure) at severity of illness level show similar rates of decline 3 with no mental health diagnosis. A clinical categorical model is similar to a regression in its approach but with The trends in the raw all-cause, unplanned, and potentially many more interaction terms. For any given base DRG, preventable readmission rates from 2008 to 2016 were readmission rates increase with patient severity (and in similar, although the magnitudes differed (Figure 1-2). general increase with age), are higher for men, and are The unplanned readmission rate was slightly lower than higher again if the patient has a mental health diagnosis. the all-cause rate, which is as expected since the number To get a reasonably reliable average readmission rate of exclusions for planned surgeries is relatively small. for each category, we required at least 25 cases in each The PPR rate was about 5 percentage points lower than category. (See online Appendix 1-B, available at http:// the unplanned readmission rate. This lower rate is the www.medpac.gov, for more details on risk adjustment result of counting only clinically related readmissions that methods). are potentially preventable and not counting subsequent 5 readmissions that are part of a readmission chain. Over Unplanned versus potentially preventable the period examined, the basic trend lines for all of these measures of readmissions were similar, suggesting In our analysis we examined three types of readmissions: that using either of the two risk-adjusted measures of all-cause, unplanned, and potentially preventable. We did readmissions would yield similar results. Historically, the this to examine whether the rate of change in readmissions Mandated report: The effects of the Hospital Readmissions Reduction Program 14

37 FIGURE Title here... 1-XX FIGURE Trends in raw rates of readmission across all conditions, 2008–2016 1–2 20 17.9 17.9 17.7 17.7 17.3 16.8 16.5 16.6 15.8 16.7 16.8 16.7 16.4 16.8 15 15.9 15.8 15.6 15.0 11.2 11.4 11.3 10 11.2 10.9 10.6 10.6 10.5 9.7 (in percent) All cause Readmission rate Unplanned 5 Potentially preventable HRRP penalties PPACA passes No HRRP 0 2011 2010 2015 2012 2014 2016 2013 2009 2008 Note: HRRP (Hospital Readmissions Reduction Program), PPACA (Patient Protection and Affordable Care Act of 2010). MedPAC analysis of Medicare claims files for Medicare fee-for-service beneficiaries ages 65 or older. Source: Commission has reported the trend in PPR rates. However, points. Across conditions not covered by the program, because our mandate is to evaluate the HRRP, we used unplanned readmissions fell at a slower rate than for Note: Note and Source are in InDesign. 6 the unplanned readmission rate for this report. Doing so HRRP conditions, from 16.3 percent to 14.9 percent. Source: allowed us to examine changes in readmissions, mortality, A comparison of pre-HRRP rates of change (2008 to and service use for the specific population of admissions 2010) with rates after the HRRP was introduced (2010 subject to the HRRP. Notes about this graph: to 2016) shows an accelerated annual rate of decline The average unplanned readmission rate was flat before in raw rates of unplanned readmission: 0.3 percentage • Data is in the datasheet. Make updates in the datasheet. the start of the HRRP (16.7 percent in 2008 and in 2010). point faster on average after 2010 when the Act was • I deleted the years from the x-axis and put in my own. After the HRRP passed in 2010, the raw unplanned passed. Specifically, we examined readmission rates readmission rate declined by an average of 0.15 percentage for all inpatient prospective payment system hospitals • I had to manually draw tick marks and axis lines because they kept resetting when I changed any data. point per year from 2010 and 2012. After the penalties with available data from 2010 to 2016. For the 1,819 • The dashed line looked ok here, so I didn’t hand draw it. started to take place in 2013, the rate declined by 0.35 hospitals with more than 50 heart failure discharges in • I can’t delete the legend, so I’ll just have to crop it out in InDesign. percentage point per year on average. 2008, 2010, and 2016, the rate of decline in heart failure readmissions was faster after 2010. The difference is • Use direct selection tool to select items for modification. Otherwise if you use the black selection tool, they will reset to graph In the 2010 to 2016 period, raw rates of readmission fell statistically significant using a Wilcoxon signed-rank test default when you change the data. for each condition covered by the HRRP (Figure 1-3, p. ( < 0.01). Similarly, for the 2,270 hospitals with more p • Use paragraph styles (and object styles) to format. 16). Of the conditions initially included in the HRRP, than 50 pneumonia discharges, the rate was faster after AMI saw the largest decline in raw rates of readmission < 0.01). For the 946 hospitals with more than 50 2010 ( p during that period, falling from 19.0 percent to 15.4 AMI discharges in each year, the rate of decline in AMI percent. Readmission rates for heart failure also declined p = 0.03). readmissions was also greater after 2010 ( substantially, falling from 23.6 percent to 20.6 percent. Pneumonia, the third condition initially covered by the However, raw rates of readmission are not fully illustrative HRRP, also saw a sizable decline, falling 2.3 percentage because the mix of cases admitted to hospitals has June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 15

38 FIGURE Title here... 1-XX FIGURE Raw unplanned readmission rates for conditions covered by the HRRP, 2008–2016 1–3 30 HRRP penalties No HRRP PPACA passes 23.8 23.6 25 22.7 21.9 21.7 21.3 21.1 20.6 21.0 20.6 19.8 19.8 19.5 19.7 20 19.0 19.1 18.0 17.2 17.1 16.8 18.7 16.4 18.6 17.9 17.2 16.5 16.4 15.4 16.3 15 16.1 16.3 15.7 15.7 15.6 14.9 (in percent) Pneumonia Heart failure 10 Readmission rate COPD Non-HRRP conditions AMI Hip and knee replacement 5 5.5 5.6 5.8 5.2 5.4 4.8 4.4 4.4 4.1 0 2014 2013 2012 2011 2010 2009 2008 2016 2015 Note: HRRP (Hospital Readmissions Reduction Program), PPACA (Patient Protection and Affordable Care Act of 2010), COPD (chronic obstructive pulmonary disease), AMI (acute myocardial infarction). The pneumonia measure reflects the expanded definition used starting in fiscal year 2016, which includes simple pneumonia, aspiration pneumonia, and sepsis with pneumonia as a secondary diagnosis. Source: MedPAC analysis of Medicare claims files for Medicare fee-for-service beneficiaries ages 65 or older. changed. From 2010 to 2016, Medicare admissions The decline in risk-adjusted readmission rates was Note: Note and Source are in InDesign. per capita fell by 17 percent, suggesting that the easier steeper than that in raw readmission rates. Between 2010 Source: cases were no longer being treated on an inpatient and 2016, across non-HRRP conditions, risk-adjusted basis. Admission rates for the three HRRP-covered readmissions fell 2.6 percentage points, from 16.8 percent conditions also declined substantially: Per capita heart to 14.2 percent. As shown in Figure 1-4, the declines Notes about this graph: failure admission rates dropped 14 percent, per capita were even greater for the HRRP-covered conditions as of • Data is in the datasheet. Make updates in the datasheet. pneumonia rates fell 11 percent, and per capita AMI rates 2010: heart failure (3.9 percent), AMI (3.7 percent), and • I deleted the years from the x-axis and put in my own. declined 9 percent. A number of factors contributed to this pneumonia (3.0 percent). Even the rate of readmissions decline in inpatient admissions, including technological for hip and knee replacements, which was already low, fell • I had to manually draw tick marks and axis lines because they kept resetting when I changed any data. improvements, general practice pattern changes, 1.4 percentage points. The trends in raw readmission rates • The dashed line looked ok here, so I didn’t hand draw it. accountable care organizations (ACOs), the impact of and the trend in risk-adjusted readmission rates suggest • I can’t delete the legend, so I’ll just have to crop it out in InDesign. RACs denying the necessity of certain admissions, and that the HRRP helped to contribute to the reduced hospital the “two-midnight” rule that discouraged short-stay readmission rates. • Use direct selection tool to select items for modification. Otherwise if you use the black selection tool, they will reset to graph 7 admissions. Many of these policies occurred concurrently default when you change the data. On average, across HRRP conditions, the rate of decline with implementation of the HRRP. in the risk-adjusted readmission rates was faster after • Use paragraph styles (and object styles) to format. the program’s passage (2010 to 2016) than in the earlier The steep decline in admission rates underscores the period (2008 to 2010) by about 0.2 percentage point per importance of adjusting for the change in mix of patients year. The decline in heart failure readmissions was steeper because those admitted after the more restrictive policies after 2010 and was statistically significant when measuring would generally have a higher severity of illness with the percentage point change or percentage change in heart a greater likelihood of being readmitted (that is, higher failure readmission rates. expected readmission rates). Mandated report: The effects of the Hospital Readmissions Reduction Program 16

39 FIGURE Title here... 1-XX FIGURE Risk-adjusted changes in unplanned readmission rates by condition, 2008–2016 1–4 30 No HRRP PPACA passes HRRP penalties 24.8 24.0 25 22.7 21.9 21.9 21.3 21.1 21.4 20.1 20.5 20.0 19.6 19.1 19.2 20 19.0 18.2 18.0 19.7 17.4 17.0 19.0 16.7 18.1 16.0 17.4 17.5 16.5 16.5 16.8 15 16.0 15.5 15.4 15.0 15.0 14.2 (in percent) Pneumonia Heart failure 10 Readmission rate COPD Non-HRRP conditions AMI Hip and knee replacement 5 5.6 6.0 5.7 5.2 5.3 4.8 4.5 4.5 4.2 0 2016 2008 2009 2011 2012 2010 2013 2015 2014 HRRP (Hospital Readmissions Reduction Program), PPACA (Patient Protection and Affordable Care Act of 2010), COPD (chronic obstructive pulmonary disease), Note: AMI (acute myocardial infarction). The pneumonia measure reflects the expanded definition used starting in fiscal year 2016, which includes simple pneumonia, aspiration pneumonia, and sepsis with pneumonia as a secondary diagnosis. Source: MedPAC analysis of Medicare claims files for Medicare fee-for-service beneficiaries ages 65 or older. Pneumonia and AMI readmission rates also fell faster after An interesting finding is that raw readmission rates were Note: Note and Source are in InDesign. 2010, but the tests for statistical significance were mixed. generally not declining before 2010, but risk-adjusted Source: For pneumonia, if we measure the change in percentage rates were declining. The difference could be in part points, the readmission rate fell faster after 2010, but due to increasing complexity of patients, but another = 0.12). p not by a statistically significant amount ( possibility is that the introduction of MS–DRGs in 2008 Notes about this graph: However, it was harder to achieve the same percentage affected the rates. The MS–DRGs may have caused • Data is in the datasheet. Make updates in the datasheet. point reduction in readmissions in later years because of greater increases in coding during the years immediately • I deleted the years from the x-axis and put in my own. declining readmission rates. Therefore, we also measured after their introduction (2008 to 2010), which in turn may the percentage change in the rate of decline in readmission have resulted in overstating the decrease in risk-adjusted • I had to manually draw tick marks and axis lines because they kept resetting when I changed any data. rates (as opposed to percentage point change). Using this readmission rates during these years (Centers for Medicare • The dashed line looked ok here, so I didn’t hand draw it. percentage change method, pneumonia readmission rates & Medicaid Services 2012, Medicare Payment Advisory • I can’t delete the legend, so I’ll just have to crop it out in InDesign. fell faster after 2010 by a statistically significant amount Commission 2011a). By 2010, the effect of the new MS– p ( < 0.01). Similarly, the risk-adjusted readmission rate DRGs on coding had largely been built into the system, • Use direct selection tool to select items for modification. Otherwise if you use the black selection tool, they will reset to graph for AMI declined 0.2 percentage point per year faster which could explain the slower growth of coding from default when you change the data. after 2010 on average. The difference is not statistically 2010 onward. • Use paragraph styles (and object styles) to format. significant when measuring change in percentage points ( p The decline in readmission rates reflects = 0.09) but is significant when measuring the percentage more than coding changes change ( p < 0.01). Therefore, while readmission rates for AMI and pneumonia were falling more rapidly after 2010, To gain some insight into the degree that coding changes the difference is statistically significant only when using affected risk-adjusted rates, we examined raw and risk- one of the two methods of statistical testing. adjusted readmission rates for AMI. AMI readmissions are less discretionary than pneumonia or heart failure June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 17

40 FIGURE Title here... 1-XX FIGURE Reduction in AMI readmission rates was not driven by coding, 2008–2016 1–5 25 20.0 19.7 19.2 20 18.6 18.2 17.4 19.7 19.4 16.5 16.5 19.0 18.9 18.0 15.5 17.2 15 16.5 16.4 15.4 10 (in percent) Risk-adjusted AMI Readmission rate Raw AMI 5 HRRP penalties PPACA passes No HRRP 0 2014 2015 2013 2010 2008 2009 2011 2012 2016 Note: AMI (acute myocardial infarction), HRRP (Hospital Readmissions Reduction Program), PPACA (Patient Protection and Affordable Care Act of 2010). Source: MedPAC analysis of Medicare claims files for Medicare fee-for-service beneficiaries ages 65 or older. readmissions. For example, a readmission that was conditions (AMI, heart failure, and pneumonia) was preceded by a test indicating an ST-elevation myocardial smaller than the decline observed across conditions not Note: Note and Source are in InDesign. infarction would be seen as less discretionary and less covered by the HRRP. If hospitals were avoiding admitting likely to be challenged by the RAC. Therefore, unless patients in these conditions to potentially circumvent Source: coding changed, we would expect profiles of AMI- readmission penalties, we would have observed a admitted patients to change relatively little, causing larger reduction for the HRRP-covered conditions. The Notes about this graph: raw and risk-adjusted readmission rates to be similar. combined effect of falling admission rates and decline in However, if coding had driven the change in risk profile, readmission rates meant the number of readmissions per • Data is in the datasheet. Make updates in the datasheet. we would have expected even AMI raw and risk-adjusted Medicare beneficiary (across all beneficiaries) declined by • I deleted the years from the x-axis and put in my own. readmission rates to diverge. more than 20 percent on average. For example, from 2010 • I had to manually draw tick marks and axis lines because they kept resetting when I changed any data. to 2016, heart failure admissions declined by 14.4 percent In fact, the risk-adjusted and raw rates for AMI tracked per capita. Among this smaller number of admissions, • The dashed line looked ok here, so I didn’t hand draw it. closely after 2010 (Figure 1-5). For this reason, we readmissions fell by 16.2 percent. The combined effect of • I can’t delete the legend, so I’ll just have to crop it out in InDesign. contend that the increased risk profile in other conditions, fewer admissions and fewer readmissions per admission such as heart failure and pneumonia, at least partially • Use direct selection tool to select items for modification. Otherwise if you use the black selection tool, they will reset to graph was a 25.3 percent reduction in heart failure readmissions reflects true differences in the characteristics of admitted default when you change the data. per capita (Figure 1-6). patients from 2010 through 2016. • Use paragraph styles (and object styles) to format. Admission rates declined while observation stays Because the reported characteristics of inpatient and emergency department visits increased admissions have changed, we also examined changes in Along with the drop in admission rates, the Medicare admissions per capita. For all conditions other than hip program has seen a steady rise in beneficiaries’ use of and knee replacements, admissions per capita between observation stays and EDs (Figure 1-7). These trends in 2010 and 2016 declined (Figure 1-6). Interestingly, the rising observation and ED use started before the HRRP fall in admission rates for the three initial HRRP-covered was implemented. Mandated report: The effects of the Hospital Readmissions Reduction Program 18

41 FIGURE Cumulative change... FIGURE 1-X Percent change in per capita admission and readmission rates, 2010–2016 1–6 15 9.2 Risk-adjusted unplanned readmissions Per capita admissions Per capita readmissions 10 5 0 –5 –10 –8.6 –11.2 –15 –14.4 –15.1 –15.6 –15.9 Percent change –16.2 –17.8 –20 –19.3 –19.0 –22.1 –25 –24.3 –25.7 –25.3 –25.8 –25.9 –30 –32.6 –35 COPD Pneumonia Heart failure AMI Non-HRRP conditions Hip and knee Note: HRRP (Hospital Readmissions Reduction Program), AMI (acute myocardial infarction), COPD (chronic obstructive pulmonary disease). Pneumonia measure reflects the expanded definition used starting in fiscal year 2016, which includes simple pneumonia, aspiration pneumonia, and sepsis with pneumonia as a secondary diagnosis. Source: MedPAC analysis of Medicare claims files for Medicare fee-for-service beneficiaries ages 65 or older. FIGURE Title here... 1-XX FIGURE Per capita admission rates fell, while observation and ED use increased, 2008–2016 1–7 0.45 0.405 0.401 0.385 0.374 0.40 0.375 0.358 0.351 0.339 0.332 0.35 0.30 0.328 0.316 0.317 0.309 0.290 0.280 0.25 0.269 0.267 0.262 Per capita ED use 0.20 Per capita initial admissions Per capita observation 0.15 (age 65 or older) Events per FFS beneficiary HRRP penalties No HRRP PPACA passes 0.10 0.049 0.047 0.045 0.043 0.039 0.034 0.05 0.030 0.026 0.022 0.00 2016 2014 2013 2012 2011 2015 2010 2009 2008 Note: Note and Source are in InDesign. Note: ED (emergency department), FFS (fee-for-service), HRRP (Hospital Readmissions Reduction Program), PPACA (Patient Protection and Affordable Care Act of 2010). Source: MedPAC analysis of Medicare claims files for Medicare fee-for-service beneficiaries ages 65 or older. Source: Report to the Congress: Medicare and the Health Care Delivery System June 2018 | 19 Note: Note and Source are in InDesign. Source: Notes about this graph: • Data is in the datasheet. Make updates in the datasheet. • I deleted the years from the x-axis and put in my own. • I had to manually draw tick marks and axis lines because they kept resetting when I changed any data. • The dashed line looked ok here, so I didn’t hand draw it. • I can’t delete the legend, so I’ll just have to crop it out in InDesign. • Use direct selection tool to select items for modification. Otherwise if you use the black selection tool, they will reset to graph default when you change the data. • Use paragraph styles (and object styles) to format.

42 FIGURE Title here... 1-XX FIGURE Readmissions per discharge declined as observations and ED visits increased, 2008–2016 1–8 20 17.8 17.5 17.1 16.3 16.4 15.7 15.3 15.2 14.4 15 Risk-adjusted unplanned readmission rate ED visit 8+ hour observation 10 7.3 6.8 6.9 7.0 (in percent) 6.7 5.2 4.9 5.1 5.0 5 Rate of return to the hospital HRRP penalties PPACA passes No HRRP 1.7 1.5 1.3 1.5 1.2 0.8 0.6 0.7 0.5 0 2013 2014 2015 2016 2010 2008 2009 2012 2011 Note: ED (emergency department), HRRP (Hospital Readmissions Reduction Program), PPACA (Patient Protection and Affordable Care Act of 2010). Source: MedPAC analysis of Medicare claims files for Medicare fee-for-service beneficiaries ages 65 or older. Between 2010 and 2016, per capita admission rates of the average inpatient admission and the cost of an ED (initial admission for qualifying conditions) dropped 17 visit is about 5 percent of the cost of an inpatient stay. Note: Note and Source are in InDesign. percent, from 0.316 per capita to 0.262 per capita (Figure Observation and ED use increased for both those 1-7, p. 19). At the same time, however, per capita use of Source: admitted and those not admitted to the hospital observation care grew 63 percent, from 0.030 per capita to 0.049 per capita. The increase in observation stays may As readmission rates declined, use of observation and the Notes about this graph: have been partially a response to the RAC audits and two- ED after inpatient stays increased (Figure 1-8). The largest midnight rule implemented by CMS during this period. increases occurred in 2012, two years after the HRRP was • Data is in the datasheet. Make updates in the datasheet. Because observation stays increased steadily from 2008 to passed. From 2012 on, the increases in observation and • I deleted the years from the x-axis and put in my own. 2016 (including the period before the RAC incentive), we ED use have been more modest. • I had to manually draw tick marks and axis lines because they kept resetting when I changed any data. expect that more than the RAC incentive was at work. Per In 2010, for beneficiaries who were not readmitted, about capita use of ED between 2010 and 2016 also increased, • The dashed line looked ok here, so I didn’t hand draw it. 0.7 percent of cases were followed by an eight-hour or rising 15 percent, from 0.351 visits per capita to 0.405. • I can’t delete the legend, so I’ll just have to crop it out in InDesign. longer observation stay. By 2016, 1.7 percent of cases Almost half of this increase took place from 2010 to 2012. were followed by an eight-hour or longer observation stay. • Use direct selection tool to select items for modification. Otherwise if you use the black selection tool, they will reset to graph The joint timing of a decline in inpatient admissions with In that same period, ED use increased from 5.1 percent of default when you change the data. an increase in observation stays and ED visits suggests that cases to 7.3 percent of cases with a prior admission. there was some substituting of outpatient care for inpatient • Use paragraph styles (and object styles) to format. care. From the Medicare patients’ perspective, patients However, this ED growth appears to be broad based and may prefer avoiding a hospital stay if they can achieve an not focused on ED visits after discharge. We compared equal or better outcome in an outpatient setting. From the Medicare beneficiaries’ ED-visit growth for those Medicare program’s financial perspective, avoiding an discharged from a hospital and those not discharged from inpatient stay helps to reduce program spending since the a hospital within the prior 30 days. We found that the cost of an observation stay is about 20 percent of the cost Mandated report: The effects of the Hospital Readmissions Reduction Program 20

43 TABLE Change in risk-adjusted rate of return to the hospital 1–2 for non-HRRP and HRRP conditions, 2010–2016 Percentage point change in the share of patients returning to the hospital within 30 days categorized as: Readmission, observation, Readmission or ED visit Readmissions or observation 0.6 –1.5 –2.5 Non-HRRP admissions HRRP conditions 0.1 –3.1 –2.1 AMI –3.7 –2.0 0.6 –0.7 –3.9 –2.7 Heart failure –2.1 Pneumonia –3.0 –0.1 0.0 COPD –3.3 –2.2 Hip or knee replacement –1.4 –1.1 1.0 HRRP (Hospital Readmissions Reduction Program), ED (emergency department), AMI (acute myocardial infarction), COPD (chronic obstructive pulmonary disease). Note: MedPAC analysis of Medicare claims files for Medicare fee-for-service beneficiaries ages 65 or older. Source: growth rates were similar for the two groups. In fact, the discharge increased by 0.1 percentage point for HRRP shares of ED visits that were postdischarge visits in 2010 conditions and 0.6 point for other conditions. From 2010 and 2016 were exactly the same (4.8 percent, data not to 2016, the change in rates of return to the hospital varied shown). Therefore, it appears that the growth in ED use is by HRRP condition: AMI returns to the hospital rose, and a broad phenomenon and cannot be primarily attributed to heart failure returns to the hospital fell. growth in postdischarge ED visits. While the increase in ED and observation use coincided The decline in readmission rates coincided with increases with the decline in readmissions, we cannot conclude in the rate of observation and the rate of ED use (Figure that the decrease in readmissions caused the increase 1-8). The next question is: To what degree did the in observation visits or ED use. Observation and ED increase in observation and ED use offset the decline in use increased for all Medicare beneficiaries in this time readmissions? Looking across conditions not covered period, not just for those who were admitted to the by the program, we see a 2.5 percentage point reduction hospital. For example, from 2010 to 2016, the share of in readmissions (Table 1-2). For those covered by the discharges that were followed by an observation visit program, the reduction averaged 3.1 percentage points. grew by 1 percentage point. At this same time, the share At the same time, observation stays rose 1.0 percentage of all observation stays (including stays by beneficiaries point and ED use within 30 days after discharge, by 2.3 never admitted to a hospital) grew by 1.9 percentage percentage points (data not shown). However, adding points, meaning observation stays grew faster for patients together 3.1 percentage point decline in readmissions who had not been admitted. Similarly, the share of and the 1.0 percentage point increase in observation stays beneficiaries with a postdischarge ED visit increased results in a combined decline of 2.1 percentage points in 2.1 percentage points, and the per capita ED use for all the sum of inpatient stays and observation stays (Table Medicare beneficiaries grew by 5.4 percentage points. 1-2). We also saw an increase in ED use within 30 days The faster growth in ED visits and observation stays for postdischarge of about 2.3 percentage points (data not those without a recent admission to the hospital allows us shown). The net effect was that the share of Medicare to conclude that the readmission policy was not likely the patients returning to the hospital for some type of care driver behind the ED and observation growth experienced. (readmission, observation, or ED) within 30 days of June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 21

44 FIGURE FIGURE Cumulative change... HRRP conditions have unusually large declines in readmissions without 1–9 1-X unusually large increases in observation or ED stays, 2010–2016 3 2.1 2.1 2 1.0 1.1 1 0.6 0.1 0 –1 Risk-adjusted unplanned readmissions –2 8+ hour observation Percentage point change –2.5 ED –3 Return to hospital (sum of all three) –3.1 –4 5 HRRP conditions Conditions not included in HRRP Note: HRRP (Hospital Readmissions Reduction Program), ED (emergency department), AMI (acute myocardial infarction), COPD (chronic obstructive pulmonary disease). The five HRRP conditions include pneumonia, heart failure, AMI, COPD, and hip and knee replacement. The reasons for returning to the hospital are all measured in events per 100 initial admissions. Source: MedPAC analysis of Medicare claims files for Medicare fee-for-service beneficiaries ages 65 or older. Greater readmission declines among HRRP Small correlations were found between reductions in readmissions and increases in observation and conditions did not trigger greater observation-stay or ED-visit growth ED visits Risk-adjusted readmission rates fell more for conditions In addition to looking at national trends, we examined the data on readmissions, observation stays, and ED visits at covered by the HRRP (i.e., 3.1 percentage points for HRRP conditions versus 2.5 percentage points for non- the hospital level to determine whether the hospitals with 8 the biggest declines in readmissions also had unusually HRRP conditions) (Figure 1-9). The difference is modest, large increases in observation and ED use rates. In this suggesting there may be some spillover of behavior from analysis, we found a small negative correlation coefficient HRRP conditions to non-HRRP conditions. Nevertheless, (−0.13) between changes in readmission rates and changes a hospital-level analysis indicates that the difference in postdischarge observation use. Similarly, adding between 3.1 and 2.5 percentage points is statistically changes in observation use and ED visits together, we < 0.01). p significant using a Wilcoxon signed-rank test ( found a small negative correlation coefficient (−0.19) with While the HRRP conditions had bigger reductions in changes in readmission rates, suggesting that hospitals readmissions, changes in observation and ED visits with above-average declines in readmissions did tend were almost exactly the same for HRRP conditions and to have increases in observation and ED use. However, other conditions (Figure 1-9). If hospitals were using taken together with the data in Figure 1-6 (p. 19) and the observation and ED visits to avoid readmission penalties, national growth rates in observation and ED use for those we would expect to see larger increases for conditions a recent admission, the data suggest that only a without covered by the program, but we did not. Because the Note: Note and Source are in InDesign. small share of the increase in observation and ED use was greater reduction in readmission rates did not trigger a Source: related to the HRRP. greater growth in observation stays and ED visits, the Mandated report: The effects of the Hospital Readmissions Reduction Program 22

45 observation and ED-use growth does not appear to be readmission rate for HRRP conditions declined by 16.4 primarily a function of declining readmission rates. percent compared with an 8.3 percent decline for non- HRRP conditions. The difference in the rate of decline of Sensitivity of findings to different methods of < 0.01). p raw readmission rates is statistically significant ( statistical testing Next, we examined whether the decline in raw Our congressional mandate is to examine whether readmission rates was offset by a raw increase in ED reductions in readmissions caused an offsetting increase visits or observation stays. As with the risk-adjusted in ED visits and observation stays. Therefore, the method model, the raw change in postdischarge ED use and to test for offsetting increases in observation stays and ED observation was almost identical for HRRP conditions visits needed to use a unit of analysis that is comparable and non-HRRP conditions. A version of Figure 1-9 across readmissions, ED visits, and observation stays. using raw (not risk-adjusted) data is provided in online As a result, Figure 1-9 presents data that are measured as Appendix 1-C, available at http://www.medpac.gov. events per 100 stays. The changes in events are equivalent Looking at the totality of the different tests, there to percentage point changes in readmission, ED-use, and is fairly consistent evidence that the HRRP caused observation rates. When we tested for percentage point some reduction in readmissions, with most of the tests differences between HRRP and non-HRRP conditions, showing statistical significance. However, given the we found that the difference (3.1 percentage points vs. 2.5 differences in the magnitude of the effects across different percentage points) is statistically significant. methods of testing for a HRRP effect on readmissions, it is not possible to say what portion of the reduction in However, there is a question of how robust the two readmissions was due to the HRRP and what portion was findings (that readmission rates for HRRP conditions due to other concurrent factors such as ACOs or changes fell faster than for non-HRRP conditions and the finding 10 in coding practices. that the readmission reductions did not trigger large increases in ED visits and observation stays) are to Medicare program costs declined as different methodological approaches. Therefore, we readmissions declined first estimated whether the difference in rate of decline for HRRP conditions and non-HRRP conditions would The Medicare program’s savings from the drop in be statistically significant if we measured change in readmissions was much greater than the increase in percentage rather than percentage points. Second, we payments for the additional observation stays and ED investigated whether the finding—that greater readmission visits. As shown in Table 1-3 (p. 24), the program declines for HRRP conditions did not trigger more ED or spent $2.28 billion less on readmissions in 2016 than observation stays—held for raw (not risk-adjusted) data. it would have if readmissions had occurred at the same rate as in 2010. Even though use of observation and ED When using percentage changes rather than percentage visits increased, the effect had a relatively small impact point changes, we found that the risk-adjusted readmission on spending, with observation spending increasing rate for HRRP conditions declined by 16.9 percent from postdischarge by $170 million and ED spending 2010 to 2016 compared with a 15.1 percent decline for increasing by $70 million. The net reduction in spending non-HRRP conditions. The difference (1.8 percent) is on readmissions was $2.04 billion. While it is clear 9 modest and statistically significant. The fact that the that readmission spending was reduced, it is not clear percentage differences are modest could reflect HRRP what share of the reduction was due to the HRRP. Other incentives spilling over into other conditions, coding factors such as ACOs or technological changes may have difference across conditions, and other factors outside contributed to the reduction in readmission rates. of the HRRP such as ACOs’ practices also affecting readmission rates (Winblad et al. 2017). Changes in mortality rates and readmissions rates are not highly correlated One concern is that the difference (1.8 percent) may be We also examined whether there was any relationship due to greater coding changes for conditions covered between changes in readmissions and changes in mortality, by the HRRP. We found no evidence of this concern using two prior studies cited earlier as a starting point. given that the difference in the change in raw rates Both studies found a slight increase in risk-adjusted heart of readmission for HRRP and non-HRRP conditions was larger than the risk-adjusted differences. The raw June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 23

46 TABLE Changes in costs after the 2010 enactment of the HRRP, 2010–2016 1–3 Change in the cost of return visits to the hospital (in billions) Type of care Readmissions within 30 days of the initial admission $−2.28 Observation stays, initial and postdischarge 0.17 0.07 Emergency department visits (without admission), initial and postdischarge −2.04 Net change in spending Note: HRRP (Hospital Readmissions Reduction Program). Reductions in spending on readmissions, observation stays, and emergency use pertain to reductions for all conditions including those not covered by the HRRP. It is not clear the degree to which these reductions are due to the HRRP or other factors. MedPAC analysis of Medicare claims data. Source: failure mortality from 2010 through 2014. One study, a portion of the decline in mortality is real. We believe which conducted a hospital-level analysis regarding the the decline in risk adjusted rates is real due to consistent relationship between readmissions and increased mortality, evidence that the patients admitted in 2016 had a higher did not find a correlation (Dharmarajan et al. 2017). We risk of mortality than the patients admitted in 2010. repeated the time-trend analysis and the hospital-level Our finding of declines in risk-adjusted heart failure correlation analyses with two changes. First, our analysis mortality rests on a finding that the patients admitted extended through 2016. Second, we measured mortality as in 2016 had a higher risk of mortality than the patients the combination of mortality during the inpatient stay and admitted in 2010. From our data, it appears that the 30 days after the stay ended. This approach addressed the large decline in initial heart failure admissions per capita problem of a possible shift in the location of the mortality caused the 2016 cohort of heart failure admissions to due to hospice use increasing during this time period (as consist of patients with higher expected mortality than discussed in online Appendix 1-B, available at http://www. the 2010 cohort of heart failure admissions. In other medpac.gov). words, it appears that hospitals were admitting fewer easy We found that, after the HRRP’s introduction, raw rates cases in 2016. While our categorical model’s finding of of mortality materially increased for one of the HRRP higher expected mortality for the 2016 cohort compared conditions (heart failure) but materially declined for two with the 2010 cohort could partially be due to greater other HRRP-covered conditions (pneumonia and AMI) coding intensity, we also find that patients admitted in (Figure 1-10). On average, raw rates of mortality declined 2016 tended to have higher risk scores based on the prior for the five HRRP conditions and increased for non-HRRP year’s diagnoses, tended to have greater intensive care conditions. The increase in raw rates of mortality for heart unit use, and were less likely to be discharged home for 11 failure and non-HRRP conditions may have been related self-care than the cohort of patients admitted in 2010. to the decline in initial admissions and increases in the These factors all suggest that the 2016 cohort of heart severity level of those admitted. The literature has tended failure admissions were less healthy than the 2010 cohort. to focus on the one metric where mortality increased (heart Therefore, our findings of both increasing raw rates of failure) rather than the conditions for which mortality mortality and declining risk-adjusted mortality for heart declined. failure admissions is plausible. These findings do not mean that no clinician ever erroneously failed to admit a patient, From 2010 to 2016, we found that risk-adjusted mortality or even that the HRRP did not affect the rate of appropriate rates during the inpatient stay and the following 30 days readmissions. It means only that, on net, care continued declined for all conditions (Figure 1-11). While greater to improve during the time the HRRP was in effect. This coding intensity over time may be responsible for some improvement could indicate that the positive effects of of the decline in risk-adjusted mortality, we believe that Mandated report: The effects of the Hospital Readmissions Reduction Program 24

47 FIGURE Title here... FIGURE 1-XX Raw 30-day mortality (in-hospital through 30 days postdischarge) rates 1–10 have risen for some and fallen for other conditions covered by the HRRP 20 18.8 18.9 19.0 18.3 17.8 17.8 17.5 15.4 16.9 16.9 14.9 14.4 14.4 15 13.6 13.4 13.0 12.9 12.1 12.4 12.4 12.1 11.9 11.9 11.9 11.1 11.4 11.3 10 9.0 8.4 8.5 8.4 8.5 8.3 8.1 8.0 8.2 8.3 8.6 8.5 8.4 7.9 7.7 7.4 7.3 7.3 Pneumonia Heart failure 5 COPD Non-HRRP conditions (share of patients admitted) AMI Hip and knee replacement PPACA passes No HRRP HRRP penalties 30-day postdischarge mortality rate 0.4 0.3 0.3 0.3 0.3 0.3 0.2 0.2 0.2 0 2013 2009 2015 2016 2008 2010 2011 2012 2014 Note: HRRP (Hospital Readmissions Reduction Program), COPD (chronic obstructive pulmonary disease), AMI (acute myocardial infarction), PPACA (Patient Protection and Affordable Care Act of 2010). The pneumonia measure reflects the expanded definition used starting in fiscal year 2016, which includes simple pneumonia, aspiration pneumonia, and sepsis with pneumonia as a secondary diagnosis. Source: MedPAC analysis of Medicare claims files for Medicare fee-for-service beneficiaries ages 65 or older, 2008–2016. FIGURE Title here... FIGURE 1-XX Risk-adjusted mortality (in-hospital through 30 days postdischarge) 1–11 fell for conditions covered by the HRRP 25 Note: Note and Source are in InDesign. Pneumonia Heart failure 21.7 Source: COPD Non-HRRP conditions 20.3 AMI Hip and knee replacement 19.4 20 18.3 17.3 Notes about this graph: 16.7 15.8 15.7 15.0 14.7 • Data is in the datasheet. Make updates in the datasheet. 14.6 14.2 15 13.8 13.5 13.3 12.8 • I deleted the years from the x-axis and put in my own. 12.6 11.9 13.6 12.7 12.2 • I had to manually draw tick marks and axis lines because they kept resetting when I changed any data. 11.8 9.4 11.2 11.1 10 9.0 8.6 10.5 8.1 10.2 7.7 • The dashed line looked ok here, so I didn’t hand draw it. 7.7 9.4 7.2 6.9 8.7 6.5 8.3 7.8 Risk-adjusted 30-day • I can’t delete the legend, so I’ll just have to crop it out in InDesign. 7.4 7.1 7.0 6.7 6.5 6.2 postdischarge mortality rate 5 • Use direct selection tool to select items for modification. Otherwise if you use the black selection tool, they will reset to graph No HRRP PPACA passes HRRP penalties default when you change the data. 0.4 0.4 0.3 0.3 0.3 0.2 0.2 0.3 0.2 • Use paragraph styles (and object styles) to format. 0 2016 2013 2011 2010 2009 2012 2008 2015 2014 Note: HRRP (Hospital Readmissions Reduction Program), COPD (chronic obstructive pulmonary disease), AMI (acute myocardial infarction), PPACA (Patient Protection and Affordable Care Act of 2010). The pneumonia measure reflects the expanded definition used starting in fiscal year 2016, which includes simple pneumonia, aspiration pneumonia, and sepsis with pneumonia as a secondary diagnosis. Source: MedPAC analysis of Medicare claims files for Medicare fee-for-service beneficiaries ages 65 or older, 2008–2016. June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 25 Note: Note and Source are in InDesign. Source: Notes about this graph: • Data is in the datasheet. Make updates in the datasheet. • I deleted the years from the x-axis and put in my own. • I had to manually draw tick marks and axis lines because they kept resetting when I changed any data. • The dashed line looked ok here, so I didn’t hand draw it. • I can’t delete the legend, so I’ll just have to crop it out in InDesign. • Use direct selection tool to select items for modification. Otherwise if you use the black selection tool, they will reset to graph default when you change the data. • Use paragraph styles (and object styles) to format.

48 FIGURE Title here... FIGURE 1-XX No material relationship between hospital-level changes in 1–12 heart failure readmission rates and hospital-level changes in mortality 0.4 Correlation: 0.3 Change in risk-adjusted readmission rates and change in risk-adjusted mortality rates: 0.2 0.058 0.1 0.0 x = 0.0493 – 0.022 y –0.1 2 R = 0.0034 2010–2011 to 2015–2016 –0.2 –0.3 Change in risk-adjusted mortality rates, –0.4 –0.2 –0.1 0.0 0.1 0.2 0.3 –0.5 –0.4 –0.3 Change in risk-adjusted unplanned readmission rates, 2010–2011 to 2015–2016 Source: MedPAC analysis of Medicare claims files for Medicare fee-for-service beneficiaries ages 65 or older. changes in care patterns (better prescription reconciliation, their readmission rates more than average tended to also Note: Note and Source are in InDesign. better care transitions, and better coordination with post- improve their mortality rates a bit more than average. Source: acute care providers) may have outweighed any negative While statistically significant, the magnitude of the changes in care patterns. Our finding for heart failure correlation is small (0.058). The correlations for mortality differs from the earlier two studies, which found a slight and readmissions for the other four HRRP conditions are Notes about this graph: increase in risk-adjusted heart failure rates. It could be due also small (and positive), but also statistically significant. • Data is in the datasheet. Make updates in the datasheet. to our combination of inpatient and post-acute mortality, Interestingly, the correlation found by the Dharmarajan differences in risk adjusters, or simply our use of two more study, which used a different measure of mortality and • I deleted the years from the x-axis and put in my own. years of data. The 2016 data may differ in that the RAC different years of data, was almost the same (0.066) • I had to manually draw tick marks and axis lines because they kept resetting when I changed any data. audits had been removed by that time. (Dharmarajan et al. 2017). • The dashed line looked ok here, so I didn’t hand draw it. Little hospital-level correlation was found between • I can’t delete the legend, so I’ll just have to crop it out in InDesign. changes in readmission rates and changes in • Use direct selection tool to select items for modification. Otherwise if you use the black selection tool, they will reset to graph Policy implications mortality default when you change the data. In addition to looking at national trends—which can be Readmission rates clearly declined from 2010 to 2016. confounded by many concurrent changes—we conducted • Use paragraph styles (and object styles) to format. Given the totality of the evidence and the findings in the a hospital-level analysis of the relationship between literature, it appears that at least some of this reduction change in readmission rates and mortality rates over was due to the incentives in the HRRP. The exact share time. As shown in Figure 1-12, we found almost no that is due to the HRRP and the share due to other factors correlation. The small correlation between changes in is difficult to disentangle. The reduction in readmission readmission rates and changes in mortality rates that we rates appears to have been achieved without an increase in did see was positive, meaning that hospitals that improved Mandated report: The effects of the Hospital Readmissions Reduction Program 26

49 risk-adjusted mortality or a material increase in spending cover all conditions, and the magnitude of the penalty on other services. While use of observation care and ED for each excess readmission could be reduced. This postdischarge increased after the HRRP was introduced, budget-neutral change would create a broader incentive for providers to reduce readmissions and would allow these increases were program wide and likely strongly the Medicare program to reduce penalties to a level that influenced by other factors such as the RAC audits and is more proportionate to the cost of excess readmissions. two-midnight policy implemented by CMS over this In addition, as we discuss in Chapter 7 of this report, the period. While the program has achieved some of its objectives, the program could still be improved. As we system of hospitals’ financial incentives could be adjusted discussed in 2013, the program could be expanded to to balance readmission, mortality, and patient experience incentives. ■ June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 27

50 Endnotes cases (e.g., trauma) and counts a sequence of readmissions The 30-day measures used for the Hospital Readmissions 1 as only one readmission. Instead of counting individual Reduction Program (HRRP) are essentially the same measure readmissions, the PPR approach counts readmission chains, as reported on the Hospital Compare website except that which are defined as sequences of one or more PPRs that readmissions to Veterans Health Administration hospitals and are all clinically related to the same initial admission. critical access hospitals are not included. A person who is In calculating PPR rates, readmission chains rather than discharged from a prospective payment system hospital and is individual readmissions are used as the numerator. For more later readmitted to a critical access hospital is not considered a information, see: https://www.cms.gov/Research-Statistics- readmission for purposes of the HRRP. Data-and-Systems/Research/HealthCareFinancingReview/ 2 The use of principal discharge diagnosis raises an issue of downloads/08Fallpg75.pdf. double counting admissions when the policy was expanded 6 We also examined the percentage change (rather than the to include chronic obstructive pulmonary disease (COPD) in percentage point change) and again found the raw rate of fiscal year 2015 and coronary artery bypass graft (CABG) decline is about twice as fast for HRRP conditions as non- surgery in fiscal year 2016. For example, many patients who HRRP conditions. are admitted to the hospital with a heart attack receive either percutaneous transluminal coronary angioplasty or a CABG The two-midnight rule specified that CMS would not target 7 during their stay. The principal discharge diagnosis for these admissions that lasted two midnights or longer for medical patients is usually AMI. Thus, these cases could be counted necessity review. In online Appendix 1-B, available at http:// under both the current AMI readmission measure and the www.medpac.gov, we illustrate how the share of heart failure CABG readmission measure. readmissions that were one-day stays fell during the period that the RAC medical necessity reviews of inpatient stays 3 During the 2008 to 2014 time frame, heart failure admissions were occurring. dropped significantly. Dharmarajan and colleagues report a decline in their data set of roughly 16 percent, from 449,135 The differences in raw rates were even larger (2.9 percentage 8 to 385,222 (Dharmarajan et al. 2017). The cases that may points for HRRP conditions and 1.3 percentage point for non- have continued to be admitted may have been the more HRRP conditions). difficult cases. It is not clear that the risk adjuster would have fully accounted for changes in case mix over time. This issue The percentage decline in readmission rates was larger for 9 is discussed further in the online Appendix 1-B, available at HRRP conditions using a Wilcoxon signed-rank test ( p < http://www.medpac.gov. 0.01). Gupta and colleagues found that risk-adjusted mortality rates 4 Ibrahim and colleagues suggested that about two-thirds of 10 increased faster than raw mortality rates. This combination of the reduction in risk-adjusted readmission rates was due to findings implies that the expected rate of mortality decreased, patients being coded as being more severely ill, which could meaning the post-HRRP group had a lower risk of death than reflect true changes in the severity of illness among admitted the pre-HRRP group. The conclusion that the post-HRRP patients or changes in coding practices (Ibrahim et al. 2017). group had a lower expected one-year mortality is difficult to Only about one-third of the change in readmission rates was reconcile with the descriptive statistics stating that the post- not related to coded severity. The study was conducted by HRRP group had an older mean age (80.9 years compared comparing inpatient prospective payment system hospitals to with 80.1 years), was more likely to have had a previous critical access hospitals that were not affected by the HRRP. stroke/transient ischemic attack (17.2 percent compared with As we discuss in online Appendix 1-B, available at http:// 15.6 percent), was more likely to be discharged to hospice www.medpac.gov, critical access hospitals are an imperfect (4.6 percent compared with 2.5 percent), and less likely to comparison group, and the share of the readmissions be discharged to home (63.7 percent compared with 69.0 reduction caused by the more intensive coding practices percent) (Gupta et al. 2017). However, the post-HRRP group cannot be precisely estimated. The fact that we also found did have a shorter length of stay: 4.8 days versus 5.4 days for large changes for raw readmission rates suggests that the the pre-HRRP group. It is not clear what factors in the Gupta effect of coding practices may have been modest (see online model led to the post-HRRP group being assigned a lower Appendix 1-B, available at http://www.medpac.gov). one-year mortality risk. 11 The 3M risk of mortality measure we use to examine expected While the rates of change in the PPR and HRRP methods 5 mortality indicates that the severity of illness of heart failure are similar, the rates of readmission in the PPR program are patients increased from 2010 to 2016. However, some of this generally lower because the PPR methodology excludes more Mandated report: The effects of the Hospital Readmissions Reduction Program 28

51 increase may be due to greater coding rather than truly greater also examined discharge destination as another indicator of health that is not dependent on coding. We saw the share health needs. Therefore, we also examined indicators that are of heart failure patients that were discharged home for self- based on patient conditions before admission and indicators that are not dependent on coding. We found that the average care decreased by 3.9 percentage points from 46.5 percent hierarchical condition category score for patients admitted of patients in 2010 to 42.6 percent of patients in 2016. In contrast, the share discharged to hospice increased by 1.1 with heart failure increased from 2.74 in 2010 to 2.88 in 2012 and to 3.06 in 2016. This means that the diagnoses codes percentage points; the share of those discharged to home with and other factors from the year before admission indicated home health care increased by 1.8 percentage points; the share discharged to an inpatient rehabilitation facility increased by that 2016 cohort of heart failure patients had higher expected 0.8 percentage points; and the share discharged to a skilled annual healthcare costs (relative to the national average for that year) than the 2010 cohort. To examine factors unrelated nursing facility increased by 0.5 percentage points. The coding-based indicators of health and the indicators that are to coding, we also examined intensive care unit use. We found the share of heart failure admissions with one or more not dependent on coding both point toward the 2016 cohort of patients being less healthy than the 2010 cohort. days in the intensive care unit increased from 34.3 percent in 2010 to 35.5 percent in 2012 to 35.8 percent in 2016. We June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 29

52 References Himmelstein, D., and S. Woolhandler. 2015. Quality Birmingham, L. E., and W. H. Oglesby. 2018. Readmission rates improvement: ‘Become good at cheating and you never need to in not-for-profit vs. proprietary hospitals before and after the blog, August 27. Health Affairs become good at anything else’. BMC Hospital Readmission Reduction Program implementation. http://healthaffairs.org/blog/2015/08/27/quality-improvement- 18, no. 1 (January 19): 31. Health Services Research become-good-at-cheating-and-you-never-need-to-become-good- Cary, M. P., Jr., V. Goode, N. Crego, et al. 2018. Hospital at-anything-else/. readmissions among total hip replacement patients in 2009 and Hu, J., M. D. Gonsahn, and D. R. Nerenz. 2014. Socioeconomic Archives of Physical Medicine and Rehabilitation (January 2014. status and readmissions: Evidence from an urban teaching 30). 33, no. 5 (May): 778–785. Health Affairs hospital. Centers for Medicare & Medicaid Services, Department of Health Hu, J., A. J. H. Kind, and D. Nerenz. 2018. Area Deprivation and Human Services. 2012. Medicare program; hospital inpatient Index predicts readmission risk at an urban teaching hospital. prospective payment systems for acute care hospitals and the (January 1): 1–9. American Journal of Medical Quality long-term care hospital prospective payment system and fiscal year 2013 rates; hospitals’ resident caps for graduate medical Ibrahim, A. M., J. B. Dimick, S. S. Sinha, et al. 2017. education payment purposes; quality reporting requirements for Association of coded severity with readmission reduction after specific providers and for ambulatory surgical centers. Proposed JAMA Internal the Hospital Readmissions Reduction Program. Federal Register rule. 77, no. 92 (May 11): 27870–28192. Medicine (November 13). https://jamanetwork.com/journals/ jamainternalmedicine/fullarticle/2663252. Chatterjee, P., and K. E. Joynt Maddox. 2018. U.S. national trends in mortality from acute myocardial infarction and heart failure Jack, B. W., V. K. Chetty, D. Anthony, et al. 2009. A reengineered policy: Success or failure? JAMA Cardiology (Published online hospital discharge program to decrease rehospitalization: March 14). https://jamanetwork.com/journals/jamacardiology/ 150, no. 3 Annals of Internal Medicine A randomized trial. article-abstract/2674715?redirect=true. (February 3): 178–187. Chen, M., and D. C. Grabowski. 2017. Hospital Readmissions Jencks, S. F., M. V. Williams, and E. A. Coleman. 2009. Medical Reduction Program: Intended and unintended effects. Rehospitalizations among patients in the Medicare fee-for-service Care Research and Review (December 1): 1–18. program. 360, no. 14 (April 2): New England Journal of Medicine 1418–1428. Coleman, E. A., C. Parry, S. Chalmers, et al. 2006. The care transitions intervention: Results of a randomized controlled Joynt, K. E., J. E. Figueroa, J. Oray, et al. 2016. Opinions on the Archives of Internal Medicine trial. 166, no. 17 (September 25): Hospital Readmission Reduction Program: Results of a national 1822–1828. American Journal of Managed Care survey of hospital leaders. 22, no. 8 (August 1): e287–294. Dharmarajan, K., and H. M. Krumholz. 2017. Consequences Journal of the of reductions in hospital readmissions: Reply. Homeward bound: Nine patient-centered Kanaan, S. 2009. 318, no. 19 (November 21): 1934. American Medical Association Oakland, CA: California HealthCare programs cut readmissions. Foundation. Dharmarajan, K., Y. Wang, Z. Lin, et al. 2017. Association of changing hospital readmission rates with mortality rates after Khera, R., A. Pandey, C. R. Ayers, et al. 2017. Contemporary Journal of the American Medical Association hospital discharge. epidemiology of heart failure in fee-for-service Medicare 318, no. 3 (July 18): 270–278. beneficiaries across healthcare settings. Circulation: Heart Failure 10, no. 11 (November): 1–9. Gu, Q., L. Koenig, J. Faerberg, et al. 2014. The Medicare Hospital Readmissions Reduction Program: Potential unintended Klevens, R. M., J. R. Edwards, C. L. Richards, Jr., et al. 2007. Health consequences for hospitals serving vulnerable populations. Estimating health care-associated infections and deaths in U.S. 49, no. 3 (June): 818–837. Services Research 122, no. 2 (March–April): Public Health Reports hospitals, 2002. 160–166. Gupta, A., L. A. Allen, D. L. Bhatt, et al. 2017. Association of the Hospital Readmissions Reduction Program implementation McHugh, J. P., A. Foster, V. Mor, et al. 2017. Reducing hospital JAMA with readmission and mortality outcomes in heart failure. readmissions through preferred networks of skilled nursing Cardiology (November 12). https://jamanetwork.com/journals/ facilities. Health Affairs 36, no. 9 (September 1): 1591–1598. jamacardiology/article-abstract/2663213?redirect=true. Mandated report: The effects of the Hospital Readmissions Reduction Program 30

53 Medicare Payment Advisory Commission. 2016. Report to the QualityNet. 2017. Measure methodology reports. https://www. Washington, DC: MedPAC. Congress: Medicare payment policy. qualitynet.org/dcs/ContentServer?cid=1219069855841&pagenam e=QnetPublic%2FPage%2FQnetTier4&c=Page. Medicare Payment Advisory Commission. 2015. The hospital readmission penalty: How well is it working? Salerno, A. M., L. I. Horwitz, J. Y. Kwon, et al. 2017. Trends MedPAC blog. March 24. http://www.medpac.gov/-blog-/ in readmission rates for safety net hospitals and non-safety net medpacblog/2015/03/24/the-hospital-readmission-penalty-how- hospitals in the era of the US Hospital Readmission Reduction well-is-it-working. Program: A retrospective time series analysis using Medicare administrative claims data from 2008 to 2015. BMJ Open 7, no. 7 Medicare Payment Advisory Commission. 2013. Report to (July 13): e016149. the Congress: Medicare and the health care delivery system. Washington, DC: MedPAC. Sheingold, S. H., R. Zuckerman, and A. Shartzer. 2016. Understanding Medicare hospital readmission rates and differing Medicare Payment Advisory Commission. 2011a. Comment letter Health Affairs penalties between safety-net and other hospitals. to CMS on the proposed rule entitled: Medicare program; hospital 35, no. 1 (January): 124–131. inpatient value-based purchasing program, May 4. http://www. medpac.gov/docs/default-source/comment-letters/06172011_ Winblad, U., V. Mor, J. P. McHugh, et al. 2017. ACO-affiliated FY12IPPS_MedPAC_COMMENT.pdf?sfvrsn=0. hospitals reduced rehospitalizations from skilled nursing facilities 36, no. 1 (January 1): Health Affairs faster than other hospitals. Medicare Payment Advisory Commission. 2011b. Report to 67–73. the Congress: Medicare and the health care delivery system. Washington, DC: MedPAC. Zuckerman, R. B., K. E. Joynt Maddox, S. H. Sheingold, et al. 2017. Effect of a hospital-wide measure on the Readmissions Medicare Payment Advisory Commission. 2008. Report to the New England Journal of Medicine Reduction Program. 377, no. Congress: Reforming the delivery system. Washington, DC: 16 (October 19): 1551–1558. MedPAC. Zuckerman, R. B., S. H. Sheingold, E. J. Orav, et al. 2016. Naylor, M. D., L. H. Aiken, E. T. Kurtzman, et al. 2011. The care Readmissions, observation, and the Hospital Readmissions span: The importance of transitional care in achieving health New England Journal of Medicine 374, no. Reduction Program. reform. Health Affairs 30, no. 4 (April): 746–754. 16 (April 21): 1543–1551. Office of the Assistant Secretary for Planning and Evaluation, Department of Health and Human Services. 2016. Report to Congress: Social risk factors and performance under Medicare’s Washington, DC: ASPE. value-based purchasing programs. June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 31

54

55 CHAPTER 2 Using payment to ensure appropriate access to and use of hospital emergency department services

56 RECOMMENDATIONS The Congress should: 2-1 • allow isolated rural stand-alone emergency departments (more than 35 miles from another emergency department) to bill standard outpatient prospective payment system facility fees and • provide such emergency departments with annual payments to assist with fixed costs. COMMISSIONER VOTES: YES 17 • NO 0 • NOT VOTING 0 • ABSENT 0 ... 2-2 The Congress should reduce Type A emergency department payment rates by 30 percent for off-campus stand-alone emergency departments that are within six miles of an on- campus hospital emergency department. COMMISSIONER VOTES: YES 17 • NO 0 • NOT VOTING 0 • ABSENT 0

57 CHAPTER 2 Using payment to ensure appropriate access to and use of hospital emergency department services In this chapter Chapter summary Medicare’s payment policies should foster adequate access to care and Background • encourage efficient delivery of services. Maintaining access to emergency • Rural areas: Maintaining department (ED) services can be a challenge in isolated rural areas, where a access to emergency single hospital may be the sole source of ED care. If that sole hospital closes, department services access to emergency care can be lost. In contrast, efficiency can be a challenge in urban areas, where EDs can be in oversupply. New urban stand-alone EDs Urban areas: Incentives have • (medical facilities providing ED services that are located apart from a hospital led to an abundance of urban campus and can be either affiliated or unaffiliated with a hospital) could result stand-alone EDs in cases shifting from lower cost settings such as urgent care centers and • Future analyses physician offices, which do not provide ED services and are generally not open 24 hours per day, to the higher cost ED setting, which is generally open 24 hours per day. New stand-alone EDs could also siphon off lower acuity (less severely ill) patients from on-campus hospital-based EDs. In this chapter, we recommend two ways to change the way Medicare pays for ED services to reduce the risk of ED services being undersupplied in rural areas and oversupplied in urban areas. Medicare payment rates to isolated rural stand- alone EDs would increase, and payment rates to urban stand-alone EDs close to other sources of emergency care would decrease. We first review basic information on how Medicare pays for emergency services in rural and urban areas. Second, we outline concerns regarding preserving access to ED services in rural areas, which is a continuation of our June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 35

58 2016 discussion of rural EDs (Medicare Payment Advisory Commission 2016a). Third, we discuss limiting excess volume of ED services in urban areas, which is an extension of our 2017 discussion of stand-alone urban EDs (Medicare Payment Advisory Commission 2017). Maintaining access to emergency department services in rural areas Maintaining access to ED services can be challenging in isolated rural areas with low population densities. In many isolated rural areas, inpatient hospitals’ volumes have fallen dramatically, with many hospitals admitting fewer than one patient per day. However, Medicare will pay a facility for emergency services only if it maintains inpatient services. Therefore, small isolated communities that want an ED must maintain a low-occupancy inpatient department in the hospital. In 2016, approximately 130 hospitals averaged less than 1 admission per day (all payers) and were more than 35 miles from other hospitals. EDs at these hospitals serve as important sources of emergency care, but to maintain these isolated EDs, hospitals must maintain their largely empty inpatient beds. As an alternative to maintaining empty inpatient beds, the Commission is recommending a new payment model that would allow Medicare to pay for emergency services at stand-alone EDs in isolated rural areas (more than 35 miles from another ED). The rural facility would have an ED that is open 24 hours a day, 7 days a week, but would not provide acute inpatient care. The facility could retain other services such as ambulance services and outpatient clinics, and we refer to the combination of the stand-alone ED and its affiliated outpatient services as an outpatient-only hospital. Isolated rural full-service hospitals that choose to convert to outpatient-only hospitals would receive the same standard prospective payment rates for ED visits as a full-service hospital. In addition, a set annual payment (common across all outpatient-only hospitals) would be made to help cover the facility’s fixed costs. The new payment option would allow rural communities that cannot support a full- service hospital a way to maintain access to emergency care in their community, while retaining the option to convert back to a full-service hospital if circumstances change. The recommendation would increase Medicare spending by less than $50 million per year. Encouraging efficient delivery of emergency services in urban areas Urban hospitals can set up stand-alone EDs that bill Medicare as if they are a part of the hospital’s main ED as long as they are located within 35 miles of the main hospital campus. We refer to these hospital-affiliated facilities as off-campus EDs (OCEDs). The number of OCEDs has increased rapidly in recent years, particularly Using payment to ensure appropriate access to and use of hospital emergency department services 36

59 in areas with high household incomes. ED visits overall and their coded severity levels have increased. Under Medicare’s payment system for ED visits, providers have incentives to add new OCEDs rather than urgent care centers, which are paid less than half the hospital ED rates. Patients who are served at off-campus EDs appear to have less complex care needs than those of patients served at on-campus hospital EDs. Ambulance operators typically take trauma, stroke, and heart attack patients to on-campus hospital EDs, which provide trauma services, operating rooms, and inpatient services. As a result, off-campus EDs do not incur the standby costs of these resource-intensive services. While urban off-campus EDs may provide some services not available at doctors’ offices and urgent care centers, we conclude Medicare overpays these facilities relative to what is paid to on-campus hospital EDs for more difficult cases. Medicare currently has two levels of payments for OCEDs. One is for EDs open 24 hours a day, 7 days a week (Type A payment rates), and the other is for EDs open less than 24 hours a day, 7 days a week (Type B payment rates). Type B ED rates are lower under the rationale that these facilities have lower standby costs. In 2018, Type B payment rates are roughly 30 percent lower than Type A rates. Evidence from three states indicates that urban OCEDs likely have lower standby costs than on-campus hospital EDs. The Commission is therefore recommending that Medicare pay urban OCEDs the Type A payment rates reduced by 30 percent— which would better align payments with standby costs and make off-campus ED rates similar to Type B rates. An exception would be needed for the one-quarter of OCEDs that are located relatively far (more than six miles) from on-campus EDs and that likely provide unique access to ED services for their local community (other exceptions could be contemplated when an urban OCED is essential to retain access—for example, if the OCED is the result of its parent hospital closing). Paying these more isolated urban OCEDs the full Type A payment rates would be justified to ensure continued appropriate access to emergency services. The Commission’s urban recommendation would better align payment with the standby costs of urban OCEDs in close proximity to on-campus hospital EDs, while maintaining higher payment rates for urban OCEDs that are located farther from on-campus EDs and may provide unique access to ED services. Medicare beneficiaries served at OCEDs close to on-campus EDs would have lower cost sharing, and access to ED services would be preserved in areas where it is most needed. Overall, this policy would reduce the incentive to develop new off-campus EDs and would lower Medicare spending by between $50 million and $250 million annually. ■ June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 37

60

61 FIGURE Medicare margins... X-X FIGURE All-payer and Medicare emergency department visits per capita 2–1 grew faster than Medicare physician office visits per capita, 2010–2016 16 14.1% 14 12 10 7.2% 8 (2010 to 2016) Percent change 6 3.9% 4 2 0 All payers: Medicare: Medicare: Outpatient ED visits per Part B beneficiary Outpatient ED visits per capita Physician visits per Part B beneficiary ED (emergency department). Outpatient ED visits are those in which the patient was treated in the ED but not admitted to the hospital. ED visits occurring at on- Note: campus hospital EDs and off-campus hospital EDs are both included. American Hospital Association and Medicare claims data. Source: Note: Note and Source are in InDesign. Source: sharing in EDs and, in some cases, denying payment for Background services not deemed emergent (Glatter 2017, Livingston 2018). Higher copayments are unlikely to work for fee- Ideally, Medicare payment policies should encourage for-service (FFS) Medicare given the widespread use of Notes about this graph: the appropriate use and efficient delivery of emergency supplemental insurance. In this chapter, we discuss two department (ED) services to both rural and urban • Data is in the datasheet. Make updates in the datasheet. ways to change the way emergency services are paid. The beneficiaries. Given that ED services can be critically objectives are to reduce the risk of undersupply in rural • WATCH FOR GLITCHY RESETS WHEN YOU UPDATE DATA!!!! important to supporting the care needs of Medicare areas and oversupply in urban areas. • The column totals were added manually. beneficiaries, adequate access needs to be maintained in ED services are most commonly delivered at the roughly rural and urban areas. In rural areas, the challenge can be • I had to manually draw tick marks and axis lines because they kept resetting when I changed any data. 4,500 on-campus hospital EDs that are typically open to maintain access to a single ED. In contrast, in some • I can’t delete the legend, so I’ll just have to crop it out in InDesign. 24 hours per day, 7 days a week (24/7). However, urban areas, concern exists about excessive expansion in increasingly, these services are also provided at OCEDs. the number of EDs, which could result in a shift of care • Use direct selection tool to select items for modification. Otherwise if you use the black selection tool, they will reset to graph Between 2010 and 2016, the number of hospital outpatient from lower cost urgent care centers and physician offices default when you change the data. ED visits (those not resulting in an inpatient hospital to higher cost EDs. Off-campus EDs (OCEDs)—those • Use paragraph styles (and object styles) to format. stay) nationwide increased by more than 7 percent per EDs located apart from the hospital campus—could then 1 • Data was from: R:\Groups\MGA\data book 2007\data book 2007 chp1 capita across all payers (Figure 2-1). Over the same benefit by treating lower cost patients while receiving period, Medicare outpatient ED visits per beneficiary rates equal to on-campus EDs that treat higher acuity increased 14 percent, while Medicare physician office (more severely ill) patients. Private insurers try to manage visits per beneficiary increased about 4 percent. Faster demand for emergency services by charging higher cost | June 2018 Report to the Congress: Medicare and the Health Care Delivery System 39

62 TABLE Medicare ED visits in the two highest paying levels grew 2–1 as a share of all Medicare ED visits, 2010–2016 Outpatient ED visits 2010 2016 ED Percentage point Change in payment change in share number of level Number of ED visits Number Share Share ED visits –21,230 Level 1 682,180 4.4% 660,950 3.6% –0.8 –468,983 Level 2 1,781,920 11.5 –4.4 1,312,937 7.1 32.8 5,198,704 28.0 95,584 –4.8 5,103,120 Level 3 Level 4 4,963,920 32.0 6,426,367 34.6 1,462,447 2.6 7.4 1,956,109 4,960,439 19.3 3,004,240 Level 5 26.7 100.0 0.0 18,559,397 100.0 3,023,927 Total 15,535,380 Note: ED (emergency department). ED payment levels are commonly used as a proxy for the severity of patient illness. Level 1 is the lowest paying level, suggesting these are the lowest severity patients. Level 5 is the highest paying level, suggesting these are the highest severity patients. Data include Medicare Type A and Type B ED visits. Outpatient ED visits are those in which the patient was treated in the ED but not admitted to the hospital. ED visits occurring at on-campus hospital EDs and off-campus hospital EDs are both included. Source: CMS hospital outpatient claims data. growth at EDs relative to physician office visits suggests coding is at least partially responsible for the increased some movement of lower severity cases from lower cost reported severity. Another possibility is that cases formerly physician offices to higher cost ED settings. In 2016, admitted to the hospital are now treated on an outpatient Medicare beneficiaries accounted for 28.4 million ED basis, increasing the share of higher severity cases. visits, counting both outpatient ED visits and ED visits However, the decline in admissions is too small to fully that resulted in an inpatient admission (data not shown). explain the magnitude of the increase in higher level cases seen in EDs. It is unlikely that the growth in higher level Volume of higher level cases has increased ED visits is the result of a real increase in patient severity because the growth in the number of ED visits in Levels 4 For payment purposes, Medicare and many other payers and 5 occurred concurrently with growth in total ED visits. require providers to identify ED visits in one of five levels That is, the growth in the share of higher intensity visits that are based on Current Procedural Terminology (CPT) did not reflect the movement of low-severity cases out of codes and general descriptions of the service. Between the ED. 2010 and 2016, the number of Medicare outpatient ED visits billed at the highest of the five ED levels increased Medicare payments for ED services as a share of all Medicare ED visits, climbing from 19.3 percent to 26.7 percent (Table 2-1). By contrast, during the Medicare beneficiaries who visit EDs generate a physician same period, Medicare ED visits coded in the three lowest claim and a hospital outpatient ED claim. Physician claims paying ED levels declined as a share of all Medicare ED for ED visits are paid through the Medicare physician visits. For example, as a share of all ED visits, Level 3 ED fee schedule (PFS). Hospital claims for ED visits that do visits fell from 32.8 percent to 28.0 percent. not result in an inpatient admission are paid through the hospital outpatient prospective payment system (OPPS) or, Certain factors could account for the more rapid growth in the case of ED visits at critical access hospitals (CAHs), of higher level ED services. One possibility is that 2 under the CAH cost-based payment system. providers are coding a larger share of ED visits in the higher paying levels, a practice referred to as upcoding. The PFS and OPPS both use the five-tiered scale to pay for Given the growth in the overall volume of higher level ED visits. The physician bills Medicare by identifying one visits (a 2.6 percentage point increase in Level 4 visits and of the five ED levels for each case (Table 2-2). The facility 7.4 percentage point increase in Level 5), it appears that Using payment to ensure appropriate access to and use of hospital emergency department services 40

63 TABLE Medicare payment rates for ED visits under the Medicare physician fee schedule 2–2 and hospital outpatient prospective payment system, 2018 OPPS payment amount Type B ED Type A (facility open less than Physician fee schedule rate payment (facility open 24 hours per day) 24 hours per day) for an OPPS visit level $102.49 $21.60 $68.66 Level 1 90.82 Level 2 42.12 124.65 Level 3 63.00 219.10 157.66 355.53 119.52 Level 4 209.01 285.88 520.85 176.04 Level 5 Note: ED (emergency department), OPPS (outpatient prospective payment system). ED payment levels are commonly used as a proxy for the severity of patient illness. Level 1 is the lowest paying level, suggesting these are the lowest severity patients. Level 5 is the highest paying level, suggesting these are the highest severity patients. The table reflects 2018 Medicare payment rates under the physician fee schedule and OPPS and does not include payments for ancillary services that might be incurred at the time of treatment. ED visits are those in which the patient was treated in the ED but not admitted to the hospital. ED visits occurring at on- campus hospital EDs and off-campus hospital EDs are both included. While Type A rates are on average higher than Type B rates, payment rates for Type B Level 1 ED visits are anomalously higher than Type A Level 1 ED visits. Source: CMS calendar year 2018 hospital outpatient prospective payment system final rule. bills under the OPPS, which maintains two sets of rates Medicare generally pays lower amounts for services that depend on the type of facility. Type A rates are used provided at urgent care centers, retail clinics, and for hospital EDs open 24/7. Type B rates are used for EDs physicians’ offices for similar types of patients. New 3 open less than 24/7. hospital-affiliated urgent care centers, independent In 2018, Type B rates are on average urgent care centers, retail clinics, and physician offices roughly 30 percent lower than Type A rates because Type are paid the nonfacility PFS rate and are not permitted B facilities do not incur the cost of maintaining standby 5 to bill facility fees for ED services. ED staff 24/7. While Type A rates are on average higher Using the same than Type B rates, payment rates for Type B Level 1 ED Level 4 example, at one of these non-hospital-affiliated visits are anomalously higher than Type A Level 1 ED providers, the total Medicare payment would be $167 to 4 visits. the physician for an evaluation and management (non-ED) The volume of claims paid under Type B rates is visit. low, accounting for about 1 percent of all Medicare ED claims in 2016. Facilities billing Medicare Type B claims When a beneficiary visits an ED, the facility bills serve lower acuity ED cases Medicare for the ED visit and other outpatient services In 2016, about 83 percent of the Medicare Type B claims (e.g., imaging and lab services) under the OPPS, and were in one of the three lowest ED acuity levels (i.e., the physician bills Medicare under the PFS. Under Levels 1–3; Table 2-3, p. 43). By contrast, only about 38 a hypothetical example of the most common level percent of Type A visits were in one of the three lowest ED billed—a Level 4 ED visit—the Medicare payment acuity levels. This difference may be too large to attribute rate for a hospital ED open 24/7 is $356 (not including simply to coding differences at the types of ED facilities other outpatient services) and for the physician is $120, and may demonstrate real differences in the acuity of cases totaling a Medicare payment of $476 (Figure 2-2, p. 42). treated at Type A and Type B ED facilities. These data If the same patient were treated at a hospital ED open suggest that Type B facilities, which in 2016 accounted for less than 24/7 (that is, a hospital receiving the Type B 1 percent of all Medicare ED claims, generally serve lower rate), the Medicare payment to the facility would be $209 acuity cases than Type A facilities. and payment to the physician would be $120, for a total payment of $329. June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 41

64 FIGURE FIGURE Medicare margins... Medicare Type A ED payment rates for similar services 2–2 X-X are higher than Type B ED payment rates and urgent care centers and physician offices payment rates, 2018 500 $476 Hospital outpatient prospective payment rate Physician fee schedule payment rate $120 400 $329 300 $120 200 $167 $356 100 $209 Medicare payment rates (in dollars) $167 0 Type A Urgent care center/physician office Type B (Facility open (Facility open 24 hours per day) less than 24 hours per day) Hospital emergency department ED (emergency department). Hospital outpatient prospective payment rates for Type A and Type B visits reflect Level 4 ED services. The physician fee schedule Note: payment rates for services delivered in hospital emergency departments reflect Level 4 physician ED services. Payment rates for for services delivered in urgent care centers and physician offices reflect Level 4 evaluation and management codes for new patients. In addition, the urgent care center/physician office payment of $167 reflects the rate paid to new urgent care centers or older urgent care centers not affiliated with a hospital, which do not receive a facility fee for outpatient Note: Note and Source are in InDesign. services. Figures have been rounded. Source: Source: MedPAC analysis of Medicare 2018 hospital outpatient prospective payment system payment rates and physician fee schedule payment rates. as many as 100 patients per day and the smallest facilities Current Medicare payment policies Notes about this graph: serving 20 or fewer patients per day. Larger OCEDs encourage stand-alone EDs • Data is in the datasheet. Make updates in the datasheet. and IFECs also can offer MRI and primary care, house A growing number of ED facilities are located apart from • WATCH FOR GLITCHY RESETS WHEN YOU UPDATE DATA!!!! physician specialists’ offices, and tend to take more a hospital campus and are known as stand-alone EDs. ambulance transports than smaller OCEDs and IFECs. There are two types of stand-alone EDs: hospital-affiliated • The column totals were added manually. They typically have one or more physicians on-site at all off-campus emergency departments and independent • I had to manually draw tick marks and axis lines because they kept resetting when I changed any data. times (typically under contract). These facilities often freestanding emergency centers (IFECs). advertise that they are open longer (24 hours per day) • I can’t delete the legend, so I’ll just have to crop it out in InDesign. OCEDs and IFECs generally offer a similar range of than urgent care centers and treat medical conditions such • Use direct selection tool to select items for modification. Otherwise if you use the black selection tool, they will reset to graph services. Both offer ED services 24/7; basic imaging as respiratory distress, infection, orthopedic injuries and default when you change the data. services such as X-rays, computed tomography (CT) fractures, and abdominal pain. A certain degree of overlap • Use paragraph styles (and object styles) to format. scans, and ultrasound; and on-site lab services for basic exists between the lower acuity cases treated at stand- diagnostic analyses. Neither typically provides trauma alone EDs and urgent care centers, signifying that urgent • Data was from: R:\Groups\MGA\data book 2007\data book 2007 chp1 services (e.g., care for victims of car accidents or gunshot care centers are also important in supporting the care wounds). They range in size, with larger facilities serving needs of Medicare beneficiaries. Using payment to ensure appropriate access to and use of hospital emergency department services 42

65 TABLE Medicare Type B ED claims included a larger share 2–3 of lower level ED visits than Type A ED claims, 2016 Type A ED visits Type B ED visits (facility open less than 24 hours per day) (facility open 24 hours per day) ED payment Share of visits Number of visits level Number of visits Share of visits 18.0% 627,561 3.4% 33,389 Level 1 Level 2 1,262,344 6.9 50,593 27.2 70,872 38.1 27.9 5,127,832 Level 3 14.1 26,226 34.9 6,400,141 Level 4 2.6 Level 5 4,955,541 4,808 27.0 100.0 185,888 100.0 18,373,419 Total Note: ED (emergency department). Total shares of visits may not total 100 percent due to rounding. CMS hospital outpatient claims data. Source: Between 2008 and 2016, the number of OCEDs roughly substantially more than PFS payment rates. Medicare doubled. In 2017, about 580 stand-alone EDs, including thus may be overpaying for the ED and outpatient OCEDs and IFECs, were in operation. Two-thirds of services furnished in micro-hospitals, encouraging their these facilities—377 facilities—were OCEDs, located proliferation. About 50 micro-hospitals are open or under in 35 states and affiliated with more than 300 hospitals. development in Arizona, Colorado, Idaho, Missouri, The remaining one-third of stand-alone EDs were IFECs. Nevada, Pennsylvania, Oklahoma, and Texas. In addition, We have identified about 200 IFECs, operating mostly in the for-profit hospital system Tenet Health stated in Texas but also in Colorado and Minnesota. In Texas, the its 2018 annual report to shareholders that it currently number of IFECs increased from 0 in June 2010 (when operates eight micro-hospitals (Morningstar Document state licensure of IFECs began) to 191 facilities in 2016. Research 2017b). The Commission may conduct future The proliferation of IFECs between 2013 and 2017 has focused research on micro-hospitals. been particularly rapid in the Dallas metropolitan area, In addition to EDs, more than 7,000 urgent care centers where the number of state-registered IFECs nearly tripled, 6 compete for lower acuity patients. Urgent care centers from 25 to 73. provide a broad range of nonemergency services but ED services are also provided at micro-hospitals, generally maintain somewhat less service capacity than which are smaller than full-service hospitals and offer a on-campus hospital EDs. They are typically open less than limited range of services. Micro-hospitals focus on the 24 hours per day; are staffed by physicians, nurses, and delivery of emergency services and typically have 10 or physicians’ assistants; and offer relatively limited lab and fewer inpatient beds. Some micro-hospitals also house imaging services. In addition, research suggests that urgent primary care physician practices, specialty physician care centers treat lower severity patients than on-campus practices, and labor and delivery services (Andrews hospital EDs but that there is overlap between these types 2016). However, micro-hospitals typically do not offer of facilities in terms of the types of patients they treat higher intensity services such as trauma care and intensive (Baker and Baker 1994, Mehrotra et al. 2009, Thygeson et care, and patients requiring prolonged care are regularly al. 2008). This overlap occurs among the lowest severity transferred to larger facilities (Rudavsky 2016). As a patients. A 2010 study estimated that between 13 percent result, micro-hospitals likely do not incur the standby costs and 27 percent of cases served in hospitals’ on-campus of full-service hospitals. Nevertheless, micro-hospitals EDs could be served similarly at urgent care centers or are licensed as independent inpatient hospitals and, as by other providers (Ashwood et al. 2016, Weinick et al. such, can bill Medicare under the OPPS, which pays 2010). The severity of patients treated at OCEDs appears June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 43

66 FIGURE Updating... FIGURE X-X Illustrative example of Medicare ED payment rates by facility type 2–3 $167 Urban concern: Off-campus Off-campus EDs located within the 35-mile hospital ED radius and in close proximity to their affiliated on-campus hospital ED receive the higher hospital outpatient payment rates. Rural concern: $476 Off-campus EDs located Off-campus outside the 35-mile radius hospital ED cannot receive the higher hospital outpatient payment rates. $476 On-campus hospital ED $167 Urgent care center 35-mile radius (Medicare requires $167 off-campus facilities to be Urgent care within 35 miles of their center affiliated hospital to receive the higher hospital outpatient payment rates.) ED (emergency department). The ED payment amounts displayed are for Level 4 Type A ED visits and for Level 4 office visits at an urgent care center. Note: Note: In InDesign. to be above that at urgent care centers but lower than that departments must be in compliance with Medicare at on-campus hospital EDs. Evaluating how Medicare and state hospital ED requirements, be financially and payment policy influences the treatment location of low- clinically integrated with the hospital, be publicized acuity cases at emergency departments may be an area of as an affiliate of the hospital, and be located within 35 future Commission research. miles of the main hospital campus (Centers for Medicare 7 & Medicaid Services 2008). Most private payers pay Billing for off-campus ED services OCEDs a facility fee and generally consider OCEDs in- network facilities. OCEDs bill Medicare under the OPPS for a beneficiary’s ED visit and any ancillary services (e.g., imaging and lab If a patient is treated at an OCED, Medicare pays the services), while the clinicians bill under the Medicare Type A payment rate as if the patient were at the main PFS. In order to bill Medicare, OCEDs must be deemed hospital campus. As with on-campus EDs, if the patient off-campus provider-based departments. Provider-based is transferred from the OCED to the main hospital for Using payment to ensure appropriate access to and use of hospital emergency department services 44

67 admission, then the ED visit and the ambulance transfer with hospitals, effectively converting to new OCEDs, to will not be paid separately but, instead, will be deemed gain Medicare provider–based status and begin billing part of the cost of the inpatient admission that is bundled Medicare. For example, in recent years, the largest owner into the diagnosis related group payment. of IFECs, Adeptus, modified its business model to partner with hospitals to enable its IFECs to bill Medicare and Under current law, hospitals have a financial incentive to Medicaid. In Arizona and Ohio, Adeptus partnered with build new off-campus EDs and colocate physician offices large health systems to build new stand-alone EDs. In and specialty clinics within them. The Bipartisan Budget Colorado, Adeptus partnered with the University of Act of 2015 (BBA of 2015) requires that new “provider- Colorado Health to build new hospitals with which its based” clinics owned by hospitals be paid under “the existing IFECs could then affiliate. In Texas, Adeptus applicable payment system.” The BBA of 2015 did not made two significant changes that enabled their IFECs to specify the applicable payment system, but CMS chose a begin billing Medicare. First, they began building their method of paying reduced OPPS rates that are comparable own new hospitals (without partnering with a hospital to rates paid in independent physician offices. The BBA system). Second, Adeptus partnered with hospital system of 2015 includes an exception to these reduced OPPS Texas Health Resources, and as a part of the Texas Health payment rates for any services provided in “dedicated Resources agreement, 31 IFECs in Dallas began billing EDs.” This exception, defined in Section 603, requires Medicare as OCEDs. that both ED and non-ED services (e.g., clinic visits and ancillary services) provided in off-campus EDs be paid the In addition, large for-profit hospital systems are building 8 full OPPS payments rates. OCEDs into their business development strategies. In their 2017 annual report to shareholders, Hospital Corporation The other type of stand-alone ED facility, IFECs, cannot of America reported that OCEDs are an integral part bill Medicare because they are not affiliated with a hospital of their strategy to develop comprehensive health care or considered provider-based facilities by Medicare. Thus, networks in select communities (Morningstar Document the ED payment policies discussed in this chapter do not Research 2017a). Community Health Systems also address IFECs. Private insurers do not typically contract reported that it will use OCEDs to improve market share with IFECs and instead treat them as out-of-network in certain markets (Community Health Systems 2017). providers. According to several news reports, private The investment of these large hospital systems in OCEDs insurers are charged significantly higher rates when IFECs suggests the model is viewed as beneficial to the overall are out-of-network facilities, and patients are often left to success of the system. pay the balance of these charges when claims are denied in part or in full (Rice 2016, Sutherly 2016). Growth in private-payer payment rates also encourages the development of stand-alone Location of OCEDs can impact Medicare payment EDs rate The proliferation of stand-alone EDs is at least in part due Medicare requires provider-based off-campus facilities, to incentives created by commercial insurance contracts such as OCEDs, to be within a 35-mile radius of their to expand ED services. The Health Care Cost Institute affiliated hospitals to receive the higher OPPS payment reported that the price paid per emergency room visit by rates. Figure 2-3 combines the payment rate example used private insurers increased by 31 percent from 2012 to 2016 in Figure 2-2 (p. 42) with Medicare’s 35-mile threshold. (Health Care Cost Institute 2018). Given the growth in the OCEDs located within a 35-mile radius of their affiliated number of stand-alone EDs during these years, it appears hospital are paid $476, the same as an on-campus hospital that the providers’ pricing power is sufficient to encourage ED. By contrast, OCEDs located outside of the 35-mile expansion. Private insurers try to manage demand for radius are paid $167 for a comparable service, which is the emergency services by having higher cost sharing in same as an urgent care center located within or outside the emergency departments and, in some cases, denying 35-mile radius. payment for services not deemed emergent. Higher copays are unlikely to work for fee-for-service (FFS) Medicare, More stand-alone EDs will begin billing Medicare given the widespread use of Medicare supplemental by converting from IFECs to OCEDs insurance. Therefore, other mechanisms for preventing Although Medicare does not pay for services provided in excess use of EDs are needed for the Medicare program. IFECs, many of the 200 IFECs are taking steps to affiliate June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 45

68 FIGURE Title here... 1-XX FIGURE Inpatient use of critical access hospitals declined, 2003–2016 2–4 1,000 900 Median 10th percentile 800 700 624 600 500 per CAH 400 335 300 Total annual discharges 170 200 71 100 0 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2003 2004 2016 CAH (critical access hospital). Note: MedPAC analysis of hospital cost report data from CMS. Source: inpatient facilities, which are operating at a scale that may Rural areas: Maintaining access to not be optimal from a quality or cost-of-care standpoint emergency department services (Medicare Payment Advisory Commission 2016a). Note: Note and Source are in InDesign. Source: Maintaining emergency access in rural areas is challenging Medicare’s existing programs for preserving due to declining populations in many rural areas, coupled rural hospitals are inpatient-centric with a delivery system that is tied to an expensive The Medicare program has several rural payment Notes about this graph: inpatient delivery model. In addition, rural hospitals are programs designed to preserve rural hospitals. Most of • Data is in the datasheet. Make updates in the datasheet. losing volume as rural patients often bypass their local these programs are inpatient-centric models. The sole rural hospital for larger (and more distant) rural or urban • I deleted the years from the x-axis and put in my own. community hospital (SCH) program increases inpatient facilities. In many cases, the bypass occurs even when the and outpatient payments by about $900 million per year • I had to manually draw tick marks and axis lines because they kept resetting when I changed any data. services are available locally (Liu et al. 2008, Medicare above inpatient prospective payment system (IPPS) rates • The dashed line looked ok here, so I didn’t hand draw it. Payment Advisory Commission 2012, UnitedHealth to over 300 SCHs. The Medicare-Dependent Hospital Center for Health Reform & Modernization 2011). By • I can’t delete the legend, so I’ll just have to crop it out in InDesign. (MDH) Program increases inpatient payments by about 2016, the urban hospital occupancy rate was 66 percent $100 million per year above IPPS rates to about 150 rural • Use direct selection tool to select items for modification. Otherwise if you use the black selection tool, they will reset to graph compared with 40 percent for all rural hospitals and hospitals. Sixty percent of rural hospitals (1,300) receive default when you change the data. 31 percent for rural hospitals with fewer than 50 beds cost-based payment through the CAH program. This cost- (Medicare Payment Advisory Commission 2018). In • Use paragraph styles (and object styles) to format. based payment program increases payments to CAHs by 2016, approximately 130 hospitals were more than 35 about $2 billion per year relative to prospective payment miles from other hospitals and averaged fewer than 1 system rates for acute care hospitals (Medicare Payment admission per day (a map is included in online Appendix Advisory Commission 2012). 2-B, available at http://www.medpac.gov). The question Despite the SCH, MDH, and CAH programs, rural is whether emergency services can be provided by these hospital closures have increased in the last three years. isolated facilities without having to maintain the hospitals’ Using payment to ensure appropriate access to and use of hospital emergency department services 46

69 FIGURE Title here... FIGURE 1-XX The number of Medicare outpatient emergency department 2–5 visits at critical access hospitals grew, 2010–2016 2,500 2,012 2,000 1,735 1,500 Median 10th percentile 1,000 ED visits per CAH Total annual Medicare 280 376 500 0 2012 2015 2014 2013 2016 2011 2010 ED (emergency department), CAH (critical access hospital). Note: Source: MedPAC analysis of Medicare outpatient claims data. Some closures reflect excess capacity, but in other occupancy rate (including post-acute swing-bed patients) instances, the closed hospitals were the sole providers of between 2006 and 2016 fell from 38 percent to 31 percent. Note: Note and Source are in InDesign. emergency services in their area. From 2013 through 2017, While hospitals’ inpatient volume continues to decline, the Source: 51 rural hospitals closed (67 if we include rural areas of use of the emergency services at CAHs increased slightly 9 metropolitan counties) (Young 2018). Among the closures in recent years (Figure 2-5). This increase suggests the were 22 CAHs. While 28 of the closures were less than 20 Notes about this graph: community still values local emergency access. Figure 2-4 miles from the nearest hospital (suggesting there may have and Figure 2-5 together illustrate how CAHs have shifted • Data is in the datasheet. Make updates in the datasheet. been excess capacity in these markets), 21 were 20 to 35 substantially to outpatient rather than acute inpatient miles from the nearest hospital, and 2 were over 35 miles • I deleted the years from the x-axis and put in my own. services. In contrast, rural payment models continue to be from the nearest hospital. • I had to manually draw tick marks and axis lines because they kept resetting when I changed any data. inpatient-centric. The financial challenges faced by CAHs can include • The dashed line looked ok here, so I didn’t hand draw it. To maintain access to care in communities where inpatient declining populations, declining volume of patients with volume is declining, there is an interest in payment • I can’t delete the legend, so I’ll just have to crop it out in InDesign. commercial insurance, difficulty recruiting physicians, models that are focused on outpatient access rather continued uncompensated care costs, and patients • Use direct selection tool to select items for modification. Otherwise if you use the black selection tool, they will reset to graph than maintaining inpatient services (American Hospital bypassing the local CAH for larger hospitals. In particular, default when you change the data. Association 2016, Iglehart 2018, Thompson 2015). A key the decline in admissions is difficult for hospitals built • Use paragraph styles (and object styles) to format. question is whether a rural hospital could cease providing on an inpatient payment model. From 2003 to 2016, the its inpatient services and still generate enough outpatient median number of annual all-payer discharges among revenue to maintain an ED. This approach works in some CAHs fell from over 600 to 335, and 10 percent of CAHs communities, but they are generally rural communities had 71 or fewer discharges in 2016 (Figure 2-4). Despite with a fairly high ED volume and payer mixes that having 25 or fewer beds per CAH, the median CAH June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 47

70 Examples of rural off-campus emergency departments to which one OCED belonged centralized many n August 2017, Commission staff conducted administrative services (e.g., billing, legal services, and multiple site visits to off-campus emergency contract negotiations) and charged the stand-alone ED departments (OCEDs) located outside of major I a fee. The costs to provide those services would likely metropolitan areas. The facilities we visited were have been much higher if the facility had provided them located within 35 miles of their parent hospitals independently. Given these circumstances, the estimates and therefore considered OCEDs for the purposes we heard during our site visits were in line with of Medicare billing. The OCEDs were located in previously published research suggesting a minimum communities that experienced hospital closures, often budget of roughly $5 million per year to operate a rural due to low inpatient volumes that led to financial losses; OCED (Williams et al. 2015). some of the OCEDs were located in the same physical facilities that once housed the closed hospitals. We At each of our site visits, the facility representatives toured the facilities and spoke with representatives of said receiving Medicare’s facility payments is critical those facilities, representatives of their parent hospitals, to ensure the viability of their stand-alone EDs. To and local emergency medical services (EMS) providers demonstrate that point, representatives of the parent to better understand the challenges associated with hospital of one of the freestanding rural EDs we operating an OCED in more rural locations and to visited said other struggling inpatient hospitals have inform our discussion of potential policy changes. contacted them to inquire about converting their facilities to stand-alone EDs. The offers were turned The representatives with whom we spoke said the cost down because none of their own hospitals were within to run their OCEDs was anywhere from approximately 35 miles of the struggling facilities, which would have $3 million to $5 million a year. Some of these estimates made the struggling inpatient hospitals unable to bill are likely low because they did not include costs such as an outpatient department of the larger hospital and as depreciation or rent and represent efficiencies of receive facility payments from Medicare. The need to belonging to a system. For example, one facility we receive facility payments for their Medicare patients is toured rents its building from the county government particularly acute for rural facilities because more of for a nominal fee. Its representatives asserted that if their patients tend to be covered by Medicare and fewer the ED had to pay market rates for the building, their tend to have private insurance. Some representatives costs would be higher. In another instance, the system said their stand-alone EDs were not financially viable (continued next page) include a large share of privately insured patients. Most ED model work without additional subsidies. For conversions of rural hospitals to stand-alone EDs are cases example, after three rural Georgia hospitals closed, some in which the closed hospital is within 35 miles of another discussed operating them as stand-alone EDs. However, a hospital and can be deemed an outpatient department of committee formed by the state concluded that the stand- another hospital. That arrangement allows the hospital to alone EDs would not have enough volume to be viable obtain facility fees. (See text box for more detail on how without additional support (Rural Hospital Stabilization this model of rural OCED can work.) In contrast, stand- Committee 2015). In addition, if a closed hospital is more alone EDs that cannot bill for facility fees are often not than 35 miles from another hospital, the hospital cannot financially viable. operate as a department of another hospital and receive facility fees. This situation is at odds with the objective Some rural communities have too few ED patients and of preserving access: Isolated communities are the ones too few private-pay patients to make the stand-alone that currently cannot receive Medicare’s facility fees for Using payment to ensure appropriate access to and use of hospital emergency department services 48

71 Examples of rural off-campus emergency departments (cont.) even with Medicare’s outpatient prospective payment health care facilities, including stand-alone EDs and system Type A ED payment rates and therefore hospitals. While patients may request to go to a specific required additional subsidies to remain open. For facility, the EMS providers said their staff make example, one stand-alone ED initially received a recommendations to patients and select the facility subsidy from the system to which it belonged to remain for those who are unconscious or otherwise unable to viable, and one stand-alone ED remained viable only make a decision. For example, the stand-alone EDs we because it was an off-campus department of a critical visited were generally bypassed or used only to stabilize access hospital that received cost-based reimbursement patients with ST-elevation myocardial infarctions, a from Medicare. life-threatening type of heart attack during which one of the heart’s major arteries is blocked. This dynamic The facility representatives said viability also depended whereby more serious cases routinely bypass stand- on achieving a certain volume of ED visits. They said alone EDs may be somewhat different for facilities that they generally need 30 to 40 visits per day, or roughly are farther away from other hospitals because bypassing 10,000 to 15,000 visits per year, for a rural off-campus such facilities means a longer transport than bypassing a ED to remain sustainable, although they noted that the stand-alone ED that is located near another hospital. In number of ED visits required to remain viable varies general, the representatives of systems that operated both based on factors such as payer mix. For the stand-alone rural and urban OCEDs said that patients at rural stand- EDs we visited, facility representatives said the vast alone EDs tended to present with more serious injuries majority of their patients were walk-ins, as opposed to or illnesses than those at urban stand-alone EDs because patients arriving by means of ambulance or helicopter. the rural facilities are often a longer distance from other While representatives said their EDs treat patients hospitals with an ED than urban stand-alone EDs. with a variety of severity levels—from patients in cardiac arrest to those who need a simple X-ray—they In addition to ED visits, all the facilities we visited suggested that patients treated at their stand-alone EDs had some colocated services and used their equipment tended to present with less severe injuries or illnesses for dual purposes. For example, all the facilities we compared with patients at on-campus EDs. visited rented space to local physicians, including primary care physicians and specialists. Some local The EMS providers we interviewed said their staff residents also used the facilities for nonemergent care, are familiar with the capabilities of all the local most commonly for imaging and laboratory services. (continued next page) OCEDs. These rural facilities may see the only option payment system (PPS) payments (Medicare Payment under current payment policy is to continue as a CAH and Advisory Commission 2016a). These supplemental receive cost-based payment; however, that is not efficient payments primarily reflected high rates that CAHs receive and may not be financially sustainable. for post-acute care. The extra payments for inpatient care were not sufficient Extra inpatient payments do not always to keep these hospitals open because the extra payments keep the emergency department doors open were absorbed by these hospitals’ high inpatient costs per High inpatient payments have not always kept rural day of care. For policymakers, a key question is whether hospitals open. In 2016, we conducted an examination these hospitals could have retained emergency capacity of all CAHs that closed in 2014. We found that, before if the Medicare program had directed the supplemental their closure, the seven hospitals received an average of payments toward preserving emergency services rather $500,000 in payments above the comparable prospective than subsidizing acute and post-acute inpatient services. June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 49

72 Examples of rural off-campus emergency departments (cont.) some hospitals were able to avoid paying a subsidy for All the EDs we visited offered a range of imaging their ED physicians because the system to which they services, including X-rays, ultrasounds, and computed belonged negotiated a contract for all of the system’s tomography (CT) scans, and sometimes including EDs, which included urban facilities and facilities with additional imaging services such as mammography, better payer mixes. nuclear medicine, and magnetic resonance imaging. Because the stand-alone EDs we visited were Finally, some of the facility representatives said that considered hospital outpatient departments, the being part of a larger hospital system was critical facilities received hospital outpatient rates rather than to making their stand-alone ED financially viable the lower physician office rates for imaging services. and more medically capable. According to the representatives, being part of a system helped them The facility representatives said that rural hospitals decrease costs (e.g., by centralizing nonclinical traditionally staffed their EDs by relying on community functions and increasing their purchasing power for physicians to cover the ED. The use of this model drugs and supplies) and increase revenues (e.g., stand- is decreasing because it has become harder to find alone EDs benefit from the higher private-payer rates physicians willing to maintain a community practice negotiated by the larger system). Clinically, they also plus cover the ED. They said rural EDs are increasingly mentioned that being part of a system gave their stand- staffing their EDs with dedicated personnel. All the alone EDs better access to physicians by, for instance, stand-alone EDs we visited were staffed 24/7 with a allowing the hospital system’s employed physicians physician board-certified in emergency medicine that to rotate through rural areas (e.g., attend a clinic one was contracted through a physician staffing company day a week) and increasing the timeliness of specialist (e.g., Apollo or EmCare), and some supplemented their consults through telehealth. All the facilities we physicians with midlevel practitioners during peak visited had some telehealth capabilities. For example, hours. Facility representatives said it can be difficult physicians at a more remote stand-alone ED would to recruit and retain such personnel to practice in rural take a CT scan of a patient who suffered a stroke and areas. They also noted that rural facilities might have project that image on a screen along with a live video to pay such companies subsidies amounting to several of a neurologist who was based at an urban hospital. hundred thousand dollars per year to recruit physicians This approach allowed the ED physician access to to practice in a rural ED. For example, the physician the expertise that is often unavailable in rural areas staffing company would receive all the professional but is critical in determining the appropriate course of billings for the services their physicians perform in treatment, such as whether to administer a clot-busting the ED plus an additional subsidy from the hospitals. drug and whether the patient needs to be transported by We heard that some rural EDs have faced difficulties means of ambulance or helicopter. financing such subsidies. The representatives noted that ■ in response to a closure 40 miles from other hospitals—in Limitations on growth of rural OCEDs a community that truly lacks access to ED services—the To bill as an OCED, a rural ED must be within 35 miles of hospital setting up that OCED would be paid at physician the main hospital campus. For urban EDs, this requirement office rates. The net result is that the Medicare program is largely not a problem unless a hospital system seeks currently pays more for care in OCEDs that are close to to open a stand-alone ED in a distant market, but for alternative sources of emergency care than it does for EDs rural areas, the 35-mile criterion can be a challenge. For that are the only source of ED care. As we discussed in example, if a rural hospital wants to set up an ED in a the text box on rural off-campus emergency departments, community 10 miles away, it can do so and receive full Type A ED rates. But if the same hospital opened an ED Using payment to ensure appropriate access to and use of hospital emergency department services 50

73 these isolated EDs appear to receive more difficult cases the annual subsidy). Our June 2016 report discussed the than the higher paid OCEDs that are close to a hospital. option of having a clinic open 12 hours a day 365 days a year as an alternative for very low-volume providers A new policy to preserve isolated rural (Medicare Payment Advisory Commission 2016a). emergency departments However, rather than form a new payment model for such facilities, it may make sense for them to be operated as There is a growing interest in trying to preserve access federally qualified health centers (FQHCs). The FQHC to 24-hour emergency services in rural areas without program provides federal grant funds and a per visit requiring hospitals to provide inpatient services (American payment to support stand-alone clinics in rural and urban Hospital Association 2016, Iglehart 2018, Morse 2015). areas. Any such policy should achieve three objectives: All hospitals that convert to an outpatient-only facility provide a mechanism for preserving emergency access • would receive equal annual fixed payment amounts. in isolated areas Unlike a cost-based model, hospitals with higher cost not materially increase overall Medicare spending • structures (often those with more financial resources) would not receive a higher payment. The fixed payment • improve efficiency of the health care delivery system would also not increase with volume because standby ED costs would not materially shift with volume changes, and Under a proposed rural 24/7 ED model discussed in Medicare would not want to encourage unnecessary ED our June 2016 report, Medicare would pay the Type A use. It would also differ from cost-based models because outpatient ED rates plus a fixed payment to partially hospitals would no longer have an incentive to offer cover overhead services (Medicare Payment Advisory services for which their costs are not competitive (e.g., Commission 2016a). This approach would encourage the post-acute services, MRI services) because additional outpatient facility to focus on ED services, ambulance volume would not lead to increases in supplemental services, and primary care. The fixed payment could Medicare payments. be used to support the rural ED’s standby costs and the cost of other services that help preserve access, such as If a hospital with inpatient services converted to an telehealth services. While a few rural PPS hospitals as outpatient-only facility, we expect that the financing and well as a few rural clinics could convert to a model of delivery of care would change as follows: an outpatient-only hospital, the providers most likely to convert would be CAHs with very low inpatient volume. Isolated hospitals choosing to eliminate acute inpatient • services and accept PPS rates would qualify to receive To fund the additional fixed payment without materially an annual fixed base payment from Medicare. The increasing overall Medicare spending, Medicare could inpatient volume would flow to neighboring hospitals, use the savings generated from discontinuing inpatient potentially improving the neighboring hospitals’ payments at the CAHs participating in this model— financial viability. roughly $500,000 on average—to fund the fixed payment. A subsidy of this magnitude would represent about 10 • Given that the fixed payment would be directed to percent of the cost of operating a small stand-alone preserving emergency access, some hospitals could 10 ED. The rationale for this approach is that if standby convert their hospital beds to skilled nursing facility emergency and primary care capacity are the desired (SNF) beds for which they would receive SNF PPS services, then Medicare should subsidize the cost of rates for the SNF services provided under the existing facilities’ standby capacity with an annual fixed payment eligibility rules. rather than increased payments per inpatient day. The • Converting facilities would make it possible to place a fixed Medicare payment and the annual local support from priority on emergency care. the town, hospital district, or county would help maintain emergency access, even in a low-volume environment. See • Outpatient clinics would continue to operate (e.g., online Appendix 2-A, available at http://www.medpac.gov, FQHCs and freestanding rural health clinics). for a summary of the proposed rural policy. The facilities would have greater flexibility to use • There may be some rural communities where the telehealth consultations. They would still receive the population is too low to support a 24/7 ED (even given June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 51

74 hospitals’ OPPS telehealth fee, but they could also use example, the larger hospital could help with peer review of the fixed payment to help support telehealth. physicians, purchasing supplies, and billing for services. Under this option, the new outpatient-only facility could • Eliminating services that can be more efficiently work cooperatively with other health care providers to delivered in centralized regional facilities (e.g., MRI provide continuity of care across settings. services) would substantially lower costs relative to the CAH models. Who would receive the rural fixed payment to maintain a 24/7 ED? Rural stand-alone EDs could switch back to A facility that eliminated inpatient services (acute and CAH status post-acute swing services), accepted outpatient PPS rates, In determining whether or not to participate in the rural and converted to an outpatient-only facility would receive outpatient-only hospital model, existing hospital boards the fixed payment. To ensure that the funds were used as would have to decide whether they were willing to intended, the facility could be required to use the fixed discontinue providing inpatient services and convert to payment for emergency standby capacity, ambulance outpatient-only hospitals. Discontinuing inpatient services service losses, telehealth capacity, and uncompensated would be a difficult process for rural communities that care in the ED. The 24/7 ED could be required to be have long been served by hospitals that focused on periodically recertified to determine that the facility inpatient care. To reduce the communities’ perceived risk was still isolated from full-service hospitals and was of conversion, Medicare could allow all CAHs that convert appropriately spending the annual fixed payments to to stand-alone EDs the option of converting back to CAH operate a 24/7 ED. We refer to the combination of the status in the future if the community demographics change stand-alone ED and its affiliated services (e.g., telehealth, so that a full-service hospital is once again needed in the ambulance, clinic services, rehabilitation services) as an community. Conversion back to a hospital, although rare, outpatient-only hospital. is occurring in one of the communities we visited. As discussed in the text box on rural OCEDs (pp. 48–50), It is not clear how many providers would choose to we visited three communities where the only hospital convert from a PPS hospital or CAH status to an outpatient within 20 miles closed. In two of the three communities, hospital under this policy. The decision would in part the population of the town grew fairly rapidly after the be determined by the size of the fixed payment and how hospital closed. In both cases, population growth led the program was targeted. The fixed-payment model we to opening stand-alone EDs where two hospitals were isolated discuss is targeted to isolated providers only; once located. In one of the communities, the population could be defined as a certain driving distance from other has continued to grow to the point where the operator of EDs. (See online Appendix 2-B, available at http://www. the ED is now going to build a new full-service hospital medpac.gov, showing a map of all isolated low-volume attached to the stand-alone ED. While we expect this hospitals more than 35 miles from another hospital. We option of converting back to a CAH will be rarely used, it use the 35-mile criterion because EDs less than 35 miles should make the initial decision to convert to a stand-alone from a traditional hospital have the option to become ED easier. an outpatient department of a neighboring hospital. In addition, the 35-mile criterion is the limit currently used in To be willing to shift to a stand-alone ED model, the SCH and CAH programs.) small communities’ hospital boards may need to better understand the limited economic effect of conversions Shifting from CAH status to a stand-alone ED of hospitals to outpatient-only facilities. While the would reduce patient cost sharing two communities that grew after hospital closures are Another consideration with regard to CAHs shifting to anecdotal observations, we are not aware of any research stand-alone ED status is the degree to which beneficiaries’ showing the conversion of a hospital to an outpatient-only cost-sharing obligations would decline when hospitals facility had large economic effects on rural communities. shifted from CAH status to PPS rates. Past Commission work suggests that the Medicare program’s share of A converted outpatient facility would also have the cost-based payments to CAHs for outpatient services option of aligning with its area’s larger hospital system to (net of patients’ coinsurance liabilities) is roughly equal support some functions at the outpatient-only facility. For Using payment to ensure appropriate access to and use of hospital emergency department services 52

75 that (1) never had a hospital, (2) had a hospital that closed, to PPS rates (Medicare Payment Advisory Commission 2012). Although the Medicare program would not or (3) have an open full-service hospital that they want to convert to an outpatient-only facility. realize significant program savings from shifting from CAH cost-based rates for outpatient services to PPS IMPLICATIONS 2-1 rates, beneficiary cost would decline dramatically. The reason is that beneficiaries’ coinsurance at CAHs is set Spending at 20 percent of charges, which is roughly 50 percent of • Most rural stand-alone EDs would be former CAHs. the cost-based payment and often close to the full PPS Under this recommendation, Medicare would make payment rate (Medicare Payment Advisory Commission annual lump sum payments to CAHs that convert to 2016a, Medicare Payment Advisory Commission 2011). become a rural stand-alone ED and maintain only When facilities switch from CAH status to PPS rates outpatient services. These payments, if in the range of under stand-alone ED status, Medicare beneficiaries $500,000, would be offset by savings from reduced could see their coinsurance fall by 70 percent or more. payments for post-acute care (PAC) services as For example, if the CAH billed $700 for a Level 3 ED beneficiaries who might have received PAC services visit that cost $350, the beneficiary would owe the CAH at the CAH are shifted to other PAC providers at a 20 percent of $700 ($140) in cost sharing. If the facility lower cost to Medicare. However, a small share of converted to a stand-alone ED, the payment rate for the the outpatient-only facilities would be either former service would fall to $200 and PPS ED coinsurance would PPS hospitals or hospitals that would have closed be $40 (71 percent less than CAH coinsurance). However, without the new program. Preserving these hospitals given the widespread use of Medicare supplemental and access to emergency care in these communities insurance that shields many FFS Medicare beneficiaries will add to program spending. The Congressional from coinsurance, the benefit for some beneficiaries with Budget Office estimates that the policy would increase medigap policies in rural states would be a small reduction spending by less than $50 million per year. in medigap premiums. Beneficiaries and providers RECOMMENDATION 2-1 Rural communities would have a new option for • preserving emergency department access without The Congress should: having to maintain expensive inpatient capacity. allow isolated rural stand-alone emergency • Medicare beneficiaries would benefit from preserved departments (more than 35 miles from another local access to emergency care and the reduced emergency department) to bill standard outpatient coinsurance. prospective payment system facility fees and • provide such emergency departments with annual payments to assist with fixed costs. Urban areas: Incentives have led to an RATIONALE 2-1 abundance of urban stand-alone EDs Struggling hospitals within 35 miles of another hospital The number of stand-alone EDs and the share of patient can eliminate inpatient services and reduce their costs visits taking place in EDs have increased rapidly in recent by becoming an outpatient department of a neighboring years. These facilities improve access to services not hospital. However, isolated rural facilities more than 35 available at doctors’ offices and urgent care centers, but miles from another hospital do not have an option to their Medicare payment rates need to be better aligned convert to a stand-alone emergency department. Therefore, with the cost of care they provide. communities that most need an emergency room but cannot support inpatient services also have the fewest Some researchers believe the growth in ED use may be payment options. This situation results in stand-alone EDs partially due to patients’ lack of access to other providers, being financially unviable in most isolated rural markets. changing practice patterns, or patients’ desire for more Creating a way to pay stand-alone EDs in isolated rural immediate access to care (Gindi et al. 2016, Morganti communities will help these areas maintain emergency et al. 2013, Pines et al. 2013). However, the increase in department capacity. The option would be available to the number of stand-alone EDs and the increase in the communities more than 35 miles from another hospital volume of ED visits may also partly reflect incentives in June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 53

76 TABLE Seventy-five percent of urban stand-alone emergency departments 2–4 are located within 6 miles and a 10-minute drive of the nearest on-campus hospital emergency department, 2018 Distance to the nearest on-campus hospital ED (in miles) 0–2 2–4 4–6 6–8 8–10 10–12 12 or more 13 Number of stand-alone EDs 5 5 5 23 26 35 Share of stand-alone EDs 31% 23% 12% 4% 4% 4% 21% Cumulative share 21% 52% 75% 87% 91% 96% 100% Average minutes from the nearest 10.3 on-campus hospital ED 21.6 8.4 19.8 14.3 4.4 14.0 Note: ED (emergency department). The five market areas include Charlotte, NC; Cincinnati, OH; Dallas, TX; Denver, CO; and Jacksonville, FL. Components may not sum to totals due to rounding. MedPAC analysis of the location of hospitals and stand-alone EDs using ArcGIS data software and Google mapping. Source: both Medicare’s payment system and commercial insurer share and control patient service use. They also stated that payment systems (Wilson and Cutler 2014). Recent a real estate analysis method—using variables such as analysis from three states suggests that stand-alone EDs the location of other EDs, population growth, household treat patients who are more similar to patients treated at income, and insurance coverage—is used to identify areas urgent care centers than patients treated at on-campus with unmet demand for convenient ED services (Adeptus hospital EDs. Despite this analysis, under Medicare, Health Inc. 2016). OCEDs are paid the same as on-campus hospital EDs, Urban stand-alone EDs are in close making the OCED model of care financially attractive to proximity to on-campus hospital EDs hospitals in many markets. Our analysis of stand-alone EDs sought to distinguish Stand-alone EDs locate in certain markets between urban stand-alone EDs that provide access to and higher income zip codes rather than urban areas that are relatively isolated from ED services underserved areas and stand-alone EDs that create redundancies in access The stand-alone EDs identified in our June 2017 report because they are in close proximity to existing on-campus were concentrated in 20 large metropolitan statistical areas hospital EDs. We examined five markets with urban stand- (MSAs) in 2016 and accounted for over 60 percent of all alone EDs (Charlotte, NC; Cincinnati, OH; Dallas, TX; 11 stand-alone EDs. These facilities tend to locate in zip Denver, CO; and Jacksonville, FL) and considered the codes with higher than average incomes and higher shares distance of stand-alone EDs from the nearest on-campus of patients with private insurance coverage (Medicare hospital ED, both in driving distance (in miles) and driving Payment Advisory Commission 2017, Schuur et al. 2016). time (in minutes). While we measured proximity as the We found that, in Houston and Denver, about 65 percent of distance to an on-campus hospital ED, policymakers could stand-alone EDs were located in zip codes that represented also opt to measure proximity from the stand-alone EDs to only 35 percent of the city’s population but had an average any other ED (on-campus ED or other stand-alone ED). household income above $90,000. We found similar Overall, our analysis found that stand-alone EDs tend to patterns in Charlotte, NC; Jacksonville, FL; Oklahoma be located in close proximity to on-campus hospital EDs. City, OK; and Seattle, WA (markets without IFECs). In 2018, 75 percent of urban stand-alone EDs in the five Recent research has found that IFECs may be even more markets studied were within six miles of the nearest on- likely to locate in high-income areas (Dark et al. 2017). campus hospital ED, and 25 percent were more than six In interviews, stand-alone ED representatives stated that miles from the nearest on-campus hospital ED (Table 2-4). hospitals use stand-alone EDs to capture patient market Using payment to ensure appropriate access to and use of hospital emergency department services 54

77 In addition, using publicly available mapping software, we This study found that the standby costs of stand-alone estimated that, on average, the EDs within 6 miles of the EDs fall between the costs of on-campus hospital EDs and urgent care centers (Ho et al. 2017). Stand-alone EDs and nearest on-campus hospital ED were roughly a 10-minute drive from the nearest on-campus hospital ED. Therefore, on-campus hospital EDs must provide continuous access a beneficiary living exactly in between a stand-alone ED to emergency clinicians, laboratory services, and imaging services. The cost of meeting these requirements is higher and an on-campus hospital ED six miles apart would need to travel three miles, or spend five minutes to drive, to the than the costs at urgent care centers, which typically are ED nearest their residence. not open 24/7 and are generally not staffed with physicians specializing in emergency medicine. While the costs of Patients served at stand-alone EDs in three stand-alone EDs are higher than urgent care centers, the states have lower acuity than patients at on- authors also contend that their costs are lower than on- campus EDs campus hospital EDs, in part because stand-alone EDs largely do not maintain the on-call physician capacity for Three recent analyses of stand-alone EDs in Texas, specialists necessary to serve patients with major trauma Colorado, and Maryland demonstrate that patients served injuries (e.g., head and neck wounds or gunshot wounds), by stand-alone EDs tend to have less complex conditions stroke, and ST-elevation myocardial infarctions. The than patients served by on-campus ED patients, but the authors suggest this difference in patient severity is linked prices paid to the stand-alone EDs are typically the same to the fact that ambulances preferentially route higher as on-campus EDs. Moreover, the analyses highlight that acuity patients to on-campus hospital EDs that maintain stand-alone EDs generally do not incur the same standby operating rooms and overnight inpatient bed capacity. In costs as on-campus EDs. our interviews with ambulance industry representatives, Texas they confirmed this view, stating that ambulance drivers are instructed to take any potential inpatient admission A study of commercial insurance claims for enrollees of to an on-campus hospital ED because stand-alone EDs Blue Cross and Blue Shield of Texas from 2012 to 2015 do not have operating rooms or overnight beds. Another suggests that stand-alone EDs serve patients who are study found that ambulance drivers transporting trauma similar to those served by urgent care centers while being cases typically bypassed an isolated rural stand-alone ED paid rates similar to hospital EDs (Ho et al. 2017). This because on-campus hospital EDs were more likely to have study found substantial overlap in the type of cases seen at trauma care capacity (Lawner et al. 2016). on-campus EDs, off-campus EDs, and urgent care centers, but it also found that on-campus EDs tend to receive the Colorado most difficult cases, such as open head or neck wounds. Among the top 20 most common diagnoses treated at Claims data for privately insured patients in Colorado 12 each facility type, 12 diagnoses were common to all 3. in 2014 show that most patients served by stand-alone EDs were treated for non-life-threatening conditions, Three diagnoses at on-campus hospital EDs were not similar to conditions treated at urgent care centers. These common to either stand-alone EDs or urgent care centers: data also show that the patients served by stand-alone kidney stones, nausea and vomiting, and complications of EDs are somewhat different from those served at on- pregnancy. Five of the most common diagnoses at urgent campus hospital EDs. In July 2016, Colorado’s Center care centers were not common to either stand-alone EDs for Improving Value in Health Care (CIVHC) compared or on-campus hospital EDs: eye inflammation, flu, other claims data from nine stand-alone EDs with claims from upper respiratory disease, pneumonia, and viral infections. urgent care centers and on-campus hospital EDs. CIVHC All of the most common diagnoses at stand-alone EDs concluded that, among the top 10 conditions for which were also most common to on-campus EDs or urgent care privately insured patients sought care at stand-alone EDs, centers. Despite the similarity in cases treated across the 7 were for non-life-threatening conditions. At urgent care three facility types, stand-alone EDs appear to occupy a centers, all 10 of the top 10 conditions were non–life middle ground between urgent care centers and on-campus threatening, whereas at on-campus hospital EDs, only 3 EDs with regard to the severity of patients they serve. For of the top 10 were for non-life-threatening conditions. example, more acute medical diagnoses such as pregnancy Between stand-alone EDs and urgent care centers, six complications and kidney stones and less complicated of the most common conditions overlapped, and none medical diagnoses such as eye inflammation and viral of them were life threatening. Between stand- infections are not common to stand-alone EDs. alone EDs June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 55

78 and on-campus hospital EDs, four of the most common more often transported to on-campus hospital EDs where conditions overlapped, and three were non–life threatening. there are backup surgical capabilities, operating rooms, cardiac reperfusion capabilities, and specialized stroke Using the same data, CIVHC found that, in 2014, privately care. Other research has reported ambulances bypassing insured patients paid higher amounts—exceeding 10 times stand-alone EDs, specifically studies examining the stand- the amount—for treatment at stand-alone EDs compared alone ED phenomenon in Maryland (Lawner et al. 2016, with treatment at urgent care centers. For example, Maryland Health Care Commission 2015). Rural EDs in 2014, the average payment amount for an upper that are especially far from other care are the exception; respiratory infection (a non-life-threatening condition) in these areas, ambulances might rely on rural EDs to at stand-alone EDs was $1,114, compared with $124 at stabilize trauma patients, and in some cases might use urgent care centers. Similar differences existed for other them as a location to begin clot-busting drugs on stroke 13 conditions. patients. This exception suggests that isolated off-campus EDs that are a substantial distance from any hospital-based Maryland ED can be called on to have a wider range of standby A 2015 report from the Maryland Health Care capacity than OCEDs located 10 or 15 minutes from a Commission (MHCC) concluded that the patients served hospital campus. by three stand-alone EDs generally had lower acuity conditions compared with on-campus EDs (Maryland Aligning payments to urban stand-alone EDs Health Care Commission 2015). MHCC reported that, in with the acuity of their patients 2014, between 3 percent and 6 percent of patients served The growth in stand-alone EDs in recent years suggests by the three stand-alone EDs were later admitted as that existing Medicare and private-insurer payment inpatients to a hospital compared with between 15 percent policies encourage providers to treat patients in higher and 19 percent of patients served at the nearest competing paying settings such as EDs rather than lower paying hospital EDs. In addition, at the Maryland stand-alone settings such as urgent care centers. The Commission’s EDs in two towns, 97 percent and 95 percent of patients, position on aligning payment rates across settings is respectively, arrived as walk-ins rather than by ambulance. that Medicare should ensure that patients have access to By contrast, the Emergency Department Benchmarking settings that provide the appropriate levels of care and Alliance and the American College of Emergency that Medicare should base payment rates on the resources Physicians reported that, in 2013, 17 percent of all ED needed to treat patients in the most efficient setting. The patients nationally arrived at the ED by ambulance concern in the case of stand-alone EDs is that providers (Augustine 2014). seek to gain market share for low-severity conditions that could be treated more efficiently in other settings. MHCC also concluded that patients served by the three For example, some hospitals are building ED facilities or Maryland stand-alone EDs in 2014 were younger, partnering with IFECs to enable them to bill for the higher more likely to have private insurance coverage, and had ED rates, when these conditions could be treated at urgent treatment options other than the ED available to them. care centers. Compared with all EDs in Maryland, the stand-alone EDs tended to treat a larger share of children and a smaller Options for paying urban OCEDs less than full share of patients older than 41, tended to serve a slightly Type A ED rates larger share of privately insured patients, and tended to To account for the lower standby costs and the lower serve a lower share of Medicare and Medicaid patients. acuity of patients served at OCEDs, the Commission considered two alternatives to current Type A ED payment Required standby capacity of urban stand- rates. The Commission’s intent was to better align alone EDs is less than that of on-campus payment rates with the costs of OCEDs, thereby reducing hospital EDs the incentive to shift lower acuity cases to the ED setting. Information gathered from site visits to stand-alone These two alternative payment rates were intended to lie EDs and recent research supports ambulance suppliers’ between the rates of on-campus hospital EDs and urgent statements that stand-alone EDs generally do not maintain care centers. the capacity to treat major trauma cases such as major head injuries. Trauma, stroke, and heart attack patients are In public discussion, the Commission initially considered paying OCEDs Type B ED rates because the acuity of Using payment to ensure appropriate access to and use of hospital emergency department services 56

79 their patients is similar to the mix of patient conditions alone EDs tend to treat lower intensity patients and incur served at EDs receiving Medicare Type B ED payment less standby capacity costs than on-campus hospital EDs rates. However, current Type B payment rates contain because they generally do not maintain services such as trauma care or operating rooms. To better align their an anomalous characteristic that results in payments for Type B Level 1 cases (the lowest level) being higher than payments and costs, Medicare should pay OCEDs at lower Type B Level 2 cases ($102 for Level 1 cases versus $91 rates than on-campus hospital EDs, but at higher rates than for Level 2 cases) and higher than Type A Level 1 cases urgent care centers. ($102 for Type B Level 1 cases versus $69 for Type A However, paying the current higher Type A ED payment Level 1 cases) (Table 2-2, p. 41). This anomaly causes rates to urban OCEDs that are not in close proximity the difference between Type A and Type B payment rates 15 to on-campus EDs may be justified. These more 14 to vary widely across each of the five ED levels. On isolated OCEDs are more likely to be providing their average, across all five ED service levels, Type B rates are local community with unique access to ED services. 30 percent lower than Type A rates. The Commission estimates that 25 percent of urban stand-alone EDs are located more than six miles from To establish payment rates for OCEDs that lie between an on-campus hospital ED, and 75 percent are located those for on-campus hospital EDs and urgent care centers, within six miles. In response to industry concerns and while reducing payments consistently across the five levels for operational simplicity, the Commission used a of ED services, Medicare should reduce Type A ED rates threshold based on the measurement of distance in road by a flat percentage. Reducing the Type A rates by 30 miles rather than driving time, and the six-mile threshold percent would be roughly equivalent to using Type B rates appeared to be a natural breaking point in the proximity and would avoid the anomaly in the Type B rates. See data. In addition, the Commission found that the 6-mile online Appendix 2-A, available at http://www.medpac.gov, distance translated into roughly a 10-minute drive. for a summary of the proposed urban policy. Our six-mile proximity threshold could result in stand- Urban stand-alone ED recommendation alone EDs locating just beyond the six-mile threshold and in relatively close proximity to other stand-alone Urban OCEDs may provide the benefit of some services EDs. To avoid this dynamic, should the Commission’s that are not available at urgent care centers and doctors’ recommendation be implemented, policymakers might offices, but Medicare appears to be overpaying these consider an alternative measure of proximity as the facilities relative to what is paid to on-campus hospital distance between the stand-alone ED and any other ED EDs that receive more difficult cases. While most urban (on campus or stand alone). stand-alone EDs are in close proximity to on-campus hospital EDs, some are located far from on-campus The Commission’s recommendation to reduce payment hospital EDs and likely provide unique access to ED rates to OCEDs is intended to align payment rates with services for their local community. Paying these more the relative costs of OCEDs. Timely congressional isolated urban stand-alone EDs higher Type A rates, with action in response to this recommendation would help no percentage reduction applied, may be justified. ensure that hospital systems do not invest significant amounts of capital in OCEDs that are not necessary RECOMMENDATION 2-2 16 to ensure appropriate access to emergency care. Our The Congress should reduce Type A emergency recommendation to reduce payment rates to certain department payment rates by 30 percent for off-campus urban OCEDs by 30 percent, making those rates more stand-alone emergency departments that are within six comparable with Type B payment rates, may reduce the miles of an on-campus hospital emergency department. 17 incentive to invest in such facilities. The 30 percent reduction reflects the current best information available, RATIONALE 2-2 but we note that the Secretary of Health and Human The structure of the Medicare payment system for ED Services could be given the authority to gather additional services creates incentives for providers to treat lower information on OCEDs’ Medicare claims data and intensity patients in EDs rather than at urgent care centers, OCEDs’ costs. This information will enable policymakers which are paid less than half the Type A payment rates to adjust the 30 percent reduction in the future as new for ED services. The Commission found that urban stand- information becomes available and the marketplace shifts. June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 57

80 IMPLICATIONS 2-2 To gather the necessary claims and cost data on OCEDs, policymakers must make two specific administrative Spending changes to hospital-related datasets. First, Medicare will Medicare payment rates for the five levels of ED • need to identify OCEDs’ Medicare claims. In 2016, the services would each decline by 30 percent for Commission recommended that “the Congress should urban off-campus EDs located within six miles of require hospitals to add a modifier on claims for all an on-campus hospital ED. Urban off-campus EDs services provided at off-campus stand-alone ED facilities” located more than six miles from an on-campus ED (Medicare Payment Advisory Commission 2016b). To would see no change in payment for ED services. date, this recommendation has not been enacted. Second, The Congressional Budget Office estimates that Medicare will need to require hospitals to report the costs this policy would result in an overall reduction in of OCEDs separately on annual hospital cost reports made Medicare outpatient hospital spending of between $50 to CMS. Once OCED claims can be tracked and OCED million and $250 million annually. Over five years, cost and charge data gathered, CMS could estimate the this policy could result in a reduction to Medicare relative costs of on-campus EDs and OCEDs. At that outpatient hospital spending of less than $1 billion. point, the Secretary could modify the magnitude of the This reduction represents less than 1 percent of total recommended 30 percent reduction to Type A ED payment Medicare outpatient hospital spending. rates. Beneficiaries and providers The Commission has made a judgment in determining • Medicare beneficiaries served at urban OCEDs located that OCEDs located farther than six miles from an on- within six miles of an on-campus hospital ED would campus ED should be paid the full Type A rates. Other, have lower cost sharing. In addition, this policy would more restrictive options could be considered. One option reduce the incentive to develop new OCEDs in close is to limit the full Type A rates to EDs more than six miles proximity to on-campus hospital EDs. By leaving from any ED (including other OCEDs). This option would Medicare payment rates unchanged at urban OCEDs prevent a clustering of OCEDs. A second option is to located more than six miles from an on-campus ED, impose a moratorium on new OCEDs. A third option is Medicare would continue to ensure access to ED to reduce payment rates for non-ED services at OCEDs, services in areas where they are needed most. such as paying office visits at affiliated clinics the rate The implications of this policy for hospitals and • paid to freestanding physician offices. This option would hospital systems is that 75 percent of existing urban eliminate the exception written into Section 603 of the OCEDs will see a 30 percent decline in payments for BBA of 2015, which requires that both ED and non-ED ED services. The remaining 25 percent of OCEDs, services (e.g., clinic visits and ancillary services) provided those located more than six miles from an on-campus in off-campus EDs be paid the higher OPPS rates. The ED, will not see a change in payment for ED services. Commission also discussed a less restrictive option, in which OCEDs within six miles of an on-campus ED could continue to receive full Type A ED payment rates if they operated in a location where a hospital closed. Future analyses The Commission has expressed interest in future research concerning the standby costs of different types of EDs and Medicare payment rates for urgent care centers and micro- hospitals. That research could lead to better alignment of payment rates for on-campus hospital EDs, OCEDs, urgent care centers, and micro-hospitals. The objective would be to create incentives for providers to use the appropriate setting to treat patients’ needs. ■ Using payment to ensure appropriate access to and use of hospital emergency department services 58

81 Endnotes broader definition used by the Federal Office of Rural Health 1 Data separating Medicare and non-Medicare ED use for Policy, which incorporate non-MSAs and rural portions of 2015 and 2016 were not available at the time of publication. counties within MSAs. Therefore, all-payer data were used to demonstrate the trend in outpatient ED use from 2010 to 2016. 10 A few rural facilities currently operate stand-alone EDs with an attached outpatient clinic. A study by the University of 2 Hospitals’ ED claims that result in a hospital admission are North Carolina estimates that the cost of operating a low- bundled into a diagnosis related group and paid through the volume 24/7 ED facility with an attached outpatient clinic inpatient prospective payment system. is about $5 million per year (Williams et al. 2015). Our The relative weights placed on Type A payment rates are 3 discussions with rural hospital accountants and administrators based on the geometric mean cost of services in Type A of small rural stand-alone EDs support estimates in this range. EDs relative to the average cost of a clinic visit. The relative We defined large MSAs as those with 500,000 or more 11 weights placed on Type B payment rates are based on the residents in 2015. In 2017, stand-alone EDs were located in geometric mean cost of services in Type B EDs, which tend to 95 MSAs and 35 states. be lower on average. The 12 diagnoses common to stand-alone EDs, on-campus 12 The anomaly in which Type B Level 1 ED visits are paid 4 hospital EDs, and urgent care centers were abdominal more than Type A Level 1 ED visits is due to the idiosyncratic pain, allergic reactions, bronchitis, wounds, connective cost and charge structure of the few hospitals billing Type B tissue disease, lower respiratory disease, upper respiratory rates. infections, skin infections, back problems, sprains, superficial Older urgent care centers affiliated with a hospital are still 5 injuries, and urinary tract infections. permitted to bill hospital OPPS rates, which are on par with 13 Private insurers in Colorado pay stand-alone EDs more for what the Type B facilities receive for an ED visit. They were other services associated with non-life-threatening conditions grandfathered under a new site-neutral policy that eliminated compared with the same services at urgent care centers, facility fees for new hospital-affiliated urgent care centers and including abdominal pain—other specified site ($5,635 vs. physician practices (Medicare Payment Advisory Commission $151, respectively), bronchitis ($1,139 vs. $123, respectively), 2016b). sinus infection ($786 vs. $125, respectively), and open finger The number of urgent care centers was obtained from the 6 wounds ($1,035 vs. $134, respectively). These high private- Urgent Care Association of America’s website at http://www. payer payments to stand-alone EDs in Colorado are consistent ucaoa.org/?page=IndustryFAQs#Size%20of%20Industry. with data from Blue Cross Blue Shield of Texas (Ho et al. 2017) and with anecdotal reports in the popular press in other Provider-based ED facilities are eligible for Medicare 7 states (Kliff 2018). payment if they are in compliance with Medicare’s provider- based department regulations, Medicare’s conditions of The difference between Type A ED payment rates and Type B 14 participation, and the requirements of the Emergency Medical ED payment rates varies by level of ED service. Type B Level Treatment and Active Labor Act. 1 payment rates are 49 percent higher than Type A Level 1 rates. Type B Level 2 payment rates are 27 percent lower than Section 603 defines dedicated EDs as facilities at which at 8 Type A Level 2 rates. Type B Level 3 payment rates are 28 least one-third of a facility’s outpatient visits for the treatment percent lower than Type A Level 3 rates. Type B Level 4 rates of emergency medical conditions are on an urgent basis are 41 percent lower than Type A Level 4 rates. Type B Level without requiring a previously scheduled appointment. 5 payment rates are 45 percent lower than Type A Level 5 rates. 9 as all areas outside of metropolitan We generally define rural statistical areas (MSAs). This definition of rural includes 15 Policymakers may identify other situations where higher micropolitan areas. Others have a broader definition of rural payments to urban OCEDs are warranted—for example, areas that includes some small towns within MSAs. For when an urban OCED is the result of the closure of its parent example, others may categorize towns as rural if they are hospital. outside the commuting zone of larger cities, even if the county they are located in is considered part of an MSA. Given 16 The Commission’s goal of adjusting payment rates to prevent these different definitions of rural, we present information on the misallocation of capital based on mispriced services is hospital closures using both our definition (non-MSA) and the not new. In earlier years, the Commission recommended | Report to the Congress: Medicare and the Health Care Delivery System June 2018 59

82 17 The extent to which the incentive to invest in OCEDs is changing the inpatient prospective payment system to prevent reduced by a Medicare payment policy change would depend overpayment to specialty hospitals treating relatively easier on the share of a given OCED’s revenues that are tied to cases (Medicare Payment Advisory Commission 2006, Medicare patient visits. Medicare Payment Advisory Commission 2005). Using payment to ensure appropriate access to and use of hospital emergency department services 60

83 References Ho, V., L. Metcalfe, C. Dark, et al. 2017. Comparing utilization Adeptus Health Inc. 2016. Form 10–K annual report submitted to and costs of care in freestanding emergency departments, hospital the Securities and Exchange Commission. http://d1lge852tjjqow. Annals of emergency departments, and urgent care centers. cloudfront.net/CIK-0001602367/1421ea7d-761a-4286-a10c- Emergency Medicine (February 15). 1df276639ca6.pdf?noexit=true. Iglehart, J. K. 2018. The challenging quest to improve rural health American Hospital Association. 2016. Task Force on Ensuring New England Journal of Medicine 378, no. 5 (February 1): care. Access in Vulnerable Communities. Chicago, IL: AHA. 473–479. Andrews, M. 2016. Sometimes tiny is just the right size: Kliff, S. 2018. An ER visit, a $12,000 bill—and a health insurer , “Microhospitals” filling some ER needs. Kaiser Health News . January 29. Vo x that wouldn’t pay. July 19. http://khn.org/news/sometimes-tiny-is-just-the-right-size- microhospitals-filling-some-er-needs/. Lawner, B. J., J. M. Hirshon, A. C. Comer, et al. 2016. The impact of a freestanding ED on a regional emergency medical services Ashwood, J. S., M. Gaynor, C. M. Setodji, et al. 2016. Retail American Journal of Emergency Medicine system. 34, no. 8 clinic visits for low-acuity conditions increase utilization and (August): 1342–1346. Health Affairs spending. 35, no. 3 (March 1): 449–455. Liu, J. J., G. Bellamy, B. Barnet, et al. 2008. Bypass of local Augustine, J. 2014. Emergency medical services arrivals, primary care in rural counties: Effect of patient and community ACEP admission rates to the emergency department analyzed. characteristics. 6, no. 2 (March– Annals of Family Medicine , December 17. Now April): 124–130. Baker, L. C., and L. S. Baker. 1994. Excess cost of emergency Livingston, S. 2018. UnitedHealth tightens reins on emergency department visits for nonurgent care. Health Affairs 13, no. 5 , March 7. Modern Healthcare department reimbursement. (Winter): 162–171. Report on the Maryland Health Care Commission. 2015. Centers for Medicare & Medicaid Services, Department of Health operations, utilizations, and financial performance of and Human Services. 2008. Memorandum from the Director Annapolis, MD: MHCC. freestanding medical facilities. of the Survey and Certification Group regarding requirements for provider-based off-campus emergency departments and Medicare Payment Advisory Commission. 2018. Report to the hospitals that specialize in the provision of emergency services. Washington, DC: MedPAC. Congress: Medicare payment policy. January 11. https://www.cms.gov/Medicare/Provider-Enrollment- and-Certification/SurveyCertificationGenInfo/downloads/ Report to Medicare Payment Advisory Commission. 2017. scletter08-08.pdf. the Congress: Medicare and the health care delivery system. Washington, DC: MedPAC. Community Health Systems. 2017. Form 10–K annual report submitted to the Securities and Exchange Commission. February 26. Medicare Payment Advisory Commission. 2016a. Report to the Congress: Medicare and the health care delivery system. Dark, C., Y. Xu, and V. Ho. 2017. Freestanding emergency Washington, DC: MedPAC. departments preferentially locate in areas with higher household income. Health Affairs 36, no. 10 (October 1): 1712–1719. Report to the Medicare Payment Advisory Commission. 2016b. Washington, DC: MedPAC. Congress: Medicare payment policy. Reasons for emergency Gindi, R., L. Black, and R. Cohen. 2016. room use among US adults aged 18–64: National health interview Medicare Payment Advisory Commission. 2012. Report to . National Health Statistics Reports, no. 90. survey 2013 and 2014 the Congress: Medicare and the health care delivery system. National Center for Health Statistics, Centers for Disease Control Washington, DC: MedPAC. and Prevention, Department of Health and Human Services. Hyattsville, MD: NCHS. February 18. Medicare Medicare Payment Advisory Commission. 2011. copayments for critical access hospital outpatient services—2009 Glatter, R. 2017. In 3 states, if Anthem thinks you shouldn’t have Report prepared by staff from RTI International for the update. , October 16. gone to ER, it won’t pay. Forbes Medicare Payment Advisory Commission. Washington, DC: MedPAC. 2016 health care cost and Health Care Cost Institute. 2018. Washington, DC: HCCI. utilization report. | June 2018 Report to the Congress: Medicare and the Health Care Delivery System 61

84 Medicare Payment Advisory Commission. 2006. Report to Schuur, J. D., O. Baker, J. Freshman, et al. 2016. Where do the Congress: Physician-owned specialty hospitals revisited. freestanding emergency departments choose to locate? A national Washington, DC: MedPAC. Annals of inventory and geographic analysis in three states. Emergency Medicine (July 12). Report to the Medicare Payment Advisory Commission. 2005. Washington, DC: Congress: Physician-owned specialty hospitals. Sutherly, B. 2016. Free-standing emergency departments will MedPAC. , February 29. drive up costs, some warn. Columbus Dispatch Mehrotra, A., H. Liu, J. L. Adams, et al. 2009. Comparing costs Thompson, B. 2015. Rural Kansas hospitals search for ways to and quality of care at retail clinics with that of other medical , June 29. KHI News Service/Heartland Health Monitor survive. 151, Annals of Internal Medicine settings for 3 common illnesses. http://kcur.org/post/rural-kansas-hospitals-search-ways-survive. no. 5 (September 1): 321–328. Thygeson, M., K. A. Van Vorst, M. V. Maciosek, et al. 2008. Use Morganti, K., S. Bauhoff, J. Blanchard, et al. 2013. The evolving and costs of care in retail clinics versus traditional care sites. . Santa role of emergency departments in the United States 27, no. 5 (September–October): 1283–1292. Health Affairs Monica, CA: RAND Corporation. UnitedHealth Center for Health Reform & Modernization. 2011. Morningstar Document Research. 2017a. HCA Holdings, for Modernizing rural health care: Coverage, quality and innovation . 10–K. February 22. Working paper 6. Minnetonka, MN: UnitedHealth Group. Morningstar Document Research. 2017b. Tenet Healthcare Weinick, R. M., R. M. Burns, and A. Mehrotra. 2010. Many Corporation, for 10–K. February 27. emergency department visits could be managed at urgent care 29, no. 9 (September): Health Affairs centers and retail clinics. Morse, S. 2015. Kansas plan would convert many rural hospitals 1630–1636. HealthCare Finance News into primary health clinics. , August 18. Williams, J. D., P. H. Song, and G. H. Pink. 2015. Estimated costs Pines, J. M., P. M. Mullins, J. K. Cooper, et al. 2013. National Findings brief. of rural freestanding emergency departments. trends in emergency department use, care patterns, and quality of Chapel Hill, NC: North Carolina Rural Health Research Program, care of older adults in the United States. Journal of the American Cecil G. Sheps Center for Health Services Research, University 61, no. 1 (January): 12–17. Geriatrics Society of North Carolina at Chapel Hill. Rice, S. 2016. Are free-standing emergency rooms helping only Wilson, M., and D. Cutler. 2014. Emergency department profits Dallas Morning News the wealthy? , August 2. are likely to continue as the Affordable Care Act expands coverage. Health Affairs 33, no. 5 (May): 792–799. Rudavsky, S. 2016. St. Vincent to open 4 emergency micro- hospitals. Indy Star , August 22. Young, S. 2018. Personal communication with Sarah Young, Federal Office of Rural Health Policy. Rural Hospital Stabilization Committee, State of Georgia. 2015. Rural Hospital Stabilization Committee: Final report to the governor . Atlanta, GA: RHSC. https://gov.georgia.gov/sites/gov. georgia.gov/files/related_files/document/Rural%20Hospital%20 Stabilization%20Committee%20Report%20022315%20FINAL. pdf. Using payment to ensure appropriate access to and use of hospital emergency department services 62

85 CHAPTER 3 Rebalancing Medicare’s physician fee schedule toward ambulatory evaluation and management services

86

87 CHAPTER 3 Rebalancing Medicare’s physician fee schedule toward ambulatory evaluation and management services In this chapter Chapter summary Ambulatory evaluation and management (E&M) services, such as office Background on the fee • and hospital outpatient visits, are essential for a high-quality, coordinated schedule for physician and health care delivery system. These visits enable clinicians to diagnose and other health professional services manage patients’ chronic conditions, treat acute illnesses, develop care plans, coordinate care across providers and settings, and discuss patients’ • Ambulatory E&M services preferences. E&M services are critical for both primary care and specialty are underpriced relative to care. The Commission is concerned that these services are underpriced in the other services fee schedule for physicians and other health professionals (“the fee schedule”) An approach to rebalance • relative to other services, such as procedures. This mispricing may lead to the fee schedule toward problems with beneficiary access to these services and, over the longer term, ambulatory E&M services may even influence the pipeline of physicians in specialties that tend to provide a large share of E&M services. • Conclusion Payment rates in the fee schedule are based on relative weights, called relative value units (RVUs), which account for the amount of work required to provide a service, expenses related to maintaining a practice, and professional liability insurance costs. Work RVUs are based on an assessment of how much time and intensity (e.g., mental effort and technical skill) services require relative to one another. If estimates of time and intensity are not kept up to date, especially for services that experience efficiency improvements, the work RVUs become inaccurate. Because of advances in technology, technique, and clinical practice, efficiency improves more easily for procedures, imaging, and June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 65

88 tests than for ambulatory E&M services, which are composed largely of activities that require the clinician’s time and so do not lend themselves to efficiency gains. When efficiency gains reduce the amount of work needed for a service, the work RVUs for the affected services should decline accordingly. Because the fee schedule is budget neutral, a reduction in the RVUs of these services would raise the RVUs for all other services, such as ambulatory E&M services. Because of problems with the process of reviewing overpriced services, this two-step sequence tends not to occur. Therefore, ambulatory E&M services become passively devalued over time. CMS, with input from the American Medical Association/Specialty Society Relative Value Scale Update Committee (RUC), has reviewed the work RVUs of many potentially mispriced services since 2009. However, CMS’s review has taken several years and has not yet addressed services that account for a substantial share of fee schedule spending. CMS’s review is hampered by the lack of current, accurate, and objective data on clinician work time and practice expenses. To estimate clinician work time for specific services, CMS relies on data from surveys conducted by specialty societies that are reviewed by the RUC. We have concerns about these data; for example, the surveys have low response rates and low total number of responses, which raises questions about the representativeness of the results. To address this problem, the Commission previously recommended that CMS use a streamlined method to regularly collect data from a cohort of efficient practices —including service volume and work time—to establish more accurate work and practice expense RVUs. These data should be used in a “top-down” approach to calculate the amount of time that a physician worked over the course of a week or month and compare it with the time estimates in the fee schedule for all of the services that the physician billed over the same period. If the fee schedule’s time estimates exceed the actual time worked, this finding could indicate that the time estimates are too high. Contractors working for CMS and the Assistant Secretary for Planning and Evaluation in the Department of Health and Human Services found that the fee schedule’s time estimates for clinician work for a broad range of services— particularly imaging, procedures, and tests—are inflated when compared with ambulatory E&M services. Indeed, errors in some of the fee schedule’s time assumptions were very large—multiples of the actual time spent by physicians. For example, the time assumption for MRI of the brain was more than twice as high as the actual time spent by physicians on this service, according to a physician survey. By contrast, the time assumption for three ambulatory E&M services in the survey was about the same as the actual time spent by physicians. Rebalancing Medicare’s physician fee schedule toward ambulatory evaluation and management services 66

89 There is also evidence that payment rates for global surgical services—which include the procedure itself and certain services that are provided immediately before and after the procedure—are too high. The global payment rate assumes that the same physician who performs the procedure also provides all the postoperative care, such as E&M visits. However, a study by the RAND Corporation observed that postoperative care is shifting from the physician who performed the procedure to other clinicians, such as hospitalists and nonphysician practitioners, who bill separately for each postoperative visit. This change suggests that physicians who bill for a global surgical service may be receiving payments for postoperative visits that in reality are provided by other clinicians. In addition, the Office of Inspector General reviewed medical records for several types of global surgical services and found that physicians frequently provided fewer E&M visits during the postoperative period than were included in the global payment rate. There are also major problems with the accuracy of the data used to set practice expense RVUs (practice expense includes the cost of nonphysician clinical and administrative staff, medical equipment and supplies, office rent, and other expenses). First, CMS does not have a comprehensive data source with current information on the prices of medical equipment and supplies; consequently, the price estimates for these items are often outdated. Second, practice expense RVUs are based on data from a survey of total practice costs. Because this survey was conducted in 2007 and 2008, the data are unlikely to reflect current practice costs. We describe a budget-neutral approach to rebalance the fee schedule that would increase payment rates for ambulatory E&M services while reducing payment rates for other services (e.g., procedures, imaging, and tests). Under this approach, the increased payment rates would apply to ambulatory E&M services provided by all clinicians, regardless of specialty. We modeled the impact of a 10 percent payment rate increase for ambulatory E&M services, although a higher or lower increase could be considered. A 10 percent increase would raise annual spending for ambulatory E&M services by $2.4 billion. To maintain budget neutrality, payment rates for all other fee schedule services would be reduced by 3.8 percent. Certain specialties would receive a large increase in their total fee schedule payments (on net) as a result of this change. The three specialties that would receive the highest proportional increase in payments are endocrinology (6.6 percent net increase in fee schedule payments), rheumatology (5.5 percent increase), and family practice (4.9 percent increase). These specialties concentrate on ambulatory E&M services. Several specialties—including diagnostic radiology, pathology, physical therapy, and occupational therapy—would experience reductions in their fee June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 67

90 schedule payments of about 3.8 percent because they provide very few ambulatory E&M services. This change would be a one-time adjustment to the fee schedule to address several years of passive devaluation of ambulatory E&M services. Even if this approach is adopted, we urge CMS to accelerate its efforts to improve the accuracy of the fee schedule by developing a better mechanism to identify overpriced services and adjust their payment rates. If successful, these efforts would improve the accuracy of prices for ambulatory E&M and other services going forward and could reduce the need for future significant adjustments to the prices of E&M services. ■ Rebalancing Medicare’s physician fee schedule toward ambulatory evaluation and management services 68

91 improve care coordination and quality while controlling Background on the fee schedule for cost growth. physician and other health professional However, it is still important to ensure the accuracy services of fee schedule prices under traditional FFS Medicare because many beneficiaries remain in traditional FFS. In 2016, Medicare paid about $70 billion under the fee In addition, all A–APM models use FFS payment rates schedule for physician and other health professional as either the basis of payment or the reference price for services (“the fee schedule”). The fee schedule contains setting the global or bundled payment amount. Further, payment rates for over 7,000 distinct services, classified the benchmarks used to determine payments to Medicare using the Healthcare Common Procedure Coding System Advantage plans are based on FFS spending, which (HCPCS). Payment rates are based on relative weights, reflects fee schedule payment rates. Moreover, many called relative value units (RVUs), which account for the commercial plans use RVUs from the fee schedule to amount of work required to provide a service, expenses determine their own payment rates for clinicians. related to maintaining a practice, and professional liability insurance costs. Collectively, these three Pricing distortions can influence the mix of services components make up the Resource-based Relative Value provided by clinicians by encouraging them to focus on Scale. Together with the fee schedule’s conversion factor services that are relatively more profitable than others, (or base payment amount), the RVUs produce a total leading to volume increases for the higher profit services. payment rate for each service. CMS, with input from Some of these additional services may represent low-value the American Medical Association/Specialty Society care, which refers to services that have little or no clinical Relative Value Scale Update Committee (RUC), revises benefit or care in which the risk of harm from the service the RVUs for some services each year based on changes outweighs the potential benefit (see Chapter 10 in this in clinical practice, new data, and other factors. In report on Medicare coverage policy and use of low-value addition, CMS annually sets RVUs for new and revised care). In addition to increasing health care spending, low- HCPCS codes. value care has the potential to harm patients by exposing them to the risk of injury from inappropriate tests or The Medicare Access and CHIP Reauthorization Act procedures. of 2015 (MACRA) established a new set of updates for clinicians billing under the fee schedule and repealed Ambulatory evaluation and management (E&M) the prior framework—the sustainable growth rate (SGR) services—which we define as office visits, hospital 1 formula—that set the conversion factor. The SGR outpatient department visits, visits to patients in certain was established to limit total fee schedule spending by other settings such as nursing facilities, and home restraining annual updates when spending exceeded visits—are essential for a high-quality, coordinated health certain parameters. MACRA provided a new framework care delivery system. These visits enable clinicians to for updating fee schedule payments. It established two diagnose and manage patients’ chronic conditions, treat payment paths: one path for clinicians who participate in acute illnesses, develop care plans, coordinate care across advanced alternative payment models (A–APMs), such as providers and settings, discuss patient preferences, and certain accountable care organization and episode of care engage in shared decision-making with patients. These models, and another path for other clinicians known as the services are critical for both primary care and specialty Merit-based Incentive Payment System (MIPS) (Medicare care. Therefore, to ensure that clinicians have an incentive Payment Advisory Commission 2018). The Commission to provide ambulatory E&M visits, these services should has recommended that the Congress eliminate MIPS not be priced too low relative to other services. and establish a new voluntary value program in fee-for- service (FFS) Medicare (Medicare Payment Advisory In this chapter, we first discuss why ambulatory E&M Commission 2018). MACRA established incentive services tend to be underpriced in the fee schedule and payments for clinicians who participate in A–APMs to evidence that the prices for other services are inflated. We encourage them to move toward these models. A–APMs then suggest an approach to rebalance the fee schedule generally require participating entities to assume financial toward ambulatory E&M services through a one-time risk for their patients, which encourages providers to price increase for these services that would be funded by reducing payment rates for other services. June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 69

92 As services experience efficiency gains, their Ambulatory E&M services are work RVUs should decline but often do not underpriced relative to other services Work RVUs for clinician services are based on an assessment of how much time and intensity services When CMS implemented the fee schedule in 1992, one of require relative to one another. Intensity refers to the the main goals was to reduce payment disparities between mental effort, technical skill, psychological stress, and primary care physicians and specialists (Ginsburg 2012, risk of performing a service. If estimates of time and Laugesen 2016). A large share of services provided by intensity are not kept up to date, especially for services primary care physicians are ambulatory E&M services. that experience efficiency improvements, the work RVUs From 1991 to 1996 (a period that includes the first five become inaccurate. years of the new fee schedule), payment rates for office and hospital outpatient visits grew by 4.3 percent per year and Procedures, imaging, and tests are more likely to rates for nursing facility/rest home visits increased by 9.4 experience efficiency gains than ambulatory E&M percent per year (Medicare Payment Advisory Commission services 1998). During this period, payment rates for most types of Due to advances in technology, technique, and clinical procedures and imaging declined (e.g., rates for cataract practice, efficiency gains are more likely to occur for lens replacement fell by 6.5 percent per year). However, procedures, imaging, and tests than for other services. For CMS’s review of certain fee schedule services in 1996 and example, when a new test or procedure is added to the fee 2001 led to substantially more services receiving higher schedule, it may be assigned a relatively high work RVU prices than lower prices (Medicare Payment Advisory because of the additional time, technical skill, mental Commission 2006b). The budget-neutral nature of the effort, and risk associated with performing the service. fee schedule means that raising prices for certain services Over time, however, as clinicians become more familiar leads to lower prices for others, such as ambulatory E&M with the service and more efficient at performing it, they services. These issues led the Commission to express can complete it faster and with less mental effort, skill, and concern in 2006 that ambulatory E&M services were risk (Medicare Payment Advisory Commission 2006b). underpriced relative to other types of services (Medicare Payment Advisory Commission 2006b). Ambulatory E&M services, by comparison, tend to be labor intensive and so do not lend themselves to efficiency Using recommendations from the RUC, CMS increased gains (Medicare Payment Advisory Commission 2008). work RVUs for several E&M services in 2007 and 2008, They are composed largely of activities that require the such as office and hospital outpatient visits. In addition, clinician’s time, such as taking the patient’s history, practice expense RVUs for E&M services increased examining the patient, and engaging in medical decision- between 2007 and 2013 because CMS adopted new making. methods and new data to calculate practice expense values (Medicare Payment Advisory Commission 2007, Because the time and effort needed to perform procedures, Medicare Payment Advisory Commission 2011b). Since imaging, and tests generally declines over time, clinicians 2013, however, payment rates for office and outpatient should be able to provide more of these services per day. visits have changed very little. For example, total RVUs However, because it is more difficult to achieve efficiency for a Level III office or outpatient visit for an established gains for ambulatory E&M services, we can expect patient (HCPCS 99213), the most frequently billed office lower volume growth for these services. As evidence, the or outpatient visit, declined slightly from 2.14 in 2013 cumulative growth in the volume of E&M services from 2 to 2.06 in 2018. Therefore, the Commission remains 2000 to 2016 was much less than the cumulative growth in concerned that ambulatory E&M services are underpriced the volume of tests, imaging, and other procedures (Figure relative to other services. 3-1, p. 72). The Commission has made prior recommendations to Ambulatory E&M services experience passive increase payment rates for ambulatory E&M services devaluation over time provided by certain clinicians (see text box on the Ideally, when efficiency gains reduce the amount of work Commission’s prior recommendations). One of these needed for a service, the work RVUs for the affected recommendations—a temporary bonus for certain E&M services should decline accordingly. Because the fee services provided by designated clinicians—was adopted schedule is budget neutral, a reduction in the RVUs of but expired in 2015. Rebalancing Medicare’s physician fee schedule toward ambulatory evaluation and management services 70

93 Prior Commission recommendations to improve payment for ambulatory E&M services & Medicaid Services 2010). E&M services he Commission has made prior in inpatient hospital settings and emergency recommendations to increase payment rates departments, annual wellness visits, chronic for ambulatory evaluation and management T care management services, and transitional care (E&M) services provided by certain clinicians management services were not considered primary relative to other services. In 2008, the Commission care visits under the PCIP. recommended that the Congress establish a bonus for designated ambulatory E&M services billed by included providers with a Primary care providers • eligible primary care practitioners (Medicare Payment primary Medicare specialty designation of family Advisory Commission 2008). The designated E&M practice, internal medicine, pediatrics, geriatrics, services included office visits, home visits, and visits nurse practitioner and clinical nurse specialist, and to patients in certain other settings (e.g., skilled nursing physician assistant and for whom primary care and intermediate care facilities). Eligible primary care visits accounted for at least 60 percent of allowed practitioners included clinicians whose designated charges under the fee schedule. specialty is primary care (e.g., family medicine) and who received at least 60 percent of their fee schedule– In 2011, the Commission recommended that the 3 allowed charges from ambulatory E&M services. Congress replace the sustainable growth rate (SGR) To help rebalance the fee schedule, the Commission system with payment updates that would have been recommended that spending for the bonus be budget higher for certain E&M services billed by eligible neutral. While the Commission did not recommend a primary care practitioners than for other services specific amount, we analyzed two levels for the bonus: (Medicare Payment Advisory Commission 2011a). 5 percent and 10 percent. Specifically, the Commission recommended that payment rates for certain E&M services be frozen at In response to this recommendation, the Patient their current levels for 10 years and rates for all other Protection and Affordable Care Act of 2010 created services be reduced in each of the first 3 years and then a temporary primary care bonus program called the frozen for the subsequent 7 years. Although the SGR Primary Care Incentive Payment (PCIP) program. was replaced, the Congress did not adopt differential However, the program was not budget neutral and updates for E&M services and other services. thus required additional funding. The PCIP, which existed from 2011 to 2015, provided a 10 percent In addition to recommendations specific to the fee bonus payment on fee schedule payments for certain schedule, the Commission recommended that the primary care visits provided by eligible primary care Congress establish a per beneficiary payment for practitioners. The PCIP’s definitions for these terms primary care providers to replace the PCIP after were as follows: it expired at the end of 2015 (Medicare Payment Advisory Commission 2015). The payment Primary care visits were ambulatory E&M • would provide funds to support the investment in services (e.g., office visits, home visits, and visits infrastructure and staff that facilitate care management in skilled nursing facilities) (Centers for Medicare and care coordination. ■ these services would raise the RVUs for all other services, to occur. Therefore, ambulatory E&M services become such as ambulatory E&M services. Because of problems passively devalued over time. In other words, their relative with the process of reviewing mispriced services and the prices are too low because the prices for other services data used to set prices, this two-step sequence tends not have become artificially high. June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 71

94 services. Services that have had their work RVUs reviewed FIGURE since 2009—whether new services, revised services, or Cumulative growth in the volume 3–1 FIGURE of E&M services per fee-for-service Volume growth has raised... services reviewed as potentially mispriced—accounted 4-X beneficiary was much less than for 35 percent of fee schedule spending in 2016 (Figure growth in volume of tests, imaging, 3-2). Services that have not yet been reviewed accounted and other procedures, 2000–2016 for an additional 35 percent of fee schedule spending. 100 If CMS were to review these services, the agency could Major procedures identify mispriced services and redistribute payments from overpriced to underpriced services. Ambulatory E&M 80 E&M services accounted for the remaining 30 percent of fee schedule spending, and CMS updated the payment rates Other procedures 60 for many of these services in 2007 and 2008. Tests 40 Imaging Tests Imaging Cumulative percent change 20 FIGURE Other procedures CMS has not yet reviewed the 3–2 E&M services FIGURE work RVUs of services that Medicare population Major procedures accounted for a significant share of X-X 0 fee schedule payments, 2016 2008 2010 2016 2012 2000 2002 2004 2014 2006 Services by category E&M (evaluation and management). Volume growth for E&M services from Note: (as percent of estimated fee schedule spending) 2009 to 2010 is not directly observable because of a change in payment policy for consultations. To compute cumulative volume growth for E&M services through 2016, we used a growth rate for 2009 to 2010 of 1.85 percent, which is the average of the 2008 to 2009 growth rate of 1.7 percent and the 2010 to 2011 growth rate of 2.0 percent. 30% Ambulatory E&M services* MedPAC analysis of claims data for 100 percent of Medicare Source: beneficiaries. Note and Source in InDesign. Note: CMS’s review of potentially mispriced 35% 35% Notes about this graph: services has not been sufficient Other services reviewed Services not reviewed CMS, with assistance from the RUC, has reviewed the • Data is in the datasheet. Make updates in the datasheet. work RVUs of many potentially mispriced services since • I had to force return the items on the x-axis. They will reflow if I update the data. 2009, but has not yet addressed services that account 4 • I had to manually draw tick marks and axis lines because they kept resetting when I changed any data. for a substantial share of fee schedule spending. After a service has been identified as potentially misvalued, • Use direct selection tool to select items for modification. Otherwise if you use the black selection tool, they will reset to graph it can often take several years for the RUC to develop default when you change the data. a recommendation for that service (Government • Use paragraph styles (and object styles) to format. Accountability Office 2015). CMS’s review is also Note: RVU (relative value unit), E&M (evaluation and management). Percentages are each category’s share of total fee schedule–allowed charges in hampered by the lack of current, accurate, and objective 2016. Services that had their work RVUs reviewed are those listed in data on clinician work time and practice expenses. Even fee schedule final rules for 2009 to 2016 as new, revised, or potentially misvalued or reviewed during the fourth five-year review. Ambulatory among the services for which CMS reduced the work E&M services include office visits, home visits, and visits to patients in RVUs, the RVUs did not decline as much as the estimated certain other settings (e.g., nursing facilities). *CMS increased payment rates for many of these services in 2007 and amount of time needed to provide the services. Note: Note and Source are in InDesign. 2008. Source: Although CMS’s review of potentially mispriced services Source: CMS fee schedule final rules and MedPAC analysis of claims data for 100 percent of Medicare beneficiaries. began in 2009, the agency has not yet reviewed many Rebalancing Medicare’s physician fee schedule toward ambulatory evaluation and management services 72

95 TABLE The RUC recommended a decrease in work RVUs for approximately half of the 3–1 potentially mispriced services for which it reviewed work RVUs, 2009–2017 Work RVUs Number of services Percent of services reviewed No change 647 39 % Increase 210 13 48 795 Decrease 1,652 100 Total RUC (Relative Value Scale Update Committee), RVU (relative value unit). The RUC examined a total of 2,220 services from 2009 to 2017. Work RVUs were Note: reviewed for 1,652 services, practice expense RVUs (but not work RVUs) were revised for 158 services, and billing codes were deleted for 410 services. Source: American Medical Association 2017. CMS and the RUC have identified potentially mispriced reductions in the estimated amount of time needed to services for review. From 2009 to 2017, the RUC provide the services. The statute defines the work of recommended lower work RVUs for only half of the clinicians as consisting of the time spent providing a potentially mispriced services for which it reviewed service and the intensity of work effort per unit of time work RVUs, a somewhat counterintuitive outcome given (e.g., mental effort and technical skill). For a number of that CMS and the RUC identified services for review that services, CMS (with input from the RUC) reduced the were likely to be overpriced (Table 3-1). According to an estimated amount of time that clinicians spend providing American Medical Association progress report, the RUC these services and the work RVUs for these services. reviewed work RVUs for 1,652 services as of October However, CMS did not reduce the work RVUs for these 5 2017 (American Medical Association 2017). services as much as the time estimates: The agency The RUC decreased the time estimates by an average of 18 percent recommended that CMS decrease the work RVUs for but decreased the work RVUs by an average of 9 percent 795 services (48 percent) but recommended no change (Table 3-2). A potential explanation for this disparity for 647 services (39 percent) and increases for 210 6 is that decreases in time were offset by increases in services (13 percent). The RUC used several screening intensity. In the absence of an increase in intensity, CMS criteria to identify potentially mispriced services for could have reduced work RVUs by the same percentage review, such as services with new technology, surgical as the time estimates, thereby making it possible to procedures that are performed less than half the time in redistribute more money to other services. inpatient settings but include inpatient E&M services in their payment rates, services with rapid volume growth, and services that are frequently performed together by the same physician on the same date. These types of services are more likely to be overpriced than underpriced, and thus the majority of services identified TABLE Time estimates have decreased with these criteria should have been candidates for RVU 3–2 more than work RVUs, 2008–2016 reductions. For example, the amount of time required for services that experience rapid volume growth Average percent change should decline over time as clinicians become more −18% Time estimates familiar with these services and can perform them faster. −9 Work RVUs Therefore, we would have expected the RUC to have recommended lower work RVUs for more than half of Note: RVU (relative value unit). Table reflects changes to RVUs adopted by CMS. the services they reviewed. The 607 services evaluated had work RVUs and work-time estimates in 2008 and 2016 and had a decrease in work RVUs, a decrease in the work-time estimate, or both. Even among the services for which CMS reduced the work RVUs, the decreases were not consistent with MedPAC analysis of physician time and RVU files from CMS. Source: June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 73

96 FIGURE XXXXXXXXX... FIGURE X-X Fee schedule’s time estimates are most important 3–3 factor in establishing work RVUs, 2017 100 90 21% 22% 23% 22% 21% 80 70 60 50 40 79% 79% 77% 78% 78% 30 20 Percent of total variation in work RVU Intensity 10 Time estimates 0 Tests Major procedures Imaging Other procedures E&M Note: RVU (relative value unit), E&M (evaluation and management). The percentages for time estimates are from five regression analyses: one for each service type. In these analyses, the log of estimated time was the explanatory variable, and the log of work RVU was the dependent variable. The percentages for intensity are the differences between the estimated time percentages and 100 percent. Source: MedPAC analysis of time data and work RVUs from CMS. Note: Note and Source are in InDesign. Source: surveyed by specialty societies for payment year 2015, the The data used to price services are median response rate to surveys was only 2.2 percent, the inadequate median number of responses to surveys was 52, and 23 of CMS’s lack of comprehensive, current, and objective Notes about this graph: 231 surveys had fewer than 30 respondents (Government data on clinician work time and practice expense is a key Accountability Office 2015). Third, the respondents • Data is in the datasheet. Make updates in the datasheet. reason the process for reviewing and revising mispriced are generally aware of the purpose of the survey (to set services has been inadequate. Clinician work time is a key • WATCH FOR GLITCHY RESETS WHEN YOU UPDATE DATA!!!! payment rates), and therefore their responses may be component of work RVUs. To estimate clinician work time • The column totals were added manually. biased in favor of higher time estimates. for specific services, CMS relies on data from surveys conducted by specialty societies that are reviewed by the • I had to manually draw tick marks and axis lines because they kept resetting when I changed any data. To address this problem, the Commission recommended RUC. We have three main concerns about the objectivity • I can’t delete the legend, so I’ll just have to crop it out in InDesign. in 2011 that CMS use a streamlined method to regularly and quality of these data. First, the specialty societies collect data from a cohort of efficient practices—including • Use direct selection tool to select items for modification. Otherwise if you use the black selection tool, they will reset to graph that conduct the surveys have a financial stake in the service volume and work time—to establish more accurate default when you change the data. process of setting payment rates. Second, the survey data work and practice expense RVUs (Medicare Payment have weaknesses that include low response rates and low • Use paragraph styles (and object styles) to format. Advisory Commission 2011a). CMS’s response has been total number of responses, which raises questions about to contract with researchers to develop models to validate • Data was from: R:\Groups\MGA\data book 2007\data book 2007 chp1 the representativeness of the results. For example, the the RVUs. These models attempt to validate the time Government Accountability Office found that, for services estimates for services one by one (e.g., through time-and- Rebalancing Medicare’s physician fee schedule toward ambulatory evaluation and management services 74

97 motion studies). The Commission’s concern has been that type of service, time explains over 75 percent of the this approach is time consuming, costly, and likely to be variance in work RVUs (Figure 3-3). burdensome for providers and CMS. However, it may be The contractors collected data from diverse sources: useful for identifying specific services that are potentially administrative data on service volume and physician hours misvalued. worked, physician surveys, time-and-motion studies, and The Commission has recommended a different, “top- electronic health records. They analyzed the data using down” approach to validate the RVUs (Medicare Payment either a top-down approach or a bottom-up approach, Advisory Commission 2011b). This method looks at which examines each service separately. Although a the amount of time that a physician worked over the bottom-up approach is costly, the findings from this course of a week or month and compares it with the method illustrate significant distortions in the time time estimates in the fee schedule for all of the services estimates for common services. that the physician billed over the same period. If the fee Specialties other than primary care had the largest schedule’s time estimates exceed the actual time worked, differences between time assumed in the fee this finding could indicate that the time estimates are too schedule and actual time worked high. In 2014, a contractor for the Commission explored The intent of the project for ASPE was to better the feasibility of the top-down approach by collecting data understand whether there are systematic differences from a small set of physician practices on the services or errors in the fee schedule’s time assumptions across billed by their clinicians and the clinicians’ actual hours specialties or groups of services. The contractor acquired worked (Medicare Payment Advisory Commission 2015). data from three integrated delivery systems (IDSs): one If CMS used a top-down approach to validate RVUs, it located in the West, one in the Midwest, and one in the could identify groups of services that are likely overpriced, eastern United States (Merrell et al. 2014). To assess carefully review those services, and price them more the accuracy of the time assumptions from a top-down accurately. perspective, the contractor collected administrative data Evidence that estimates of clinician work on service volume by physician and billing code. These time are inflated service volumes were multiplied by the code-specific time assumed in the fee schedule and summed for each Contractors working for CMS and the Assistant Secretary physician to calculate “fee schedule time.” Data were also for Planning and Evaluation (ASPE) in the Department collected on “actual time worked,” calculated based on of Health and Human Services have gathered evidence clinical practice days per year, clinical hours per year, or that the fee schedule’s time estimates for clinician work a full-time equivalent measure, depending on the IDS. are inflated (Merrell et al. 2014, Zuckerman et al. 2016). The accuracy of the fee schedule’s time assumptions was While there was heterogeneity in the data and methods analyzed as the ratio of fee schedule time to actual time used by the contractors, the findings were consistent: 7 worked. The time assumptions for a broad range of services in the fee schedule—particularly imaging, procedures, and Analyzing the differences between fee schedule time and tests—are inflated when compared with ambulatory E&M actual time worked, the contractor concluded that the fee services. Indeed, errors in some of the fee schedule’s time schedule’s time assumptions may be distorted for some assumptions were very large—multiples of the actual time specialties. Specifically, their findings are consistent spent by physicians. The Commission’s position is that with the conclusion that primary care is disadvantaged the time assumptions—and, therefore, the fee schedule’s by the current time assumptions (Table 3-3, p. 76). The work RVUs—should be validated and corrected. In median ratio of fee schedule time to actual time worked, the meantime, a budget-neutral payment adjustment when evaluated across all specialties, was 1.35. However, would appropriately rebalance the fee schedule toward the ratio for radiology was higher, at 2.00; the ratio for ambulatory E&M services. cardiology was highest, at 2.08. By contrast, the ratios were lowest for pathology, general surgery, and primary The contractors focused on estimates of the time that it care at 1.14, 1.16, and 1.25, respectively. Primary care takes clinicians to furnish services to a typical patient. specialties tend to concentrate on ambulatory E&M These time assumptions are important because they are services. highly predictive of the work RVUs. Depending on the June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 75

98 TABLE Primary care physicians’ ratio of fee schedule time 3–3 to actual time is below the median for all physicians Median ratio of fee schedule Number of time to actual time Specialty physicians Pathology 31 1.14 1.16 General surgery 53 1.25 Primary care 231 Orthopedic surgery 45 1.35 All other specialties 1.36 345 2.00 57 Radiology 2.08 Cardiology 44 All 806 1.35 “Primary care” includes family medicine and internal medicine. “Fee schedule time” refers to the work time assumed in the fee schedule for the services provided by Note: each physician. “Actual time” refers to the actual time worked by each physician, based on their clinical practice days per year, clinical hours per year, or a full- time equivalent measure. Merrell et al. 2014. Source: data source was direct observation—project or physician Findings suggest that the fee schedule’s time estimates for services other than ambulatory E&M practice staff observing and documenting the time needed services are inflated to provide services to individual patients. The direct observation data were collected at three sites in different Two projects that examined each service separately regions of the U.S.: Mid-Atlantic, New England, and suggest that the fee schedule’s time assumptions for Pacific. EHR data were available from two of the sites. services other than ambulatory E&M services are likely too high. One study was a pilot project for CMS on In selecting services for the project, researchers considered validating the time assumptions for 60 services with data a service’s risk of being misvalued and its importance to gathered from both electronic health records and direct Medicare because of total spending on the service or other observation of the care received by individual patients policy reasons. Researchers also selected a mix of services (Zuckerman et al. 2016). The other project, for ASPE, that would allow them to test methods in a variety of assessed the feasibility of validating the time assumptions 9 clinical settings. The services were of four types: office- for 26 services with data from a survey in which based procedures, outpatient department or ambulatory physicians were asked how many minutes they typically surgical center procedures, inpatient procedures with spend when furnishing each of the services (Merrell et al. global periods, and imaging and other test interpretations. 2014). E&M services were not included. The pilot project for CMS included Pilot project for CMS In interpreting the results, the contractor concluded that the developing empirical measures of physician service fee schedule’s time assumptions were often high relative time for specific services (Zuckerman et al. 2016). to the empirical time captured in their study. For 42 of The contractor measured time in one of two ways, the 60 services studied, the ratios of fee schedule time to 8 depending on the service and data collection site. First, empirical time were over 1.1, based on the data collected administrative data were extracted from electronic health (Table 3-4). The largest discrepancies were in imaging and records (EHRs) for some services. EHRs include time other test interpretations. Electrocardiogram report—the stamps for each recorded event (e.g., start of a procedure). extreme case—had a fee schedule time of 5 minutes but The contractor calculated the service time in minutes with a median study time of only 6 seconds, making the fee start and end time stamps, excluding minutes associated schedule time 50 times the actual time observed. Other with any documented interruptions or pauses. The second Rebalancing Medicare’s physician fee schedule toward ambulatory evaluation and management services 76

99 imaging and test interpretations had smaller discrepancies, TABLE Ratios of fee schedule time to but fee schedule times were still multiples of empirically 3–4 empirical measures of physician based medians. Clinical expert reviewers consulted by the time were over 1.1 for contractor attributed the discrepancies to automation and most services reviewed personnel substitution that has become prevalent since CMS and the RUC defined the content of the services and Ratio of fee schedule time Number of to empirical physician time services valued them. Under 0.9 8 The contractor for ASPE Feasibility study for ASPE 10 0.9 to 1.1 surveyed physicians in five specialties: cardiology, family Over 1.1 42 medicine, radiology, ophthalmology, and orthopedic surgery (Merrell et al. 2014). Each physician was asked 60 Total about the time spent providing selected services relevant to their specialty. The 26 services selected—an average “Fee schedule time” refers to the work time assumed in the fee schedule. Note: “Empirical physician time” is based on data from electronic health records of 5 per specialty—were frequently provided, such as or direct observation of the time needed to provide services to individual echocardiogram, office visits, computed tomography of patients. the abdomen, cataract removal with lens insertion, and Source: Zuckerman et al. 2016. 10 knee arthroplasty. A total of 625 physicians participated in the survey. Questionnaires were administered through mixed modes: mail and internet, with telephone prompts performed by interviewers trained to solicit participation. Some of the physicians were from random samples drawn from the To summarize the results by specialty, each physician’s 11 American Medical Association Physician Masterfile. response for a service was categorized as implying Others were from multispecialty group practices that that the fee schedule’s time assumption for that service agreed to participate. Two of these practices were in was too high, too low, or about right (Table 3-5, p. 78). the South, three in the West, one in the Midwest, and Overall, the rate at which physicians said that fee schedule one in the Mid-Atlantic. The survey was administered times were too high was almost 58 percent. However, by from November 2013 through July 2014. Participants specialty, the rate ranged from a high of about 72 percent were offered a financial incentive to encourage adequate for radiologists to a low of almost 44 percent for family response to the survey. The response rate was 54 percent. medicine. The rate for family medicine means that the survey participants in this specialty were more likely to The contractor summarized the survey results as say that fee schedule times were too low or about right suggesting that, for the majority of the 26 services, the than they were to say that those times were too high. fee schedule’s time assumptions are high. At the time of the study, for example, photocoagulation of the retina Evidence that RVUs for global surgical had the highest ratio of fee schedule time assumption to services are inflated median survey time estimate: 3.78. In other words, the Currently, the payment rate for many surgical services is fee schedule time assumption was almost four times the a bundled payment that includes the procedure itself and survey estimate. The service’s time assumption was 208 certain services that are provided immediately before and minutes, but its median time estimate from the survey was 12 after the procedure; CMS calls this group of services the 55 minutes. Another example is MRI of the brain, for global package. There are three categories of global billing which the fee schedule time assumption was more than codes based on the number of postoperative days included twice the survey time estimate. By contrast, the ratios for in the global package: the three ambulatory E&M services in the survey—Level III and Level IV office visits for established patients and 0-day global codes, which include the procedure and • Level IV office visit for new patients—were 1.05, 1.00, preoperative and postoperative physician services on and 1.00, respectively. Overall, most services (20 of 26) the day of the procedure; had fee schedule time assumptions that were higher than their median survey time estimates. June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 77

100 TABLE Specialties other than family medicine physicians were 3–5 more likely to report fee schedule times that were too high Percent of survey responses that suggest fee schedule time is: Total number of service-level responses Too low Too high About right Specialty Family medicine 582 43.8 45.5 % 10.7 % % 50.5 14.9 34.5 469 Cardiology 59.1 530 8.1 Orthopedics 32.8 Ophthalmology 443 66.1 2.7 31.2 3.4 72.4 Radiology 496 24.2 34.1 8.1 57.8 2,520 All Note: Fee schedule time was defined as “too high” if the fee schedule time exceeded the survey time estimate by more than 5 percent. “About right” means that the fee schedule time was within 5 percent of the survey time. “Too low” indicates that the fee schedule time was lower than the survey time by more than 5 percent. Each specialty’s shares may not sum to 100 due to rounding. Source: Merrell et al. 2014. • 10-day global codes, which include the same of the global package or just a single value for all services as the 0-day global codes plus physician components of the package). visits related to the procedure during the 10 days The global codes contribute to payment disparities • after the procedure; and among specialties. • 90-day global codes, which include the same • The global packages are inconsistent with current services as the 0-day global codes plus preoperative medical practice (e.g., care has been shifting from services furnished one day before the procedure and individual practitioners to larger practices and postoperative services during the 90 days after the teams). procedure. CMS also cited evidence from the Office of Inspector In general, the Commission supports moving Medicare General (OIG) that the RVUs for global codes may not in the direction of bundled payments to counter the reflect the typical number and level of postoperative volume incentives intrinsic to FFS Medicare. However, visits (Office of Inspector General 2012a, Office of it is essential that the individual services that make Inspector General 2012b). OIG reviewed a sample of up a bundle have accurate values and that there is a medical records for several types of global surgical codes mechanism to ensure that the services that are part of the and counted the number of postoperative E&M visits bundle are not paid separately (unbundling). Otherwise, provided by the physicians. In many cases, OIG found the payment rate for the entire bundle will be inaccurate. that physicians provided fewer E&M visits during the CMS has raised several concerns with the 10-day and postoperative period than were included in the payment 90-day global packages for surgical services (Centers for for the global package. OIG recommended that CMS Medicare & Medicaid Services 2014): adjust the number of E&M visits in the global package to reflect the number that are actually provided. • The number and type of visits needed in the package for a given service are likely to change over time as The global payment assumes that the same physician medical practice and the patient population changes. who performs the procedure also provides all the postoperative care. However, a study by the RAND There is a lack of consistency in how the work RVUs • Corporation for CMS observed that postoperative for global codes are constructed (e.g., services may care is shifting from the physician who performed the have work RVUs that are the sum of each component Rebalancing Medicare’s physician fee schedule toward ambulatory evaluation and management services 78

101 procedure to other clinicians, such as hospitalists and Commission recommended in 2011 (Medicare Payment nonphysician practitioners, who bill separately for each Advisory Commission 2011a). postoperative visit (Mehrotra et al. 2016). This change suggests that physicians who bill for the global payment may be receiving payments for postoperative care that is provided by other clinicians. An approach to rebalance the fee schedule toward ambulatory E&M CMS proposed to convert all 10-day global codes to services 0-day codes in 2017 and convert all 90-day codes to 0-day codes in 2018. With these changes, providers Despite efforts made by CMS and the RUC over the last would bill separately for all preoperative visits and several years to review potentially mispriced services postoperative visits that occur after the day of the and adjust their payment rates, there is evidence that procedure. However, the Medicare Access and CHIP certain types of services—such as procedures—are Reauthorization Act of 2015 (MACRA) directed CMS still overpriced. Because the fee schedule is budget to not transition all 10-day and 90-day global codes to neutral, ambulatory E&M services become underpriced 0-day codes. Instead, MACRA mandated that CMS through a process of passive devaluation. One approach develop and implement a process to gather the necessary to rebalance the fee schedule toward ambulatory E&M data to appropriately value postoperative care. CMS is services is to increase payment rates for these services currently collecting the data. and to maintain budget neutrality by reducing payment The available evidence from CMS, OIG, and RAND rates for other services (e.g., procedures, imaging, and suggests that 10-day and 90-day global surgical services tests). Because these services are essential for both are overvalued. It may take CMS several years to collect primary care and specialty care, the higher payment data and revalue these services. In the meantime, a rates should apply to all clinicians who bill for an budget-neutral payment adjustment for ambulatory E&M ambulatory E&M visit, regardless of specialty. This services—excluding the ambulatory E&M services change would be a one-time price adjustment to the fee currently considered when valuing global packages— schedule to address several years of passive devaluation would rebalance the fee schedule toward ambulatory of ambulatory E&M services. This adjustment could be E&M services. phased in over multiple years to reduce the impact on other services. To reduce the need for future significant Problems with the accuracy of practice price changes and to address the mispricing of individual expense RVUs services, CMS should accelerate its efforts to identify overpriced services and adjust their payment rates. To In addition to the shortcomings with the data used do so, CMS should regularly collect data from a cohort to estimate clinician work time, there are also major of efficient practices and use this information to validate problems with the accuracy of the data used to set payment rates and establish accurate RVUs. practice expense RVUs. Practice expense includes the cost of nonphysician clinical and administrative staff, Design issues medical equipment and supplies, office rent, and other expenses. First, CMS does not have a comprehensive A key design issue is which ambulatory services should data source with current information on the prices of be included in the payment increase. For the purpose of medical equipment and supplies; consequently, the price this approach, we included E&M billing codes for office estimates for these items are often outdated (Medicare visits, home visits, and visits to patients in certain non- Payment Advisory Commission 2006a). Second, practice inpatient hospital settings (nursing facility, domiciliary, expense RVUs are also based on data from a survey rest home, and custodial care). We excluded newer E&M of total practice costs incurred by nearly all specialty services that were added to the fee schedule in recent groups. Because this survey was conducted in 2007 and years because they have not been subject to several 2008, the data are unlikely to reflect current practice years of passive devaluation. For example, we excluded costs. CMS has not developed a strategy for updating annual wellness visits (added to the fee schedule in this information. However, CMS could collect data on 2011), transitional care management services (added in total practice costs along with data on service volume 2013), and chronic care management services (added in and work time from a cohort of efficient practices, as the 2015 and 2017). We also considered whether to include June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 79

102 ambulatory psychiatric visits in the payment increase. services because there is a single base payment amount. Like ambulatory E&M services, many ambulatory Moreover, it would increase the reach of the policy psychiatric services are time-based services that do not beyond Medicare because many commercial plans use lend themselves to efficiency gains (e.g., HCPCS code the fee schedule’s RVUs to determine their payments to 90834 is for a 45-minute psychotherapy visit). However, clinicians. The second approach would make it easier we excluded them from the payment increase that we for policymakers to establish payment updates for model below because the payment rates for most of these ambulatory E&M services that are different from updates services were updated in recent years. Nevertheless, for other services in the future. Under either approach, policymakers could consider applying a payment the end result is the same: Clinicians would receive increase to the newer E&M and ambulatory psychiatric a higher payment rate for ambulatory E&M services services in addition to ambulatory E&M services. and a lower rate for other services. The results of our illustrative model, described below, would be the same Another important design issue is the size of the payment under either approach. increase for ambulatory E&M services. Although these services have become passively devalued, we were not Another critical design question is how to offset the able to precisely quantify how much these services are increase in fee schedule payments for ambulatory E&M underpriced. We considered an increase in the range of 5 services in a budget-neutral manner. We describe three percent to 30 percent. If policymakers decided to make a options: one-time price adjustment to ambulatory E&M services, an automatic reduction to the prices of new services • they would need to make a policy judgment about (after a certain amount of time) and services with the appropriate increase. One precedent to consider is high growth rates, the Primary Care Incentive Payment program, which, from 2011 through 2015, provided a 10 percent bonus • an extension of the annual numeric target for CMS for certain E&M visits provided by eligible primary to reduce the prices of overpriced services, and care practitioners (see text box, p. 71). To illustrate the impact of a budget-neutral payment increase for • an across-the-board reduction to all fee schedule ambulatory E&M services, we modeled a 10 percent services other than ambulatory E&M services. increase. However, a smaller or larger adjustment could Under the first option for budget neutrality, there would also be considered. Our model assumes that the increase be an automatic adjustment to the prices of new services would apply to both Medicare program payments and to ensure that prices declined over time, consistent beneficiary cost sharing so that cost sharing would with the expectation that the amount of time and effort continue to equal 20 percent of the total payment amount required for new services should decline over time for a fee schedule service, which is the current policy. because of advances in technology, technique, and As a result, beneficiary cost sharing would increase other factors (Medicare Payment Advisory Commission for ambulatory E&M services but decline for all other 2006b). Because the payment rates for new services services. Total cost sharing across all services would are not updated frequently enough to reflect reductions remain about the same. in time and effort, these services tend to become CMS could increase payment rates for ambulatory E&M overpriced. An automatic reduction triggered after a services in a budget-neutral manner by raising total certain number of years would ensure that payment rates RVUs for these services while reducing RVUs for all did not remain too high. Services that were recently 13 other services. reviewed by the RUC and CMS and had their RVUs Alternatively, CMS could create two reduced could be exempt from an automatic reduction. different conversion factors: a higher one for ambulatory 14 An automatic reduction could also apply to services that E&M services and a lower one for all other services. experience high volume growth because the dynamic of Currently, CMS uses a single conversion factor to learning by doing that applies to new technology should calculate payment rates for all fee schedule services. also apply to services that are being provided more The first approach is consistent with CMS’s current frequently. Savings from this automatic reduction could method for adjusting practice expense RVUs; if RVUs be used to offset increased payments for ambulatory for some services go up, RVUs for other services decline E&M services. by a corresponding amount. This approach also makes it simpler to apply the same update to all fee schedule Rebalancing Medicare’s physician fee schedule toward ambulatory evaluation and management services 80

103 Under the second option for budget neutrality, there schedule services (the third budget-neutrality option would be an extension of the annual numeric target described above). Other alternatives could be considered set by the Congress for CMS to reduce the prices of for the size of the payment increase and how to offset overvalued services. The Congress set this target for the increase. Our model assumes that the payment a three-year period (2016 through 2018). The target changes would occur in a single year, but the changes was set at 1 percent of fee schedule spending for 2016 could instead be phased in over multiple years. The net and 0.5 percent for 2017 and 2018. (CMS did not effect of these changes on specialties would vary based meet the target in any of the three years, which meant on each specialty’s mix of ambulatory E&M and other that payment rates for all fee schedule services were services. Specialties that focus on ambulatory E&M reduced by the difference between the target and the services would receive a net increase in payments, while actual aggregate reduction to the RVUs of overpriced specialties that mainly provide other services would services.) The annual numeric target was based on a receive a net decrease, assuming there is no change in Commission recommendation from 2011 (Medicare volume due to changes in payment rates. Payment Advisory Commission 2011a). Under this The increased payments for ambulatory E&M services option, the target would be extended beyond 2018 and would total $2.4 billion (based on 2016 data). To the cumulative target amount would be based on the determine the total amount of the additional payments total amount of money to be redistributed to ambulatory for ambulatory E&M services by specialty, we summed E&M services. For example, a cumulative target amount the fee schedule payments for ambulatory E&M services of 4 percent of fee schedule spending could be phased in 2016 for each specialty and multiplied this amount in through an annual 1 percent target over four years. by 10 percent. Table 3-6 (p. 82) shows the increase in Savings achieved by reducing the prices of overpriced payments for ambulatory E&M services and the net services would be redistributed to ambulatory E&M effect of the 10 percent payment increase for these 15 services. If CMS did not meet the target, payment rates services and the 3.8 percent reduction to other services, for all fee schedule services other than ambulatory E&M by specialty, for the 20 specialties with the highest share services would be reduced by the difference between the of total fee schedule payments in 2016. Online Appendix target and the actual reduction to the prices of overpriced 3-A, available at http://www.medpac.gov, displays these services. These savings would be redistributed to impacts for all specialties. ambulatory E&M services. Internal medicine and family practice would receive the Under the third option, the payment increase for largest amount of additional payments for ambulatory ambulatory E&M services would be offset by an across- E&M services ($435 million and $378 million, the-board payment reduction to all other fee schedule respectively) (Table 3-6, p. 82). The three specialties services (procedures, imaging, tests, and other E&M that would receive the highest percent increase in their services such as those provided in emergency department total fee schedule payments (on net) are endocrinology and inpatient hospital settings). To fully offset a 10 (6.6 percent net increase in fee schedule payments), percent payment increase for ambulatory E&M services, rheumatology (5.5 percent increase), and family practice for example, there would need to be a payment decrease (4.9 percent increase) (see online Appendix 3-A, of 3.8 percent for all other fee schedule services. These available at http://www.medpac.gov). These specialties payment changes could be implemented in one year or concentrate on ambulatory E&M services. Specialties phased in gradually over multiple years. This estimate that perform procedures but also provide a significant assumes that there would be no changes in service number of ambulatory E&M services—such as urology, volume as a result of the changes in payment rates. obstetrics/gynecology, and otolaryngology—would also experience a net increase in fee schedule payments. Modeling the net effect of a payment increase for ambulatory E&M services Several specialties would experience reductions in their To illustrate the impact of a budget-neutral payment fee schedule payments of 3.8 percent because they increase for ambulatory E&M services, we modeled a provide very few ambulatory E&M services. These 10 percent increase that would be offset by a 3.8 percent specialties include diagnostic radiology, pathology, across-the-board payment reduction to all other fee physical therapy, and occupational therapy (see online Appendix 3-A, available at http://www.medpac.gov). June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 81

104 TABLE Impact of a 10 percent payment rate increase for ambulatory E&M services 3–6 offset by a 3.8 percent reduction for all other services, for specialties with highest share of total fee schedule payments, 2016 Net change in Share of fee schedule payments Amount of total payment as a result of payment Current payments payment increase increase for increase for ambulatory for ambulatory for ambulatory ambulatory E&M E&M services and E&M services E&M services services (across payment reduction for Specialty (in millions) (in millions) all specialties) all other services % % 4.9 15.7 $378 $3,782 Family practice 6.8 165 4.1 1,650 Nurse practitioner 689 69 2.9 2.8 Hematology/oncology 82 824 Physician assistant 3.4 2.5 Neurology 658 66 2.7 2.0 3.1 74 Urology 745 1.9 18.0 435 4,349 Internal medicine 1.7 1.4 3.1 Podiatry 744 74 0.3 Cardiology 1,681 168 7.0 Pulmonary disease 51 0.2 2.1 507 2.1 49 495 Gastroenterology 0.1 3.9 93 933 Orthopedic surgery –0.4 –0.5 84 Dermatology 841 3.5 341 1.4 –1.5 General surgery 34 –1.6 1.5 36 356 Nephrology 50 505 Ophthalmology –2.6 2.1 Emergency medicine –3.1 0.7 18 177 8 83 0.3 –3.2 Radiation oncology 0.1 Diagnostic radiology –3.8 14 1 <1 Physical therapy –3.8 0.0 <1 Note: E&M (evaluation and management). Table includes the 20 specialties with the highest share of total fee schedule payments. “Ambulatory E&M services” includes office visits, home visits, and visits to patients in certain non-inpatient hospital settings (nursing facility, domiciliary, rest home, and custodial care). The payment increase is applied to allowed charges for ambulatory E&M services. Estimates assume there would be no changes in service volume as a result of changes in payment rates. Analysis includes services billable under the fee schedule for physician and other health professional services. Source: MedPAC analysis of claims data for 100 percent of Medicare beneficiaries, 2016. the process for reviewing mispriced services. To support Conclusion these efforts, CMS should regularly collect data from a cohort of efficient practices and use this information to We describe an approach to address the problem of passive validate the payment rates. Improving the accuracy of devaluation of ambulatory E&M services that would prices for ambulatory E&M and other services going rebalance fee schedule payment rates in a budget-neutral forward could reduce the need for future significant manner. It would also help reduce the risk of beneficiaries adjustments to rebalance the fee schedule. ■ experiencing problems accessing these services and send a more favorable signal to medical students and residents contemplating careers in specialties that provide a large share of E&M services. Even if this approach is adopted, we urge CMS to accelerate its efforts to improve the accuracy of the data used to calculate payment rates and Rebalancing Medicare’s physician fee schedule toward ambulatory evaluation and management services 82

105 Endnotes 9 Services at risk of being misvalued included those with the 1 clinicians In this chapter, the term is synonymous with fastest growth, substantial changes in practice expenses, and . physicians and other health professionals new technologies. 2 These RVUs are the national average nonfacility RVUs. Physicians in three specialties were asked about time spent 10 3 The full list of practitioners eligible for the bonus, as providing a Level IV office visit for an established patient recommended by the Commission, was family medicine, (HCPCS 99214): cardiology, family medicine, and orthopedic internal medicine, geriatric medicine, pediatric medicine, surgery. Family medicine physicians were also asked about nurse practitioner, and physician assistant. a Level III office visit for an established patient (HCPCS 99213) and a Level IV office visit for a new patient (HCPCS 4 When CMS reviews the work RVUs for a code, it also reviews 99204). the practice expense RVUs. 11 The American Medical Association Physician Masterfile The RUC examined a total of 2,220 services from 2009 to 5 includes current and historical data for more than 1.4 million 2017. Work RVUs were reviewed for 1,652 services, practice physicians, residents, and medical students in the United expense RVUs (but not work RVUs) were revised for 158 States. services, and billing codes were deleted for 410 services. 12 CMS has since reduced the service’s time assumption from Although CMS has accepted most of the RUC’s prior 6 208 minutes to 81 minutes. recommendations, we do not have information on whether CMS accepted the recommendations for these specific 13 Total RVUs include work, practice expense, and professional services. liability insurance RVUs. The analysis used the fee schedule’s payment modifiers (e.g., 7 14 There is precedent for a fee schedule with more than one assistance at surgery) to adjust service volumes. conversion factor. Under the volume performance standard policy that was replaced by the sustainable growth rate 8 Both types of data were based on intraservice time, the largest formula in 1997, the fee schedule had separate conversion component of the RUC’s time assumptions. Intraservice time factors for surgical services, primary care services, and other includes the time the clinician spends on treatment/therapy nonsurgical services. and documentation of services. The other two components of the time assumptions are preservice time—preparing to see In extending the target, the Congress would need to specify 15 the patient, reviewing records, and communicating with other that the savings would be redistributed only to ambulatory professionals—and postservice time—arranging for further E&M services. Under the target that expires at the end of services and communicating (written or verbal) with the 2018, savings are redistributed to all fee schedule services. patient, family, and other professionals. Report to the Congress: Medicare and the Health Care Delivery System June 2018 | 83

106 References Medicare Payment Advisory Commission. 2008. Report to the RUC Relativity Assessment American Medical Association. 2017. Congress: Reforming the delivery system. Washington, DC: Workgroup progress report. Updated October 2017. Chicago, IL: MedPAC. AMA. Medicare Payment Advisory Commission. 2007. Report to the Centers for Medicare & Medicaid Services, Department of Congress: Promoting greater efficiency in Medicare. Washington, Health and Human Services. 2014. Medicare program; revisions DC: MedPAC. to payment policies under the physician fee schedule, clinical laboratory fee schedule, access to identifiable data for the Center Medicare Payment Advisory Commission. 2006a. Report to the for Medicare and Medicaid Innovation Models & other revisions Congress: Increasing the value of Medicare. Washington, DC: 79, no. Federal Register to Part B for CY 2015. Proposed rule. MedPAC. 133 (July 11): 40318–40540. Medicare Payment Advisory Commission. 2006b. Report to the Centers for Medicare & Medicaid Services, Department of Health Congress: Medicare payment policy. Washington, DC: MedPAC. and Human Services. 2010. Medicare program; payment policies under the physician fee schedule and other revisions to Part B for Medicare Payment Advisory Commission. 1998. Healthcare CY 2011. Final rule. Federal Register 75, no. 228 (November 29): spending and the Medicare program: A data book. Washington, 73169–73860. DC: MedPAC. Ginsburg, P. B. 2012. Fee-for-service will remain a feature of Mehrotra, A., C. Gidengil, L. Hilborne, et al. 2016. Developing major payment reforms, requiring more changes in Medicare codes to capture post-operative care. Santa Monica, CA: RAND. 31, no. 9 (September): 1977– physician payment. Health Affairs 1983. Merrell, K., C. Schur, T. Oberlander, et al. 2014. Analysis of physician time use patterns under the Medicare fee schedule. Medicare physician Government Accountability Office. 2015. Report prepared for the Assistant Secretary for Planning and payment rates: Better data and greater transparency could Evaluation. Washington, DC: Social & Scientific Systems and the improve accuracy. GAO-15-434. Washington, DC: GAO. Urban Institute. Fixing medical prices: How physicians are Laugesen, M. J. 2016. Office of Inspector General, Department of Health and Human paid. Cambridge, MA: Harvard University Press. Cardiovascular global surgery fees often did Services. 2012a. not reflect the number of evaluation and management services Medicare Payment Advisory Commission. 2018. Report to the No. A-05-09-00054. Washington, DC: OIG. provided. Washington, DC: MedPAC. Congress: Medicare payment policy. Office of Inspector General, Department of Health and Human Report to the Medicare Payment Advisory Commission. 2015. Services. 2012b. Musculoskeletal global surgery fees often did Congress: Medicare payment policy. Washington, DC: MedPAC. not reflect the number of evaluation and management services provided. No. A–05–09–00053. Washington, DC: OIG. Medicare Payment Advisory Commission. 2011a. Moving forward from the sustainable growth rate (SGR) system. Letter to Zuckerman, S., K. Merrell, R. Berenson, et al. 2016. Collecting the Congress. October 14. empirical physician time data: Piloting an approach for validating work relative value units. Report prepared for the Report to Medicare Payment Advisory Commission. 2011b. Centers for Medicare & Medicaid Services. Washington, DC: The the Congress: Medicare and the health care delivery system. Urban Institute. Washington, DC: MedPAC. Rebalancing Medicare’s physician fee schedule toward ambulatory evaluation and management services 84

107 CHAPTER 4 Paying for sequential stays in a unified prospective payment system for post-acute care

108

109 CHAPTER 4 Paying for sequential stays in a unified prospective payment system for post-acute care In this chapter Chapter summary In 2016, Medicare fee-for-service (FFS) spending on post-acute care (PAC) • Background services—skilled nursing facilities (SNFs), home health agencies (HHAs), • Challenges with paying for inpatient rehabilitation facilities (IRFs), and long-term care hospitals sequential post-acute care (LTCHs)—totaled $60 billion. For any condition, Medicare’s FFS payments stays can differ substantially because Medicare uses separate prospective payment systems (PPSs) to pay for stays in each setting. As mandated by the Congress, • Summary of the proposed in June 2016, the Commission evaluated a prototype design and concluded PAC PPS design that it was feasible to design a unified PAC PPS that would establish accurate Definition of sequential PAC • payments using readily available data. The Commission recommended stays the necessary features of a PAC PPS that spans the four settings and bases payments on patient characteristics. Our initial work concluded that the design Characteristics of sequential • would establish accurate payments for most of the more than 40 patient groups PAC stays we examined and would increase the equity of payments across conditions. • PAC PPS payments need to In turn, providers would have less incentive to selectively admit certain types align with the cost of stays of patients over others. In June 2017, the Commission recommended that a throughout a sequence of PAC PPS be implemented beginning in 2021 with a three-year transition and a post-acute care corresponding alignment of setting-specific regulatory requirements. • Defining the beginning and The Commission continues to work on a unified PAC PPS, considering end of stays when treating in refinements that would improve the design. These refinements should not place delay implementing a PAC PPS or the Commission’s recommendation to • Conclusion improve the equity of PAC payments before the PAC PPS is implemented. June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 87

110 Refinements focus on increasing the accuracy of payment for cases that involve a course of PAC care—that is, sequential stays—which we define as PAC stays within seven days of each other. In this chapter, refinements focus on two payment issues related to sequential stays. The first has to do with the way the cost of a stay can vary depending on where it falls in a sequence of PAC stays. The reason is that, throughout a course of care, a beneficiary’s clinical condition is likely to change, so later PAC stays could have different average costs—often lower but sometimes higher—compared with initial PAC stays. As with other FFS payment systems, it will be important under the unified PAC PPS to align payments with the cost of each stay throughout a sequence of stays. If payments and costs are not aligned, providers could have a financial incentive to refer beneficiaries for unnecessary subsequent care or could have difficulty placing beneficiaries who require continued care. A second issue involves how to identify, for payment purposes, distinct levels of care for a PAC provider that treats a patient with evolving care needs “in place” rather than referring the patient to another PAC provider. Under the unified PAC PPS, such providers would be financially disadvantaged unless the payment system included a way to trigger payments for different phases of care. Of 8.9 million PAC stays in the Commission’s analysis, a majority (64 percent) were solo stays, thus, not part of a sequence of stays. Of the 1.9 million multi- stay sequences, half involved stays in the same setting; the most common of these were back-to-back home health stays. Another third involved beneficiaries who transitioned from more intensive to less intensive settings. The most common of these were SNF and IRF stays followed by home health stays. Far less frequently, beneficiaries transitioned from less intensive to more intensive settings, most commonly from home health care to SNF care. Our analysis of sequential PAC stays, if paid under our prototype PAC PPS (which adjusts payments based on patient characteristics), found that patterns of costs relative to estimated payments over the course of care differed for home health stays and institutional PAC stays. For home health stays, payments under a unified PAC PPS would decrease over the course of a sequence of stays, but the cost of stays would decline more. As a result, later home health stays in a sequence would be more profitable than earlier stays, with stays that occurred later in longer sequences being the most profitable. These results suggest that payments need to be adjusted downward for later stays, similar to the adjustment used in the current HHA PPS. By contrast, PAC PPS payments for institutional stays would remain reasonably well aligned with the cost of stays throughout a sequence of care. This finding indicates that the PAC risk adjustment adequately captures differences in the cost of Paying for sequential stays in a unified prospective payment system for post-acute care 88

111 institutional stays throughout a sequence of care, indicating no need for a separate adjustment to payments. However, under its current design, the prototype PAC PPS would not be able to appropriately pay a PAC provider that offered a range of PAC services and was able to treat in place beneficiaries with evolving care needs (that is, not refer them to another PAC provider), even though such in-place treatment might be optimal for beneficiaries requiring PAC and operationally and administratively easier for providers (assuming the regulatory flexibility to do so). Under current policy, these beneficiaries are typically discharged to a second setting, and Medicare makes two payments for the patient’s PAC, one to each provider. Under a PAC PPS, providers will have more flexibility to offer a continuum of services to patients with evolving care needs, but, for payment purposes, Medicare will need to define when one “stay” or phase of care ends and the next one begins. Otherwise, with only one admission and discharge date, providers would receive only one payment, creating a financial disincentive to treat in place. Of the approaches we examined, the most promising would involve episode-based payments; that is, Medicare would pay for all PAC provided during an episode of care. The episode would include only PAC and would exclude other services furnished during the episode, such as hospital care or physician services. Payments for the episode of PAC would be set prospectively using a unified PAC PPS, with no reconciliation to a target benchmark. Payment for the PAC could be made to a hospital, a health system, the PAC provider where the episode starts, an accountable care organization, or a third-party convener that assumes financial risk for the episode of PAC. Under this approach, Medicare would not need to define and set payments for subsequent stays because the entity would be paid for the PAC provided during the episode, regardless of how many stays, settings, or providers were included. Further, a payment adjuster for later home health stays would not be needed because payments for the episode of PAC would be based on the average cost of the PAC for the full duration of the episode, including lower cost PAC later in the episode. Though episode-based payments could require an entity receiving payment from Medicare to pay all PAC providers involved in the care, such an arrangement would be necessary only for the small share of sequential stays that involved more than one provider. We expect this share to decline under a PAC PPS as entities evolve to offer a continuum of PAC. Entities would gain valuable experience managing PAC across a continuum before they embarked on assuming more responsibility for caring for beneficiaries. The incentive for entities receiving payment to stint on the June 2017 | Report to the Congress: Medicare and the Health Care Delivery System 89

112 amount or quality of services furnished (to keep costs low) could be countered with value-based purchasing. Episode-based payments would require a certain level of infrastructure for the minority of PAC stays that involve multiple providers, but the Commission contends that the advantages of this approach substantially outweigh its complexities. The Commission will continue to explore episode-based payments over the coming year. Shifting the unit of service from a stay to an episode would change certain incentives (most notably the incentive to initiate PAC stays), but the most important features of a PAC PPS would remain: correcting the biases of the current PPSs and increasing the equity of payments across all types of stays so that providers have less incentive to selectively admit certain beneficiaries over others. Shifting to an episode-based payment would incorporate these strengths into a bolder approach to a PAC PPS. In the meantime, CMS should proceed with implementing a stay-based unified PAC PPS. ■ Paying for sequential stays in a unified prospective payment system for post-acute care 90

113 the Commission recommended that, in anticipation of a Background transition to a unified PAC PPS, CMS should begin to base payments to providers in each of the PAC sectors on Post-acute care (PAC) providers—skilled nursing a blend of the sector’s setting-specific relative weights and facilities (SNFs), home health agencies (HHAs), inpatient the unified PAC PPS relative weights. Doing so would rehabilitation facilities (IRFs), and long-term care hospitals begin to improve the equity of payments across conditions (LTCHs)—offer important recuperation and rehabilitation (Medicare Payment Advisory Commission 2018). 1 services to Medicare beneficiaries. In 2016, Medicare fee-for-service (FFS) spending on these services totaled $60 billion. However, Medicare’s payments for a similar Challenges with paying for sequential case treated in different settings can differ substantially, in part because Medicare uses separate prospective payment post-acute care stays systems (PPSs) to pay for stays in each setting. Some of the difference in payments reflects the considerably The Commission’s initial work on a unified PAC PPS, different cost structures and the regulatory and statutory presented in the June 2016 report to the Congress, requirements for each setting. At the same time, there is considered each PAC stay as an independent event a lack of evidence-based criteria guiding decisions about (Medicare Payment Advisory Commission 2016). Yet, where patients should receive PAC and how much care many PAC stays are the second or third (or more) in a they should receive. The only study to compare outcomes series of PAC stays, in which patients transition from across the settings for a broad range of clinical conditions one setting or provider to another during their course of did not find consistent differences in rates of readmission care. In an FFS payment system like the unified PAC to hospitals or in improvement in mobility or self-care PPS, sequential stays present two potential challenges to (Gage et al. 2012). These factors contribute to considerable payment accuracy. First, throughout a course of care, a variation in the supply and use of PAC providers across the beneficiary’s clinical condition is likely to change such country. Results from the Center for Medicare & Medicaid that subsequent PAC stays may have different average Innovation’s Bundled Payments for Care Improvement costs than initial PAC stays. If payments for subsequent (BPCI) initiative indicate that, while the use of PAC did not stays are too high, providers such as those that are part decline, the mix of services shifted away from institutional of a system of care or HHAs that can recertify additional PAC and toward home health care, indicating that patients stays have an incentive to refer patients for unnecessary in the settings overlap. additional PAC stays, which could expose beneficiaries to undue risk and would increase program spending. If Given the overlap among settings for treating similar payments for subsequent stays are too low, providers patients, the Commission has long promoted the idea could avoid admitting these beneficiaries for necessary of moving to a unified system to pay for PAC in FFS additional care. Medicare using a PPS that spans the four settings, with payments based on patient characteristics rather than site The second challenge related to sequential stays centers 2 of service. As mandated by the Improving Medicare Post- on how to pay institutional providers for treating 3 Acute Care Transformation Act of 2014 (IMPACT), the beneficiaries whose care needs evolve over time. Commission, in June 2016, recommended the necessary Currently, patients treated in institutional settings who features of a PAC PPS and considered the implications of need additional PAC typically transition from one moving to such a system (Medicare Payment Advisory setting to another. For payment purposes, each stay has Commission 2016). Using readily available data, the a clearly defined beginning and end, and Medicare pays Commission’s PAC PPS design accurately predicted for each stay separately. As regulatory requirements for the costs of stays for most patient groups. In June 2017, institutional PAC settings begin to be aligned under a the Commission focused on several implementation unified PAC PPS, institutional PAC providers would issues, including the need for a transition to this new have the flexibility to offer a continuum of services to payment system, the level at which to set payments when beneficiaries who require different levels of care. In such the system is implemented, and the need for continued circumstances, however, the “end” of one stay and the monitoring and periodic refinements over time to keep “beginning” of another would not be clear. Yet, being payments aligned with the cost of care (Medicare able to distinguish between the stays would be important Payment Advisory Commission 2017). In March 2018, to pay for these services accurately. Otherwise, providers June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 91

114 beginning in 2021, which is sooner than the time table for would have a financial incentive to discharge patients to the studies required by IMPACT. In the Act’s schedule of another PAC provider, exposing beneficiaries to the risks associated with transitions of care. required reports on a PAC PPS design, it is unlikely that a new payment system would be proposed before 2024 for implementation at some later date. And while the Act requires recommendations for a design, it does not require Summary of the proposed PAC PPS the implementation of a PAC PPS. design In 2017, the Commission reported that the level of current PAC payments was high relative to the cost of stays (14 Based on its analysis of 8.9 million PAC stays in 2013 percent higher) and, for that reason, determined that the and using readily available administrative data, the implementation of the new system should not be budget Commission concluded that a unified PPS is feasible neutral. In 2017, the Commission recommended, based on (Medicare Payment Advisory Commission 2016). A PAC its analysis of 2013 PAC stays (with costs and payments PPS design would establish accurate payments using a updated to 2017), that the Congress direct the Secretary to uniform unit of service (a stay, which, in the case of home implement a PAC PPS beginning in 2021, with a three-year health care, is defined as an episode) and a uniform risk transition and payments lowered by 5 percent (absent any adjustment method. The Commission found the following prior payment reductions made to any setting’s payments). factors to be important predictors of costs that should Concurrently, the Secretary should begin to align setting- be considered in the design: the patient’s age, disability specific regulatory requirements (Medicare Payment status, comorbidities (and the number of body systems Advisory Commission 2017). The Commission believes that involved), severity of illness, risk score, cognitive status, its recommended design could be adopted on this timetable. and impairments; the primary reason to treat; the length of stay in an intensive care unit during the prior hospital In March 2018, the Commission recommended that stay (if any); and the use of select high-cost services the Congress direct the Secretary to begin to increase 4 (such as dialysis and mechanical ventilation). The design the equity of each PAC setting’s PPS payments before should include an adjustment for stays provided by HHAs implementing the unified PAC PPS. To do so, CMS because of their much lower costs and for two outlier would base each PAC setting’s payments on a blend of policies—one for unusually high-cost stays and another the proposed PAC PPS relative weights and the current for unusually short stays. The Commission examined the setting-specific relative weights. Using this blend would accuracy of PAC PPS payments for more than 40 patient redistribute payments in each setting’s PPS toward groups before concluding that an initial design could be medically complex stays (Medicare Payment Advisory based on readily available data. Commission 2018). This approach would also give providers more time to adjust their costs and practices to The proposed PAC PPS would redistribute payments the incentives of the new payment system. and narrow the differences in profitability of different types of stays (Medicare Payment Advisory Commission Medicare has different regulatory requirements for PAC 2017, Medicare Payment Advisory Commission 2016). settings, in part to differentiate one level of care from Payments would decrease for rehabilitation care unrelated another, even though the conditions they treat overlap. to patient characteristics (for example, for patients Under the proposed PAC PPS, with payments based on recovering from hip surgery who receive high amounts patient characteristics (and not setting), it would be less of rehabilitation therapy services unrelated to their care important to distinguish among types of institutional needs) and increase for medically complex care (for PAC providers. Furthermore, it would be unreasonable to example, patients with comorbidities that involve multiple maintain different regulatory requirements, with varying body systems). Because PAC PPS payments would be associated costs, for providers that will be paid the same based on the average cost of stays across the four settings, amount for the same type of patient. Policymakers would the new payment system would also redistribute payments need to align the regulatory requirements across the across settings, with payments shifting from the high-cost institutional PAC settings by waiving or altering some LTCH and IRF settings to the lower cost settings. of the current requirements. The Commission proposed a two-part strategy. In the near term, concurrent with the Because payments would be more accurate and equitable, implementation of the PAC PPS, some of the current the Commission recommended implementing a PAC PPS Paying for sequential stays in a unified prospective payment system for post-acute care 92

115 differences among institutional settings in establishing regulatory requirements would be waived or modified, payments for these providers and would separately adjust thereby establishing common requirements across institutional settings that help ensure quality of care. payments for home health stays to align payments to the considerably lower costs of this setting. In the longer term, CMS could define a common set of requirements for all PAC providers for participation and additional requirements for providers opting to treat patients with specialized care needs, such as those Characteristics of sequential PAC stays requiring ventilator or severe wound care. As background to our analysis of the costs of and payments for sequential stays, we first examined the patterns of PAC (Figure 4-1, p. 95). Of the thousands of Definition of sequential PAC stays multi-stay sequence patterns, the 10 most frequent patterns made up three-quarters of these sequences. Multiple home Although a majority of beneficiaries have just one PAC health stays were the most common. Stay sequences with stay after discharge from the hospital, many beneficiaries decreasing intensity were three times as frequent as those have a series of stays before their episode of illness with increasing intensity. resolves. To examine these stays, we used beneficiary identifiers and admission and discharge dates to link Beneficiaries with solo stays differed from those with multi- sequences of PAC stays together. This method allowed us stay sequences. Among home health stays, beneficiaries to identify common trajectories of PAC use (e.g., a single with multi-stay sequences were more likely to be dually IRF stay, a SNF stay followed by a home health stay, eligible for Medicare and Medicaid, disabled, and admitted back-to-back home health stays). from the community, while beneficiaries with multiple A refers to care furnished to a sequential PAC stay institutional PAC stays were less likely to have those beneficiary with short or no gaps in between the stays characteristics. Compared with providers of solo home (see text box, p. 94, defining sequential PAC stays). For health stays, providers of multi-stay sequences were our analysis, we defined a sequential stay as one that more likely to be for profit and freestanding. In contrast, began within seven days of another PAC use. These rules institutional PAC providers of multi-stay sequences were are rough proxies for clinical relatedness while allowing more likely to be nonprofit and hospital based compared some flexibility in how quickly home health care can with providers of solo institutional PAC stays. be arranged (changes in institutional PAC setting stays Frequency of sequential PAC stays typically involve transferring the beneficiary with no days in between the stays). Sequences include stays in the We identified 5,762 combinations of PAC stays in 2013. same setting and in different settings. A “first” stay was About two-thirds (64 percent) of the stays were solo defined as having no PAC use within the previous seven events—that is, consisted of a single stay. Of solo stays, days. A SNF stay followed by a home health episode that home health stays made up the majority (67 percent), began within seven days of discharge from the SNF was while SNF stays made up 28 percent, IRF stays another 4 considered a two-stay sequence. We assigned stays to the percent, and LTCH stays about 1 percent. following groups based on the dates of the stay: About one-third (36 percent) of the combinations involved • Solo (first-and-only) stays consisted of one admission multiple stays, with beneficiaries transitioning from one to one PAC provider, with no subsequent care. PAC setting or provider to another during their course of care. Pairs of PAC stays were the most common multi- First-of-multiple stays were the first in a sequence of • stay sequence (see online Appendix 4-A, available at PAC stays. http://www.medpac.gov, for information on the 25 most common sequences). Half of the sequential stays were • Subsequent stays were the second, third, or later in a lateral transitions within the same setting. The most sequence of PAC. frequent of these lateral, same-setting sequences consisted We aggregated the three institutional-type stays into a of home health stays only. Beneficiaries who moved from single “institutional PAC” group to reflect how a PAC more intensive PAC care to less intensive care made up 5 PPS would pay for this care. The PAC PPS would ignore one-third of multi-stay sequences. Transitioning from a June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 93

116 Defining sequential PAC stays PAC. Dates were used to establish the sequence and onsistent with previous work, characteristics of assign stays to a position in the sequence, such as beneficiaries and stays were assigned based on “second stay,” “third stay,” and so on. For example, a information from claims, Medicare Advantage C second home health stay was second in a sequence (of risk scores, and the beneficiary enrollment file. any length) that included either another home health To create sequences of post-acute care (PAC), we began stay or an institutional PAC stay as the first stay. In with the 8.9 million PAC stays in 2013 that we used in our analyses of costs and payments, we examined our previous analysis of the unified PAC prospective position and sequence length to separate the effects of payment system (PPS) (Medicare Payment Advisory sequence length from position in the sequence. Our Commission 2017, Medicare Payment Advisory final analytic sample included 5,334,377 sequences Commission 2016). Beneficiaries with overlapping comprising 3,435,192 solo stays and 1,899,185 multi- start and end dates for institutional PAC stays or with stay sequences. duplicate start dates for institutional PAC stays were Current billing rules govern what constitutes a stay, and excluded from the analysis. These exclusions removed our analysis did not redefine stay parameters. Given the 12,479 stays from the analytic file. Home health separate PPS for each of the four settings, differences stays with start and end dates that overlapped with exist among settings in how intervening events, such institutional PAC stay dates remained in the analysis as hospitalizations, define stays. In SNFs, for example, because a beneficiary could discontinue a home health stays interrupted by a hospitalization are considered care episode and enter into an institutional PAC setting separately (as two stays), while a single home health before the end of the 60-day home health episode. episode continues after an intervening hospitalization. A “first” stay was defined as having no PAC use An interrupted stay in inpatient rehabilitation facilities within the previous seven days. Subsequent stays and long-term care hospitals can trigger a separate were defined as stays that began within seven days stay, depending on the length of the interruption and 6 of another PAC use. Consistent with prior work, we the intervening event. In the future, when a common aggregated a beneficiary’s separate skilled nursing set of requirements is developed for PAC providers’ facility (SNF) claims to create a stay. Sequences could participation, billing rules and the treatment of include any combination of home health or institutional interrupted stays could be defined uniformly. ■ SNF or an IRF to home health care was the most common Multiple stays in HHAs were the most common: combination of stays of decreasing PAC intensity. Far Sequential home health stays made up 42 percent of all less frequently (10 percent of multi-stay sequences), multi-stay sequences, with a pair being the most frequent beneficiaries were discharged from a lower level of PAC (21 percent of multi-stay sequences). What appears to be to a more intensive setting. Presumably, this trajectory continuous home health care during the year (six or more reflects a change in care needs of the beneficiary and episodes) made up 7 percent of multi-stay sequences. capabilities of the provider or caregiver at home. Of those, Characteristics of solo and multiple home transitions from a home health stay to a SNF stay were health stays the most frequent. The remaining 7 percent of sequences were a mixed pattern of transitions (of increasing and To assess whether there were differences between decreasing intensity over the course of care), the most beneficiaries with solo home health stays and beneficiaries frequent being transitions back and forth between SNFs with multiple stays that included home health stays, we and HHAs. compared the beneficiaries’ characteristics and primary reason for treatment. We compared home health stays that Of the thousands of multi-stay sequence patterns, the 10 were solo, first of multiple stays, and subsequent stays most frequent made up three-quarters of these sequences. Paying for sequential stays in a unified prospective payment system for post-acute care 94

117 FIGURE FIGURE Medicare margins... X-X The 20 most frequent patterns of post-acute care, 2013 4–1 H S HH SH HHH SS I HHHHHH IH HS HHHH L HHHHH SHH Pattern of post-acute care SSH IS SSS SHS HSH LS 2,500,000 1,500,000 1,000,000 500,000 2,000,000 0 Count Note: H (stay treated in home health agency), S (stay treated in skilled nursing facility), I (stay treated in inpatient rehabilitation facility), L (stay treated in long-term care hospital). A sequence shows the order and setting of the stays. For example, “LS” refers to a sequence that started with a long-term care hospital stay and was followed by a skilled nursing facility stay. The 8.9 million post-acute care (PAC) stays were provided in 5,334,377 sequences of PAC. Source: Analysis of 2013 PAC stays conducted for the Commission by the Urban Institute (Wissoker and Garrett 2018). Note: Note and Source are in InDesign. Source: in a sequence. (Home health stays that were the first of position in the sequence (second stay in a sequence, third multiple could be followed by PAC stays of any type— stay in a sequence, etc.). The shares of the most frail and including SNF, IRF, and LTCH stays. Subsequent home chronically critically ill decreased as the position in the Notes about this graph: health stays could be preceded and followed by any type sequence increased. There were not large differences • Data is in the datasheet. Make updates in the datasheet. of PAC care.) between solo home health and first-of-multiple home • WATCH FOR GLITCHY RESETS WHEN YOU UPDATE DATA!!!! health stays in the shares of very old (85 years or older), Among home health stays, first-of-multiple stays were cognitively impaired, beneficiaries with end-stage renal • The column totals were added manually. more likely to be for beneficiaries who were dually disease, and the least frail (data not shown). • I had to manually draw tick marks and axis lines because they kept resetting when I changed any data. eligible, disabled, and admitted from the community compared with solo stays (Table 4-1, p. 96). For example, The primary reasons for treatment were similar for solo • I can’t delete the legend, so I’ll just have to crop it out in InDesign. 73 percent of first-of-multiple home health stays were home health and first-of-multiple home health stays, with • Use direct selection tool to select items for modification. Otherwise if you use the black selection tool, they will reset to graph for beneficiaries who were admitted from the community two exceptions. A higher share of solo home health stays default when you change the data. (thus, 27 percent had a prior hospital stay). In contrast, (10 percent) were for beneficiaries recovering from an 55 percent of solo home health stays were admitted from orthopedic surgical condition (such as a joint replacement) • Use paragraph styles (and object styles) to format. the community (and 45 percent had prior hospital stay). compared with 2 percent for first of multiple (Table 4-1). • Data was from: R:\Groups\MGA\data book 2007\data book 2007 chp1 Among subsequent stays, the shares of dually eligible, Because home health care often follows an institutional disabled, and community admissions increased with the stay (in a SNF or IRF) for beneficiaries recovering from June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 95

118 TABLE Characteristics of beneficiaries with single or multiple PAC stays 4–1 Multiple Unusually Cardio- Chronically Dual Orthopedic Community Most body vascular Number Position in critically high ill frail admission systems Disabled eligible surgery medical of stays cost sequence All 8,877,513 10% 11% 5% 11% N/A 50% 15% 32 % 26% Home health stays 2,290,337 Solo 8% 24% 55% 7% 3% N/A 10% 13% 29 % 1,020,688 First of multiple 16 19 2 N/A 2 6 73 29 38 66 11 1,388,388 32 26 Second 3 N/A 9 17 7 36 581,866 Third 21 10 1 N/A 1 4 86 30 90 10 319,637 39 32 Fourth 4 1 N/A 1 22 9 Fifth 22 0 N/A 0 4 92 33 41 196,815 Sixth 0 N/A 0 3 94 34 43 125,718 22 8 Institutional post-acute care stays 1,144,855 Solo 11% 17% 18% 21% 24% 33% 11% 11% 8% First of multiple 9 7 25 15 12 21 7 21 24 847,483 22 479,783 10 18 8 11 11 12 24 31 Second 19 8 12 11 32 164,420 25 15 22 6 Third 12 21 6 22 15 8 33 59,590 11 Fourth 26 24,018 Fifth 6 23 8 12 12 23 27 34 15 15 Sixth 9,255 34 27 15 25 7 23 8 12 Note: PAC (post-acute care), N/A (not applicable). “Institutional post-acute care” refers to stays in skilled nursing facilities (SNFs), inpatient rehabilitation facilities (IRFs), and long-term care hospitals (LTCHs). The table shows the share of stays with the respective characteristic(s). Because each row and column is independent, the rows and columns will not sum to 100 percent. “First-of-multiple” PAC stays are stays discharged to subsequent PAC settings—either home health or institutional PAC. Second, third, fourth, fifth, and sixth stays could be preceded and/or followed by PAC stays of any type, home health or institutional. For example, a third home health stay was third in a sequence of PAC stays, and the sequence could include home health and institutional PAC stays before and after the third stay. Dual-eligible beneficiaries are eligible for Medicare and Medicaid. “Most frail” refers to stays assessed as having most frail patients using the JEN Frailty Index. (The JEN Frailty Index is an algorithm that identifies frail older adults who may be at risk for institutionalization.) “Chronically critically ill” refers to stays for beneficiaries who spent eight or more days in an intensive care or coronary care unit. “Severely ill” refers to stays for patients who were treated in institutional PAC and categorized as severity of illness level 4 during the immediately preceding hospital stay. “Multiple body systems” refers to stays for patients with diagnoses that involved five or more body systems and were treated in institutional PAC settings (thus, “not applicable” in the home health portion of the table). “Unusually high cost” refers to stays that would be included in an outlier pool About 12,000 stays were excluded from the analysis because the dates on the claims set at 5 percent for home health stays and 5 percent for institutional PAC stays. overlapped. Other combinations of visits with seven or more stays in the sequence are not shown. Source: Analysis of 2013 PAC stays conducted for the Commission by the Urban Institute (Wissoker and Garrett 2018). orthopedic surgery, the share of second stays for this (Institutional stays that were the first of multiple could be condition jumps to 9 percent. The share of stays for followed by PAC stays of any type—including SNF, HHA, beneficiaries being treated for a cardiovascular medical IRF, and LTCH. Subsequent institutional stays could be condition was higher among first-of-multiple stays preceded and followed by any type of PAC care.) compared with solo stays (19 percent vs. 13 percent). The patterns for institutional PAC stays were opposite those for home health stays. First-of-multiple stays were Characteristics of solo and multiple less likely than solo stays to be for beneficiaries who were institutional PAC stays dually eligible, disabled, or admitted from the community. To assess whether there were differences between For example, 24 percent of first-of-multiple stays were for beneficiaries with solo institutional PAC stays versus dual-eligible beneficiaries compared with 33 percent of beneficiaries with multiple stays that included one or more solo stays. The frequency of these characteristics increased institutional PAC stays, we compared the beneficiaries’ with the timing of the stay, though differences were small. characteristics and primary reason for treatment. We In multi-stay sequences, indicators of patient complexity compared institutional solo stays, first-of-multiple (the shares of beneficiaries who were most frail or who sequences, and subsequent stays in a PAC sequence. had conditions that involved multiple body systems) Paying for sequential stays in a unified prospective payment system for post-acute care 96

119 TABLE Provider characteristics of post-acute care stays, 4–2 by position of the stay in a sequence of care Hospital Number of stays Nonprofit For profit Position in sequence based Freestanding 11% All 8,877,513 27% 89% 70% Home health stays 36 Solo 2,290,337 % 61 % 14 % 86 % First of multiple 1,020,688 21 76 8 92 Second 91 9 72 25 1,388,388 581,866 79 19 7 Third 93 319,637 17 81 6 Fourth 94 94 6 82 16 196,815 Fifth 95 5 82 15 125,718 Sixth Institutional post-acute care stays % 67 89 % % % 11 1,144,855 Solo 28 First of multiple 81 19 63 32 847,483 479,783 Second 34 71 8 92 Third 164,420 23 73 8 92 74 92 Fourth 59,590 22 8 Fifth 24,018 21 76 8 92 Sixth 8 92 76 9,255 19 Note: “Institutional post-acute care” includes stays in skilled nursing facilities (SNFs), inpatient rehabilitation facilities (IRFs), and long-term care hospitals (LTCHs). “First- of-multiple” post-acute care (PAC) stays include stays discharged to subsequent PAC—either home health or institutional PAC. Second, third, fourth, fifth, and sixth stays could be preceded and/or followed by PAC stays of any type, home health or institutional. For example, a third home health stay was third in a sequence of PAC stays, and the sequence could include home health and institutional PAC stays before and after the third stay. About 12,000 stays were excluded from the analysis because the dates on the claims overlapped. The 3 percent of stays provided in government providers are not shown. Analysis of 2013 PAC stays conducted for the Commission by the Urban Institute (Wissoker and Garrett 2018). Source: increased with the sequence’s stay count. For example, and multi-stay sequences, differences also were found 15 percent of first-of-multiple stays had conditions that between home health and institutional PAC stays (Table involved multiple body systems compared with 23 percent 4-2). Among home health stays, a larger share of first-of- of institutional stays that were the fifth and 23 percent that multiple stays (76 percent) were provided by for-profit were the sixth in a sequence of PAC stays. agencies compared with solo stays (61 percent), and the share increased for stays later in the sequence, reaching Differences in the clinical reasons for treatment were 82 percent of fifth and sixth stays. By type of HHA, a similar across institutional PAC stays, except that a larger smaller share of solo home health stays (86 percent) share of first-of-multiple stays compared with solo stays were furnished by freestanding HHAs compared with 92 were for beneficiaries recovering from orthopedic surgery percent of first of multiple and, again, the shares of stays (25 percent of first-of-multiple stays vs. 17 percent of solo provided by freestanding HHAs increased for later stays, stays). Stays in longer sequences were for beneficiaries comprising 95 percent of sixth stays. who were generally more medically complex than for beneficiaries with shorter sequences. Differences in ownership and facility type were smaller among institutional PAC stays. Compared with solo Characteristics of providers of solo and institutional PAC stays, first-of-multiple stays were more multi-stay sequences frequently (32 percent) treated in nonprofit facilities compared with 28 percent of solo stays, and the share of In addition to differences in the ownership and type of providers (freestanding and hospital based) treating solo June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 97

120 Estimates of PAC costs and PAC PPS payments regression model and applied this model to the 2013 he 8.9 million post-acute care (PAC) stays PAC stays. The cost of each stay reflects, in part, the in 2013 that have been used in previous differences in costs across settings. Commission research on the unified PAC T prospective payment system (PPS) were the starting To estimate payments, the PAC PPS design relies on point for this work on sequential stays (Medicare models that predict the cost of each stay using patient Payment Advisory Commission 2016). To estimate and stay characteristics. The following patient and the costs of each stay, information from claims stay information was used to predict the cost of each and Medicare cost reports and—as required by the stay: patient demographics (e.g., age and disability), Improving Medicare Post-Acute Care Transformation primary reason to treat, comorbidities, cognitive status, Act of 2014—data from CMS’s Post-Acute Care impairments (e.g., difficulty swallowing and bowel Payment Reform Demonstration (PAC–PRD) were incontinence), measures of severity, and use of special used. Therapy and nontherapy costs were estimated treatments (e.g., ventilator care). We included these using 2013 PAC claims and 2013 Medicare cost reports factors in the risk adjustment because they captured (see online Appendix 4-A, available at http://www. different dimensions of a patient that could influence medpac.gov, for a full discussion of the methodology). the cost of care. The Secretary could consider other We took advantage of the unique stay-level information dimensions or other measures of the same dimensions on routine costs collected in the PAC–PRD (and not in the final design. available elsewhere) to estimate routine costs using a (continued next page) . Although estimated PAC costs and PAC PPS payments) stays treated in nonprofit facilities decreased in later stays. PPS payments (which adjust for differences in patient A higher share of first-of-multiple institutional PAC stays characteristics) for institutional PAC stays would be were furnished by hospital-based providers (19 percent) compared with solo institutional PAC stays (11 percent), aligned with the lower average costs of later stays, PAC but the mix across later stays was the same. PPS payments for home health stays would not be. As a result, later home health stays would be increasingly We did not explore whether providers that are members profitable. These findings suggest the need for a payment of vertically integrated systems (with different settings adjustment for later home health stays similar to the included in their holdings) have different patterns of care. adjustment in the home health PPS. Otherwise, providers Given a common financial stake, providers with these will have an incentive to furnish additional stays. linkages would have an incentive to refer beneficiaries to subsequent care. Why costs might vary throughout a sequence of care It is possible that the average costs of stays differ throughout a sequence as patients’ care needs evolve. PAC PPS payments need to align with Early stays are more likely to include beneficiaries the cost of stays throughout a sequence recovering from acute events and receiving services aimed of post-acute care at getting the beneficiaries functioning as independently as possible. Later PAC stays may focus on strengthening Our analysis found that the average cost of stays declined beneficiaries and managing chronic conditions, which over the course of sequential PAC stays, especially for may require fewer resources. In addition, stays may home health stays (see text box on estimates of PAC Paying for sequential stays in a unified prospective payment system for post-acute care 98

121 Estimates of PAC costs and PAC PPS payments (cont.) unusually high-cost stays and helps ensure beneficiary We used Poisson regression models and developed one access to services. A short-stay policy protects the model to predict the costs of routine and therapy care program and taxpayers from excessive payments that for stays in the four PAC settings and a separate model would otherwise be paid for unusually short stays. to predict nontherapy ancillary (NTA) costs for stays Instead of being paid a full stay amount, short stays in skilled nursing facilities, independent rehabilitation are paid a daily rate for the duration of the stay. (For facilities, and long-term care hospitals. We developed details of these designs, see the Commission’s June a separate model for NTA services because the home 2016 report to the Congress (Medicare Payment health care benefit does not cover these services. Advisory Commission 2016).) Payments were adjusted A home health indicator was included in the model for budget neutrality so that total payments across the to account for this setting’s considerably lower four settings are the same as under the current payment costs compared with institutional PAC. Without this systems. adjustment, home health stays would be substantially overpaid and the other PAC providers would be The payments and costs were updated from 2013 to substantially underpaid. The design does not consider 2017 (Medicare Payment Advisory Commission 2017). differences in costs across institutional settings in To estimate payments in 2017, payments were updated establishing payments for stays. using each setting’s market basket update net of the adjustments made by CMS (e.g., for productivity and Payments also include two outlier policies—one for any coding adjustments). Costs were updated to 2017 unusually high-cost stays and another for unusually using the average cost increases by PAC setting. short stays. A high-cost outlier policy protects providers ■ from incurring exceptionally large losses from treating have different average costs throughout a sequence if than the first). In a five-stay home health sequence, the they involve a different mix of settings. Beneficiaries average cost of the fifth stay was 26 percent lower than the may transition between settings as they no longer meet first stay ($1,896 compared with $2,574 for the first stay). coverage requirements for a given setting. However, Beneficiary characteristics are unlikely to explain these distinctions between the costs of home health care and large cost differences, which is consistent with findings institutional PAC were already considered in a PAC PPS from extensive work conducted for the Commission on the design, while differences across institutional PAC settings cost of home health episodes (Wissoker and Garrett 2015) are intentionally not factored into payments (payments (see online Appendix 4-A, available at http://www.medpac. are “site neutral”). Therefore, the cost differences due gov, for more information). That work found that clinical to setting should not be a factor in evaluating whether characteristics explain little of the variation in costs across payments require further adjustment. episodes. If payments are not aligned to the declining cost of stays, later stays will be increasingly profitable and The average cost of stays declines create an incentive for HHAs to furnish additional stays. throughout a sequence of care The average cost of institutional PAC stays generally The average cost of home health and institutional PAC declined throughout a sequence, though the pattern was stays declined throughout a course of care. For home a little more variable and the differences were smaller health stays, the average cost of last stays in the sequence compared with home health stays. Except for the two-stay was considerably lower than the cost of a first stay in sequence, the costs of later stays were between 7 percent the sequence (Table 4-3, p. 100). For example, in two- and 12 percent lower than first-stay costs. Compared with stay sequences, the cost of the first stay averaged $2,699 later stays, first-stay costs were higher in part because compared with $2,278 for the second stay (16 percent lower they involved a costlier mix of settings (with higher June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 99

122 TABLE Average costs of stays generally decline over a sequence of care 4–3 Sequence length (number of stays) 2 stays 1 stay Position in sequence 3 stays 4 stays 5 stays 6 stays Home health stays $2,574 First $2,190* $2,699 $2,174 $2,611 $2,592 2,278 2,565 2,430 2,356 2,056 Second Third 2,087 2,343 2,226 1,986 Fourth 1,982 2,204 1,982 1,896 Fifth 1,979 Sixth 1,790 Institutional post-acute care stays First $13,948 $15,191 $16,097 $16,740 $17,506 $14,245* 14,318 14,334 14,785 15,162 16,147 Second Third 14,100 14,821 15,205 15,966 Fourth 15,052 15,784 14,287 Fifth 14,677 16,016 Sixth 16,246 Note: Second, third, fourth, fifth, and sixth home health stays could be preceded and followed by post-acute care (PAC) stays of any type—including skilled nursing facility (SNF), inpatient rehabilitation facility (IRF), and long-term care hospital (LTCH) stays. Second, third, fourth, fifth, and sixth institutional stays could be preceded and followed by PAC stays of any type—including SNF, home health, IRF, and LTCH. For example, a third home health stay was third in a sequence of PAC stays, and the sequence could include home health and institutional PAC stays before and after the third stay. *The first stay in a one-stay sequence is a solo stay. Source: Analysis of 2013 PAC stays conducted for the Commission by the Urban Institute (Wissoker and Garrett 2018). shares of stays in IRFs: 21 percent of first-of-multiple sequence, payments for the first stay would be 5 percent stays compared with 10 percent of fifth stays (data not higher than the average cost (a payment-to-cost ratio shown)). If risk adjustment does not adequately capture (PCR) of 1.05), but payments for the third stay would be the differences in patient complexity throughout the 24 percent higher than costs (PCR = 1.24). The pattern of sequence, later stays will be less profitable, and providers increasing profitability was consistent across sequences, of subsequent stays could be discouraged from admitting and later stays in longer sequences were more profitable these beneficiaries, creating placement problems for compared with earlier stays. For example, the PCR for the beneficiaries with extended PAC needs. last stay in the two-stay sequence was 1.16 but increased to 1.41 for the last stay in a six-stay sequence. Ideally, Profitability would increase throughout a differences in the cost of stays would be captured by the sequence of home health care but remain case-mix adjusters. However, the higher profitability relatively uniform for institutional PAC stays for later home health stays suggests the need for an We found that payments estimated by our prototype PAC adjustment to payments based on the timing of the stay PPS design for home health stays were not evenly aligned to more closely align payments with costs. Otherwise, with these stays’ declining costs, so that later stays were HHAs could generate additional profits by recertifying considerably more profitable than earlier stays (Table 4-4). beneficiaries for additional home health care, assuming PAC PPS payments are risk adjusted for differences in the beneficiary continued to meet coverage rules. Such a patient characteristics (see text box on estimates of costs refinement of the PAC PPS would be consistent with the and payments, pp. 98–99). For example, in a three-stay current payment system for HHAs that lowers payments Paying for sequential stays in a unified prospective payment system for post-acute care 100

123 TABLE Under our proposed PAC PPS, payment-to-cost ratios would increase for later 4–4 home health stays but would be relatively uniform for institutional PAC stays Sequence length (number of stays) 4 stays 3 stays 2 stays Position in sequence 5 stays 6 stays 1 stay Home health stays First 1.07 1.22 1.06 1.05 1.01 1.16* Second 1.08 1.13 1.16 1.22 1.16 Third 1.24 1.16 1.21 1.31 Fourth 1.29 1.22 1.31 1.31 1.34 Fifth 1.41 Sixth Institutional post-acute care stays 1.14* 1.17 1.12 1.10 1.08 1.05 First 1.13 1.14 1.12 Second 1.12 1.08 1.07 1.14 1.11 1.10 Third Fourth 1.14 1.11 1.08 1.08 Fifth 1.13 1.06 Sixth Note: PAC (post-acute care), PPS (prospective payment system). The ratio of payments to costs is a measure of profitability. Payments are estimated PAC PPS payments. Institutional post-acute care includes stays in skilled nursing facilities (SNFs), inpatient rehabilitation facilities (IRFs), and long-term care hospitals (LTCHs). Second, third, fourth, fifth, and sixth home health (HH) stays could be preceded and followed by PAC stays of any type—including SNF, IRF, and LTCH stays. Second, third, fourth, fifth, and sixth institutional stays could be preceded and followed by PAC stays of any type—including SNF, HH, IRF, and LTCH. For example, a third home health stay was third in a sequence of PAC stays, and the sequence could include home health and institutional PAC stays before and after the third stay. *The first stay in a one-stay sequence is a solo stay. Source: Analysis of 2013 PAC stays conducted for the Commission by the Urban Institute (Wissoker and Garrett 2018). for third and later episodes of home health care. The be lower for institutional PAC stays, they would remain changes to the HHA PPS proposed by CMS in 2017 also well above the cost of stays. The results for institutional include a large adjustment for subsequent stays to reflect PAC stays indicate that the risk adjustment included in the lower average resource use for these episodes (for the proposed PAC PPS design would do a reasonable job example, a 39 percent reduction for later stays admitted capturing the differences in patients’ characteristics across from the community) (Centers for Medicare & Medicaid stays in a sequence. An additional payment adjustment Services 2017). based on the order of the stay in a sequence of care is not needed for institutional PAC stays. In contrast, PPS payments for institutional PAC stays would be more consistently aligned with the cost of stays throughout sequences, with much smaller variation in Defining the beginning and end of stays the profitability across stays in a sequence. Although the profitability of stays would generally increase for when treating in place later stays, the patterns would be more uneven and the differences would be much smaller. For example, the PAC Under a unified PAC PPS with modified regulatory PPS payments for three-stay sequences would range from requirements, some providers may choose to treat a 12 percent to 14 percent higher than the average cost of broader range of patients than they can under current stays (payment-to-cost ratio of 1.12 to 1.14). For five-stay policies, opting to treat “in place” patients who require sequences, the PCRs would range from 1.08 for the first changing levels of care during an episode of illness rather stay to 1.13 for the fifth stay. While profitability would June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 101

124 FIGURE XXXXXX x-x FIGURE Comparison of the number of stays under the proposed PAC PPS design 4–2 when institutional PAC providers refer beneficiaries to another provider and when they opt to treat in place There would be two stays when a beneficiary is Stay 2 Stay 1 referred to a second provider for additional care. When a beneficiary is treated in place by the same provider under the proposed PAC PPS, there would Stay 2 Stay 1 be one stay unless a second stay is established for the second phase of care. Time Note: PAC (post-acute care), PPS (prospective payment system). than refer them to another provider. A patient could remain untoward outcomes from a poor transfer, providers opting at the same facility and receive intensive services for the to treat in place should not discouraged. early portion of care and less intensive services as recovery Define a stay based on time progresses. IRFs and LTCHs could opt to treat patients with less intensive care needs (as opposed to transferring One approach to defining the beginning and end of stays them to SNFs), while SNFs could opt to offer services when treating in place would be to use a fixed period of 7 that previously had been furnished by IRFs and LTCHs. time—a threshold—to define when the first stay ends. A Reducing the number of handoffs between providers provider would be paid a PAC PPS amount for the initial would lower the risk of poor transitions. stay, but if the stay reaches a certain length, providers would conduct a new assessment and would receive a Defining a stay is straightforward when a beneficiary is separate payment based on it. This method would be discharged from one provider and admitted to another; the similar to the day-based definition of home health episodes stay begins at admission to the first PAC provider and ends (currently 60 days, but the Balanced Budget Act of 2018 when discharged to the second (or when discharged home changes this period to 30-day episodes beginning in 2020). for home health care) (Figure 4-2). Sequential home health Note and Source in InDesign A day-based definition of a stay could be considered for care stays are also easy to identify because the unit of all stays, not just those furnished by providers treating in service is 60 days, with another home health stay triggered place. on day 61 of service. In both cases, Medicare makes two payments, one for each stay. The advantage of an approach based on length of stay is that it would be clear and relatively simple to administer. For institutional PAC providers furnishing a continuum The large downside is that it would encourage PAC of care, the end of one stay and the beginning of another providers to extend stays beyond the pre-set threshold would be less clear. CMS will need a way to distinguish to establish a subsequent stay and receive an additional between the different phases of care. Otherwise, with payment. Providers’ likely response to this financial one admission and one discharge, a provider opting to incentive would increase the share of stays that extend treat in place would receive one payment that may not beyond the threshold. Medicare’s experience with be sufficient to cover the costs of an extended phase of thresholds illustrates how providers typically adjust their PAC. Providers that treat in place would then be at a practices in response to thresholds (e.g., HHAs and SNFs financial disadvantage compared with providers that refer have been known to provide additional therapy visits or the beneficiary to another level of care. Yet, if treating in minutes—respectively—to qualify for higher case-mix place would offer comparable care and reduce the risk of Paying for sequential stays in a unified prospective payment system for post-acute care 102

125 payments and LTCHs to extend stays beyond the short- Change the unit of service to an episode of stay outlier threshold to qualify for full payment). post-acute care Another approach would circumvent the multiple issues Strategies to counter the incentive to raised by sequential stays by shifting the unit of service increase the volume of subsequent stays from a stay to an episode of PAC. The episode would Because providers would have an incentive to extend include only PAC and would exclude other services. This care past a threshold to generate subsequent stays, CMS approach differs in a couple of ways from the “virtual” would need to undertake multiple activities to guard bundled payment the Center for Medicare & Medicaid against uncontrolled volume increases. First, it would Innovation (CMMI) is testing with the Bundled Payments need to use a relatively long unit of service that would for Care Improvement (BPCI) initiative. Under the BPCI, encompass the majority of stays. Second, it would need to Medicare continues to make FFS payments to each develop a short-stay outlier policy, which would weaken provider, with retrospective reconciliation between total the incentive to extend initial stays to garner payment for actual spending and a benchmark amount. The entity is a second stay. That is, providers would have to extend a at risk for the cost of all services furnished during the stay beyond the day threshold to a number exceeding the episode, including any hospital care, additional PAC, short-stay outlier cut-off for the stay to qualify for another physician services, and ancillary services. The approach full payment. Third, recertification by a beneficiary’s that the Commission will explore is narrower in concept. physician could be required for the PAC provider to The unit of service for the PAC PPS would include receive an additional payment. Under such a policy, the all PAC for an episode of care, but no other services. physician would be required to review the plan of care, Medicare could make one payment to an entity to cover all attest to the continued need for PAC, and estimate how PAC within the episode. There would be no benchmarks or much longer services would be required, as is done for reconciliation. recertification for home health episodes. Finally, a value- If the unit of service for the PAC PPS were an episode of based purchasing program that included a measure of PAC, Medicare would not need to define and set payments resource use, such as Medicare spending per beneficiary, for subsequent stays because the entity would be paid for could also counter the incentive to generate volume since all PAC services provided during the episode, regardless of the added spending would count against the provider’s how many stays that included. Further, a payment adjuster performance. for later home health stays would not be needed because CMS would need to monitor the frequency of subsequent payments for the episode would be based on the average PAC and examine providers with aberrant patterns. cost of the PAC for the full duration of the episode, Inevitable differences in stay-level profitability, even if including lower cost care toward the end. small, could make certain practice patterns more attractive. An episode-based payment would require one entity to be For example, a large increase in subsequent PAC could financially at risk for the entire episode of care. The entity indicate that providers are delaying care until after the could be the first PAC provider, a health care system, stays are complete, thereby obtaining full payments for a hospital, an accountable care organization (ACO), a stays and lowering their costs or taking undue advantage physician group practice, or a third-party convener. This of the ability to treat in place to generate an additional entity would need to have the infrastructure to receive stay. Periodic reevaluation of the alignment of payments a lump-sum payment from Medicare and, in turn, make and costs would indicate whether the Secretary needed payments to any “downstream” PAC provider furnishing to revise the PAC PPS. The Commission previously care during the episode. If the first PAC provider is the recommended that the Congress grant the Secretary the entity at risk, it could opt to furnish all PAC for the episode authority to revise and rebase the PAC PPS over time to or refer the beneficiary to another PAC provider that it keep payments aligned with the cost of care (Medicare would pay. Given current practice patterns, we estimate Payment Advisory Commission 2017). that a minority of episodes (about 18 percent) would While it would be feasible to design and implement these involve paying more than one provider, and we would counter-incentive strategies, Medicare’s experience with expect this share to decline substantially under a PAC PPS 8 them suggests that they would not be effective. Many of as providers opt to offer a continuum of PAC. these strategies are currently in place but have not deterred the provision of PAC of questionable value. June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 103

126 Episode-based payments for providers choosing to treat those providers not already participating in alternative beneficiaries in place underscores the need to align payment models, an episode approach would give them Medicare coverage rules and beneficiary cost-sharing valuable experience managing beneficiaries across a requirements across PAC settings. For example, a prior continuum of care. For them, episode-based payment hospital stay of three days is currently required for SNF would represent a stepping stone to accepting more risk, coverage but not for HHA, IRF, or LTCH services. As which will be required under broader payment reforms. distinctions between particular institutional settings blur As practice patterns change under episode-based and providers opt to offer a broader mix of services, it payments, CMS would need to periodically evaluate would make sense to have one set of coverage rules. whether payments continue to be aligned with the cost Likewise, beneficiary cost-sharing requirements currently of care and adjust payments as needed. The Commission vary by setting. Standardized cost sharing would enable previously recommended that the Congress grant the beneficiaries to select PAC based on their care needs and Secretary the authority to revise and rebase the PAC PPS preferences rather than on financial considerations. over time to keep payments aligned with the cost of care. Advantages of episode-based payments Disadvantages of episode-based payments Using episodes as the unit of care would have numerous There are three potential downsides to episode-based advantages. First, an episode-based payment would payments. First, providers would have a financial incentive overcome the distortions inherent in volume-driven FFS to furnish fewer services than medically appropriate or payment. Providers would have an incentive to furnish a provide lower quality care if it lowered their costs. The mix of services to meet a beneficiary’s care needs over potential for providers to stint on care is inherent in any the entire PAC episode rather than to furnish more stays. prospective payment system. Second, with more dollars Results from CMMI’s BPCI initiative indicate participants at stake, episode-based payments could encourage more lowered their use of PAC, which may translate to fewer episodes, resulting in increased program spending. sequential stays (Lewin Group 2017). However, the risk of more episodes would be lower than If providers opted to treat in place rather than transfer the risk of unnecessary subsequent stays because the beneficiaries to another provider, there would be fewer decision to use PAC would be made by the beneficiary’s handoffs between providers, and beneficiaries would be physician in consultation with discharge planning staff less likely to experience poorly coordinated care. Having (as it is now), whereas, under the length of stay approach, one entity responsible for payment could also improve the decision to generate additional stays would be made care coordination among providers. Entities would be by the PAC provider. Last, an episode-based payment incentivized to improve their follow-up care and use would require the entity at risk to have the infrastructure case managers to oversee the PAC, strategies used by needed to pay multiple providers. Although episodes some ACOs, bundled payment conveners, and Medicare that involve multiple providers represent the minority Advantage plans. In this case, beneficiaries and their of episodes, some PAC providers would not be ready to families would have a better idea of whom to contact accept this level of financial risk or have the administrative with questions and concerns, thus overcoming a common infrastructure to set and make payments to other providers. criticism of FFS care. The Commission maintains that the administrative complexities of this approach are far outweighed by the Episode-based payment should, in no way, limit a advantages of episode-based payment. beneficiary’s choice of PAC provider. Because the entity in charge could seek to influence a beneficiary’s decision Strategies to counter the potential about where to get their PAC, Medicare would need to disadvantages of episode-based payments ensure that information given to beneficiaries to aid their To counter these disadvantages, CMS would need to decision making did not limit their choice to poor-quality monitor the frequency of PAC use and examine entities providers. with aberrant utilization patterns. Given the financial incentives of the current payment systems to furnish Another advantage of episode-based payments is that they unnecessary therapy care, changes from current practice would align the incentives of PAC providers with those of would not necessarily signal a worrisome trend. To alternative payment models (such as ACOs and bundled discourage unnecessary episodes, physicians could payments) that encourage low-cost, high-quality care. For Paying for sequential stays in a unified prospective payment system for post-acute care 104

127 be required to attest to the need for PAC. Value-based alternative payment models that require them to assume more risk. The Commission will explore this approach purchasing that included a measure of resource use could over the coming year. In the meanwhile, CMS should deter providers from delaying care until after the episode proceed with implementing a stay-based unified PAC window. One such measure, the Medicare spending per beneficiary–PAC, identifies spending during the PAC stay PPS. While shifting the unit of service from a stay to an episode would change certain incentives (most notably the plus 30 days after discharge. To detect stinting, a value- incentive to generate unnecessary PAC stays), the most based policy would also need to include quality measures, such as rates of potentially avoidable (or ambulatory important features of a PAC PPS would remain: correcting the biases of the current PPSs and increasing the equity of care–sensitive) readmissions and emergency room visits. It could also consider measures of care coordination, such payments across all types of stays so that providers have as the number of days between hospital discharge and the less incentive to selectively admit certain beneficiaries first physician visit or the number of transitions while the over others. A shift to an episode-based payment should, in no way, be interpreted as a temporary retreat from a beneficiary is away from her residence. PAC PPS. Rather, building on these basic features of a PAC PPS, the Commission will explore bolder approaches that focus providers’ efforts on considering beneficiaries’ Conclusion PAC needs throughout the duration of a PAC episode. ■ An episode-based PPS would discourage the provision of unnecessary PAC stays and would ready providers for June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 105

128 Endnotes intervention. If the intervening event is three days or less and We refer to all care furnished in home health agencies, 1 the beneficiary returns to the same facility, the original IRF inpatient rehabilitation facilities, and long-term care hospitals stay continues. If the intervening event is longer than three as “post-acute care,” even though some of the beneficiaries days or the beneficiary goes to a different facility after the were admitted from the community. The chapter includes intervening event, there are two IRF stays—one before the community admissions in all of its work on the unified PAC event and another after the event. In LTCHs, the duration of prospective payment system. the interruption and whether the beneficiary returns to the 2 In this chapter, we examine PAC use by FFS beneficiaries. We same LTCH define whether a separate stay is established. do not include PAC use by beneficiaries enrolled in Medicare An LTCH stay is counted as one if the intervening stay is in Advantage. an acute hospital and shorter than 10 days, in an IRF and is shorter than 28 days, or in a SNF and is shorter than 46 days. Subsequent care in HHAs does not present the same problem 3 If the intervening stay is longer than those limits or if the because each stay is clearly defined by the 60-day episode. beneficiary is transferred to a different LTCH, there are two LTCH stays. 4 The predictors and their relative importance in estimating payments under a PAC PPS were published in 2016 in a The Commission considered another approach that would 7 report prepared for the Commission by researchers at the define stays using a phase of care. As care needs evolved, a Urban Institute (Wissoker and Garrett 2016). provider would on paper “discharge” the beneficiary from the first phase and “admit” her to the second phase, triggering two The intensity of the setting is based on the following 5 payments. It was not clear whether criteria could differentiate hierarchy: LTCHs were considered the most intensive, a new phase of care from normal disease progression or followed by IRFs, then SNFs, and the least intensive was healing without the criteria being easily manipulated by home health care. providers. The difficulty of designing and monitoring this approach seemed unworkable. 6 Current billing rules establish definitions of stays. In a home health stay, an intervening hospital or institutional PAC stay 8 The estimate is based on the share of stay combinations that that occurs entirely during a home health care episode does are solo (64 percent) and the share of sequences that include not change the counting of the 60 days that define an episode lateral stays (18 percent), neither of which would involve and does not establish separate episodes for the care before paying different providers. Our data suggest that most lateral and after the intervening stay. For SNF stays, an intervening stays involve the same provider and that most are back-to- hospital or PAC stay establishes separate SNF stays, one back home health stays. Lateral institutional PAC stays are before the intervening event and another after. In IRFs, the most likely for stays interrupted by a hospitalization that duration of the interruption (for a hospital or other PAC triggered a new PAC stay. Far less frequently, beneficiaries stay) and whether the beneficiary returns to the same facility change PAC providers for any number of reasons, including establishes whether the original IRF stay continues after the proximity to family or dissatisfaction with the initial provider. Paying for sequential stays in a unified prospective payment system for post-acute care 106

129 References Report to Medicare Payment Advisory Commission. 2017. Centers for Medicare & Medicaid Services, Department of Health the Congress: Medicare and the health care delivery system. and Human Services. 2017. Medicare and Medicaid programs; Washington, DC: MedPAC. CY 2018 home health prospective payment system rate update and proposed CY 2019 case-mix adjustment methodology Report to Medicare Payment Advisory Commission. 2016. refinements; home health value-based purchasing model; and the Congress: Medicare and the health care delivery system. Federal home health quality reporting requirements. Final rule. Washington, DC: MedPAC. 82, no. 214 (November 7): 51676–51752. Register Characteristics, costs, Wissoker, D., and B. Garrett. 2018. Post-Acute Care Gage, B., M. Morley, L. Smith, et al. 2012. and payments for stays within a sequence of post-acute care. Payment Reform Demonstration: Final report, volume 1 of 4. Washington, DC: The Urban Institute. Prepared under contract to the Centers for Medicare & Medicaid Services. Baltimore, MD: CMS. Wissoker, D., and B. Garrett. 2016. Designing a unified prospective payment system for post-acute care. Washington, DC: Lewin Group. 2017. CMS Bundled Payments for Care Medicare Payment Advisory Commission. Improvement Initiative models 2-4: Year 3 evaluation & monitoring annual report. Report prepared by the Lewin Group Wissoker, D., and B. Garrett. 2015. Simulation and analysis of an for the Centers for Medicare & Medicaid Services. Baltimore, alternative Medicare home health payment system not based on MD: CMS. https://downloads.cms.gov/files/cmmi/bpci-models2- Washington, DC: The Urban Institute. number of therapy visits. 4yr3evalrpt.pdf. Medicare Payment Advisory Commission. 2018. Report to the Washington, DC: MedPAC. Congress: Medicare payment policy. June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 107

130

131 CHAPTER 5 Encouraging Medicare beneficiaries to use higher quality post-acute care providers

132

133 CHAPTER 5 Encouraging Medicare beneficiaries to use higher quality post-acute care providers In this chapter Chapter summary About 40 percent of Medicare acute inpatient hospital discharges result in use Introduction • of post-acute care (PAC), which includes four provider types: skilled nursing • Beneficiaries seeking facilities (SNFs), home health agencies (HHAs), inpatient rehabilitation PAC often have many facilities, and long-term care hospitals. Ensuring that the patient is served by PAC options that vary the appropriate type of PAC provider is critical, but the selection of a provider substantially in quality a PAC category can be crucial because the quality of care varies widely within among providers. Increasing the use of higher quality PAC providers is • Patients referred to PAC particularly important as CMS implements value-based payment reforms, such need assistance to identify better quality providers as the Hospital Readmissions Reduction Program (HRRP), hospital value- based purchasing programs, and accountable care organizations (ACOs), • Principles for improving which hold providers accountable for the expenditures related to readmissions hospital discharge planning during a PAC stay. • Approaches for identifying Beneficiaries report that they value quality of care and that they prefer PAC higher quality PAC providers providers that are close to their home or family. Medicare discharge planning • Conclusion regulations place responsibility with hospitals for connecting inpatient acute care hospital patients with their options for PAC, including educating beneficiaries about their choices and facilitating access to PAC when necessary. Medicare regulations also require that hospitals consider patient preferences and guarantee beneficiary freedom of choice in selecting PAC providers, but hospitals are limited in the assistance they can provide. Though they are required to provide beneficiaries who need PAC with a list of nearby June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 111

134 SNFs and HHAs, Medicare regulations prohibit hospitals from recommending specific PAC providers. The Improving Medicare Post-Acute Care Transformation Act of 2014 requires hospitals to include quality data when informing beneficiaries about their options, but CMS has yet to finalize the regulations implementing this requirement. CMS has developed consumer-oriented websites that provide information on the quality of SNFs and HHAs, but many studies have concluded that these efforts have not significantly increased the use of higher quality PAC providers, possibly because beneficiaries are not always made aware of the data. The Commission’s analysis of referral patterns of Medicare beneficiaries who were sent to SNFs and HHAs indicates that, for many beneficiaries, another nearby provider offered better quality care, though not all of the higher quality providers may have had available capacity. For example, over 94 percent of beneficiaries who used HHA or SNF services had at least one provider within a 15-mile radius that had higher performance on a composite quality indicator than the provider they selected. About 70 percent of beneficiaries who received HHA services had 5 or more other HHAs within a 15-mile radius that offered better quality than their original provider, while almost half of SNF users had 5 or more options with better quality. Helping beneficiaries to identify better quality PAC providers should be a goal in a reformed discharge planning process, and authorizing hospital discharge planners to recommend specific higher quality PAC providers would further this goal. However, several design decisions would need to be resolved. First, a consistent approach to identifying better quality PAC providers would be needed, and quality standards would need to be transparent for PAC providers and beneficiaries. Second, policies would be needed to safeguard against potential conflicts of interest that could ensue from the authority to recommend specific providers. Finally, the criteria to determine what defines a quality provider would need to account for variations in quality across markets since the number of higher quality providers available in any market will depend on how quality is defined. Regardless of the approach selected to encourage the use of higher quality PAC providers, beneficiaries should retain freedom of choice. Beneficiaries may have important concerns that are not necessarily reflected in standard quality measures, such as language competency or proximity to family members. These preferences may lead them to select a PAC provider that has lower performance on some quality measures, but additional quality information would allow them to better understand the nature of their options and any trade-offs. Encouraging Medicare beneficiaries to use higher quality post-acute care providers 112

135 Medicare’s options for expanding the authority of discharge planners to recommend higher quality PAC providers could include prescriptive approaches that provide specific metrics or definitions that hospitals must use or more flexible approaches that leave key decisions to discharge planners. A hybrid approach could specify certain selection criteria hospitals would need to use while granting hospitals discretion in the application of these criteria. In a flexible approach, hospitals would be responsible for defining the criteria they would use for identifying higher quality PAC providers. Hospitals would select quality measures, collect data from PAC providers or other sources of information, and set the performance levels that PAC providers have to meet. CMS could require that hospitals establish formal vetting processes for setting the criteria and reviewing PAC provider performance to provide some degree of transparency for beneficiaries and PAC providers. This option would allow hospitals to use criteria they believe best meet the needs of their patient populations and reflect the availability of PAC providers in their local markets. However, it could be confusing for beneficiaries and PAC providers in a market area to have different hospitals use different quality definitions. In addition, this option could be administratively complex for CMS to oversee. In a prescriptive approach, CMS would select the quality measures, set the performance levels, identify and notify hospitals and PAC providers, and update the measures as new data became available. Hospitals would be required to notify beneficiaries of the PAC providers that are designated as higher quality. This option would ensure consistent standards of quality and would be less burdensome for hospitals. However, the number of PAC providers designated as high quality would vary across markets. Beneficiaries could find it difficult to select a higher quality provider in areas with limited supply. In a variation of the prescriptive approach, CMS could rate providers on a composite measure that captures various aspects of PAC quality. In each market, discharge planners could highlight the PAC providers that are higher rated and have available capacity. This approach would account for the variation in quality across markets and provide more flexibility to discharge planners. ■ June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 113

136

137 p. 116). For example, under the Hospital Readmissions Introduction Reduction Program (HRRP), the quality of the PAC providers selected by a hospital’s patients could affect While many delivery system reform options highlight the whether the hospital receives a reward or penalty. Other importance of placing patients in the appropriate type of models, such as accountable care organizations (ACOs) post-acute care (PAC)—skilled nursing facility (SNF), and payment bundles that include inpatient hospital care home health agency (HHA), inpatient rehabilitation and PAC, can create even more explicit links between facility (IRF), or long-term care hospital (LTCH)—the hospitals’ financial incentives and the use of higher quality selection of a PAC provider from among particular PAC providers. Because Medicare’s current discharge several of any given type can also be crucial for the planning regulations have not been substantially revised in clinical outcome and expenditures of an episode of care. over 20 years, opportunities exist to update them to better Beneficiaries seeking posthospital care, particularly those serve beneficiaries and advance delivery system reform. patients referred to SNFs and HHAs, frequently have many agencies or nursing facilities operating in their markets. CMS has implemented some initiatives to help Beneficiaries seeking PAC often beneficiaries identify better PAC providers, but these efforts may not be adequate. have many PAC options that vary substantially in quality Encouraging beneficiaries to use higher quality providers is also important because PAC services are costly and Though the supply of PAC providers varies widely across frequently used in traditional fee-for-service (FFS) the country, most beneficiaries have a number of PAC Medicare. In 2015, about 40 percent of hospital discharges providers in their local area. Most areas have at least one, resulted in the use of PAC services, and Medicare if not many, SNFs and HHAs participating in Medicare. spending on PAC totaled about 10 percent of all FFS For example, 86 percent of beneficiaries had five or expenditures—over $60 billion. PAC providers vary in more HHAs operating in their zip code of residence in the quality of care they provide, as we have reported 2016 (Medicare Payment Advisory Commission 2018). annually in our analyses of Medicare payment adequacy The supply of IRFs and LTCHs is more concentrated. (Medicare Payment Advisory Commission 2018). Lower In practice, most hospitals refer to many SNF and HHA quality providers have higher rates of complications such providers. For example, one study found that, in 2008, as rehospitalizations and emergency services use, resulting the average hospital referred patients needing PAC in worse health outcomes for beneficiaries and further services to 23 HHAs and 34 SNFs (Lau et al. 2014). driving up Medicare spending. Policies that encourage Two recent studies have found that readmission rates the selection of higher quality providers could yield generally decrease when a hospital’s PAC discharges are better quality of care and lower Medicare spending and concentrated with a select number of providers, so referring beneficiary cost sharing. to a wider range of providers than necessary may increase readmission rates (Rahman et al. 2013, Schoenfeld et al. Medicare discharge planning regulations place 2016). While factors in addition to supply, such as distance responsibility with hospitals for connecting inpatient acute from a beneficiary’s residence, bed availability, and any hospital patients with their options for PAC, including special clinical needs, can constrain a beneficiary’s options, educating beneficiaries about their choices and providing the substantial supply of providers in many areas indicates referrals when necessary. These regulations are designed that beneficiaries usually have a number of nearby options not only to ease the burden for arranging posthospital in selecting a PAC provider. care for beneficiaries but also to guarantee beneficiary freedom of choice in selecting PAC providers. In fact, Selecting among providers in markets with a robust current regulations do not permit discharge planners to supply is complicated by the variation in quality among recommend specific PAC providers to beneficiaries. PAC providers. For example, the Commission found the following in analyses of PAC providers: Increasing the use of higher quality PAC providers is particularly important as CMS implements value-based • Among SNFs, potentially avoidable rehospitalization payment reforms that hold hospitals accountable for the rates for the first 30 days of a stay averaged 20.2 expenditures and outcomes related to PAC (Table 5-1, percent for the lowest performing quartile of facilities June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 115

138 TABLE Medicare initiatives that place hospitals at financial risk for readmissions from PAC 5–1 Financial incentive to prevent readmissions Initiative Participation Inpatient hospital Mandatory for all Payment determination is in part based VBP incentive that pays hospitals value-based PPS hospitals on a measure of spending in the 30-day bonuses or imposes penalties based on purchasing postdischarge period. their performance program Hospital Mandatory for all The program includes a financial penalty Penalty for hospitals that exceed Readmissions for hospitals with higher than expected PPS hospitals expected rate of readmission for six Reduction readmissions. conditions Program Creates an incentive that holds Comprehensive Mandatory for all Hospitals in the CCJR program can hospitals accountable for cost and Care for Joint hospitals in 67 selected receive a bonus or penalty depending on quality of the inpatient acute care Replacement urban areas (CMS their aggregate spending in the payment services and 90 days of postdischarge intends to reduce to 34 bundle. Lowering readmissions from PAC care for joint replacement patients areas in 2018) helps keep spending below target. Bundled Includes a model that allows hospitals Participants in the BPCI initiative can Voluntary Payments to select a bundle that includes the receive bonus payments if they keep for Care inpatient stay plus PAC and all related spending below a target based on prior Improvement services up to 90 days after discharge; utilization. the beneficiary’s condition must be 1 or more of 48 diagnostic groups Accountable care Hospitals can participate in ACOs with Incentives vary depending on the program. Voluntary organizations other stakeholders to share financial Hospitals that lower readmissions relative (Next Generation risk and collaborate to improve care; to their target will have lower spending or Medicare not all ACOs include a hospital and better quality, which will influence Shared Savings whether they receive penalties or bonuses. Program) Note: PAC (post-acute care), VBP (value-based purchasing), PPS (prospective payment system), CCJR (Comprehensive Care for Joint Replacement), BPCI (Bundled Payments for Care Improvement), ACO (accountable care organization). Source: MedPAC analysis. compared with 8.4 percent for the highest performing range of performance (Medicare Payment Advisory quartile in 2015 (Medicare Payment Advisory Commission 2017). Commission 2017). These examples illustrate the importance of selecting a Among HHAs, rates of hospitalization during or • quality provider since the choice of provider can have within the 30 days after home health care in 2014 implications for the quality of care received. Beneficiaries varied from 17.5 percent for the agency at the 25th served by lower quality providers could experience percentile compared with 30.1 percent for the agency additional hospital stays, have more difficulty recovering at the 75th percentile. from the acute condition that required their hospitalization, and may have adverse long-term health outcomes (e.g., not Among IRFs, the share of patients discharged to a • recovering to a premorbid level of walking or other form SNF in 2015 almost doubled between the providers of physical function). at the 25th percentile and the 75th percentile of the Encouraging Medicare beneficiaries to use higher quality post-acute care providers 116

139 Medicare providers, beneficiaries have a “basic freedom Patients referred to PAC need assistance of choice” to select any PAC provider participating in the to identify better quality providers program (though PAC providers do not have an obligation to accept any patient that is referred). In addition, the Patients selecting an HHA or SNF after a hospitalization Medicare statute defining discharge planning indicates report that they value quality and a provider that is close to that a hospital “may not specify or otherwise limit” the the beneficiary’s residence, but several factors complicate PAC providers made available to beneficiaries. (Medicare the challenge for beneficiaries to make informed choices Advantage allows plans to establish their own networks; (BearingPoint 2003, Sefcik et al. 2016, Shugarman and these plans’ enrollees must select a provider that is in their Brown 2006). Reports of patient experience suggest that plan’s network.) The Balanced Budget Act of 1997 also many beneficiaries who need PAC do not understand the requires that hospitals provide a list of HHAs or SNFs that basic nature of the services, particularly those who have are near the beneficiary’s residence for patients identified no prior experience with posthospital care (BearingPoint as needing these services. The list is not required to 2003, Coleman et al. 2005, Shugarman and Brown 2006). include quality or performance information. In practice, Some patients report being unaware they have a choice many discharge planners are cautious about providing of provider, despite Medicare’s requirements for making advice to beneficiaries because they do not want to be seen them aware of their options (Baier et al. 2015). as limiting patient choice (Baier et al. 2015, Tyler et al. 2017). The hospital stay can be a confusing period when beneficiaries and their families are focused on the patient’s Providing PAC quality information has had acute health problem that led to admission, and they limited success in shifting volume to higher may not recognize, or may be slow to realize, that the quality providers beneficiary will require posthospital care. While provider- Medicare has made provider-level PAC quality measures level quality information is available for beneficiaries, available for PAC providers through components of the some studies suggest that patients are not always aware 1 Medicare.gov website. For each of the four settings, of it and can find the information difficult to understand consumers may search for providers by zip code, and the (Castle et al. 2009, Harris and Beeuwkes-Buntin 2008). In website provides a list of participating providers, quality addition, the decision to discharge a beneficiary can come measures, and other information describing the provider. suddenly. In one study, 30 percent of patients reported The website includes 23 quality measures for SNFs and 21 being discharged with less than a day’s notice (Horwitz quality measures for HHAs. The information is updated et al. 2013). The selection of a PAC provider may need to quarterly. Consumers search the SNF data about 158,000 happen swiftly. With these pressures, it can be challenging, times a month; the HHA data, about 33,000 times a without significant assistance, for many beneficiaries to month. identify the highest quality provider available. The information provided through Medicare.gov—such Medicare’s discharge planning policies are as staffing ratios, quality measures for short-stay patients, intended to facilitate choice and access to compliance survey results, and services offered—can be PAC useful to beneficiaries but also has some limitations for Under Medicare’s conditions of participation (COPs), patients seeking PAC. The measures generally cover broad hospitals are responsible for evaluating their patients’ categories of patients, so there is no ability to examine postdischarge needs, educating beneficiaries about those quality for specific conditions, such as outcomes for a needs, and, if necessary, arranging transfers to the selected facility’s poststroke or other rehabilitation patients. The postdischarge provider. The hospital discharge planner is site also does not identify facilities that provide specialized required to solicit patient preferences for postdischarge treatments such as ventilator care. care and consider the practicability of the patient returning In recent years, Medicare has added a star rating system to home when presenting PAC options. to make the quality reports under Nursing Home Compare Medicare statute and the hospital discharge planning and Home Health Compare easier to interpret. Under COPs are intended to protect beneficiary choice in the this system, Medicare computes a composite measure for selection of PAC providers. As they have with other SNFs and HHAs that summarizes performance on several June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 117

140 individual quality measures. The value of the composite data available through sources like Medicare.gov measure is used to rate providers: The highest scoring (Advisory Board Company 2016, Harris and Beeuwkes- providers receive 5 stars, and the lowest receive 1 star. The Buntin 2008, Sefcik et al. 2016). However, some patients quality measures in the SNF and HHA star rating systems find that physicians vary in their knowledge of the quality include patients receiving PAC, but many of the measures of posthospital care (Burke et al. 2017, Colwell 2017). also pertain to long-term care or community-admitted Hospital discharge planners might be a natural source of patients. Because the rating’s measures are not specific to recommendations since their principal responsibilities the PAC population, their utility for posthospital patients should make them familiar with the PAC options in an may be limited. area. However, Medicare discharge planning rules do not permit them to recommend specific PAC providers. In The evidence suggests that Medicare’s Nursing Home addition, a lack of knowledge about PAC quality may limit Compare and Home Health Compare data have minimal their ability to provide useful information to beneficiaries. impact in motivating beneficiaries to choose higher quality A 2004 survey of discharge planners found that, while 63 providers. Studies have assessed whether patient selection percent of planners were aware of the PAC quality data of HHAs and SNFs changed after Medicare.gov data that Medicare makes available, only 38 percent reported were made available to consumers. One study found that using it (Castle 2009). A more recent analysis found that most SNF patients did not appear to select higher quality discharge planners are not always aware of comparative providers after the Medicare.gov data were released to quality data on PAC providers or do not believe that PAC consumers, while another found that the data had a small providers differ significantly in quality (Baier et al. 2015). impact (an increase of less than 1 percent of a facility’s Discharge planners’ awareness may have increased since volume) when there was a large difference in the quality of 2004, but the survey suggests that a significant share may available providers (Werner et al. 2012, Werner et al. 2011). not use quality data even if they are aware of it. A review of the impact of the HHA data available through Medicare.gov also found minimal impact: On average, the Concern about protecting patient choice reportedly best performing agencies might have increased their market also makes some discharge planners cautious in the share by less than 1 percent (Jung et al. 2016). The lack of assistance they provide, even when patients ask for their impact is consistent with studies of the use of information opinions (Baier et al. 2015). Hospital and health system about quality for consumers in other settings. Reviews of representatives have been concerned that COPs do not the health services literature have found that, while provider adequately define permissible educational activities quality information can be useful for consumers, it has had that respect the beneficiary’s freedom to select a PAC limited or minimal success in getting beneficiaries to select provider (Kahn 2015, Thompson 2016). In practice, this higher quality providers (Goncalves-Bradley et al. 2016, lack of definition means that some discharge planners see Harris and Beeuwkes-Buntin 2008, Hussey et al. 2014). The providing more tailored information, such as highlighting limited impact of these data may indicate that patients are PAC providers that have agreed to collaborate with often unaware of this information or that they have limited the hospital, as part of their assistance responsibilities. or no access to online services when hospitalized. Patients In contrast, others report being unwilling because who are hospitalized may be too distracted or sick to they believe it violates Medicare’s freedom of choice conduct detailed research about their PAC provider options, requirements (Baier et al. 2015, Tyler et al. 2017). For and a beneficiary’s family member or other caregiver may many patients, especially those who lack family contacts also have difficulty finding and using this information. or a physician prepared to advise on PAC, the hesitancy of a discharge planner to provide additional assistance Beneficiaries seek assistance from trusted could be problematic since there may not be other medical intermediaries for selection of a PAC professionals in a better position to help beneficiaries provider consider their options. In practice, beneficiaries report soliciting the views of physicians, family members, or other associates to IMPACT mandates hospitals’ use of quality recommend a PAC provider (Advisory Board Company information, but implementation status is 2016, Harris and Beeuwkes-Buntin 2008, Shugarman unclear and Brown 2006). Beneficiaries generally view this In 2014, the Improving Medicare Post-Acute Care information as more valuable than comparative quality Transformation Act (IMPACT) required changes to the Encouraging Medicare beneficiaries to use higher quality post-acute care providers 118

141 discharge planning COPs to mandate that hospitals “take of these models. Participant hospitals and ACOs have an into account quality, resource use, and other measures incentive to encourage the use of better PAC providers. . . . in the discharge planning process.” CMS proposed In most reform models, CMS has not changed or waived regulations in 2015 to put this mandate into effect but any existing discharge planning requirements, and never finalized the regulation. The proposed rule also hospitals continue to be subject to the current regulations. would have required that beneficiaries referred to IRFs or Hospitals and health systems participating in these LTCHs be given a list of nearby providers, similar to the efforts have indicated that they seek to encourage the use current requirement for SNFs and HHAs. These policies of preferred PAC providers by educating beneficiaries had the potential to strengthen patient choice by explicitly about PAC choices and highlighting the supplemental permitting hospitals to provide and explain quality data services available in their reform model. For example, to beneficiaries during the discharge planning process. in the Bundled Payments for Care Improvement (BPCI) However, the expanded use of quality information did not initiative, hospitals can indicate that they have identified address some concerns about current discharge planning preferred PAC providers with which they collaborate; regulations. Hospital representatives wanted the rule to be beneficiaries selecting one of these providers can receive more explicit that a discharge planner could recommend additional services, such as a transitional care nurse a PAC provider to a beneficiary (Kahn 2015, Thompson that will follow the patient across settings. While some 2016). hospitals report success with encouraging beneficiaries to The proposed regulation would have required hospitals to use preferred providers, no studies have directly assessed share with beneficiaries the cross-sector PAC measures the impact of these efforts (Hargrave et al. 2014). of quality that CMS was required to develop under Another approach to the PAC selection issue is found in 2 IMPACT. Since the measures were not expected to be the Comprehensive Care for Joint Replacement (CCJR) ready before the regulation’s expected implementation, the program. CMS provides hospitals participating in the rule suggested that hospitals use other sources of quality CCJR program with the authority to recommend preferred information such as the data on SNFs and HHAs found PAC providers but leaves the beneficiary’s right to select on Medicare.gov. The regulations implementing IMPACT the PAC provider unchanged. In effect, hospitals can requirements were never finalized, and CMS has offered recommend a provider, but beneficiaries are not obligated no information about future actions on the proposed rule. to use it. While the CCJR program has been active since 2016, no studies of the impact on patient choice of PAC While CMS has made data available to beneficiaries provider have been released. through Medicare.gov, there is no regulatory requirement that hospitals inform patients about these data. If discharge Hospitals have developed preferred PAC planners do not inform beneficiaries, beneficiaries would provider networks to lower readmission have to know about publicly reported measures from their rates own research. Finding and understanding this information The changes in payment policy resulting from the Patient may be challenging for beneficiaries who have been Protection and Affordable Care of 2010 (PPACA) led recently hospitalized or who are unfamiliar with online many hospitals to establish partnerships with PAC information. providers to perform well under the new policies regarding Patient choice under Medicare’s delivery hospital readmission rates. In recognition of these new system reform efforts incentives, hospitals established PAC networks with select providers to strengthen their connections with posthospital CMS has also had to consider how to address beneficiary care. While some hospitals created these networks because choice of PAC in some of its delivery system reform of their participation in programs like BPCI or ACOs, all models. Many of these initiatives are intended to prospective payment system hospitals had an incentive to encourage partnerships or collaboration among providers scrutinize PAC quality because patients readmitted from to improve care, such as encouraging PAC providers these settings could affect their payments under the HRRP and hospitals to coordinate transitional care or quality and hospital value-based purchasing programs. Initial improvement efforts. The high cost of readmissions from efforts were reportedly focused on SNF networks, though posthospital care in many episodes suggests that the some organizations reported developing networks for the quality of PAC providers significantly affects the success June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 119

142 other provider types. These networks are widespread and million beneficiaries were referred to a SNF and about likely to increase in number. A 2016 survey of Premier 2.2 million beneficiaries were referred to an HHA after Health hospitals found that 56 percent had established a hospitalization. To understand the options available a formal or informal PAC network and that 32 percent to these beneficiaries, the Commission compared the were developing a network (Compton-Phillips and Mohta quality of the 5 closest providers within a 15-mile radius 2016). of a beneficiary’s home zip code with the quality of the 3 provider from which the beneficiary received service. To establish a network, hospitals generally release a Each provider within the radius was rated using a solicitation for PAC providers to indicate interest and composite score that included two quality measures: one to collect information about PAC providers’ ability to for adverse events such as hospitalization and a second meet criteria on a variety of metrics. Hospitals are free 4 for improvement in functional ability such as walking. to establish their metrics, which can include quality Over 94 percent of beneficiaries who used HHA services measures, clinical capabilities, performance on licensing had at least one provider within a 15-mile radius that had and accreditation surveys, compliance history, physician a higher quality score than the provider from which they staff affiliation, and geographic coverage in the hospital’s 5 received services (Table 5-2). Similarly, about 84 percent service area. Frequently, a major consideration is the of beneficiaries who used SNF services had at least one volume of patients a PAC provider currently receives from better provider within a 15-mile radius of their residence. a hospital. Focusing the network on higher volume PAC Many beneficiaries lived in an area with multiple options, providers ensures that any quality improvement efforts are though they were disproportionately located in urban targeted to the PAC providers that serve a significant share areas. About 70 percent of beneficiaries who received of a hospital’s patients. These networks are arrangements HHA services had five or more other HHAs that offered between the hospital and the PAC providers, and better quality than their selected provider, while almost beneficiaries are not required to select a PAC provider in half of SNF users had five or more options with better the hospital’s network. Once the networks are established, quality. Beneficiaries who used SNF services and resided the hospital and PAC providers can collaborate on quality in rural areas typically had fewer options: Only 9.9 percent improvement activities such as establishing new clinical had 5 or more SNFs in the 15-mile radius. protocols and case reviews. The magnitude of the quality difference between the Hospitals with preferred networks use voluntary higher performing nearby providers and the provider approaches to promote preferred PAC providers to selected was substantial in many cases. For example, beneficiaries, such as beneficiary education about the for beneficiaries with one better provider nearby, quality of preferred providers or the offer of transitional the geographically closest better SNF had a rate of care nurses that follow patients through their episode of rehospitalization 3 percentage points lower on average. care (Hargrave et al. 2014). For example, one provider The average difference between the selected provider and established an online tool that allowed beneficiaries to the higher quality providers nearby increased with market search the preferred providers by geographic location and size. For example, for beneficiaries with five nearby quality performance. Though some hospitals reported providers with better quality, the average rehospitalization success in encouraging beneficiaries to select preferred rate for the better nearby SNFs was 15 percentage points PAC providers, they also reported that discharge planners lower than the selected hospital’s rate. could be reluctant to highlight network providers because they were concerned about violating patient choice There are some limitations to this analysis. First, the requirements or disrupting current referral patterns analysis does not measure whether SNFs had available (Hargrave et al. 2014). Hospital representatives indicated capacity at the time a beneficiary was discharged from that changing the practices of hospital discharge planners the hospital. Second, CMS does not report data on quality continued to be a challenge. for smaller providers. The absence of data for small providers may be acute for the rates observed in rural areas Beneficiaries who use PAC often have a because these providers tend to have lower patient volume higher quality provider nearby than urban providers. In addition, the rural rates for the availability of SNFs could be affected because critical A review of the referral patterns of Medicare beneficiaries access hospitals are not required to report quality data for that were sent to SNFs and HHAs provides an illustration the swing beds they operate. of current policies and practices. In 2015, about 1.8 Encouraging Medicare beneficiaries to use higher quality post-acute care providers 120

143 TABLE Many beneficiaries had higher quality PAC options nearby, 2015 5–2 Number of higher quality providers available within 15-mile radius 0 5 or (No better 4 options) 1 2 3 more Total Share of beneficiaries with higher quality options nearby: 8.3% Skilled nursing facility patients 14.7% 9.8% 12.2% 100% 46.8% 8.2% 5.9 6.0 5.7 5.5 Home health patients 100 69.5 7.4 Note: PAC (post-acute care). Beneficiary and provider locations were measured using zip code centroids. A provider’s location had to be within 15 miles of the beneficiary’s zip code. Medicare Provider and Review skilled nursing facility file 2015, home health standard analytic file 2015, and Medicare Beneficiary Summary File 2015. Source: meet specific quality levels (e.g., top third nationwide) These results suggest that a significant share of beneficiaries had a nearby HHA or SNF that offered better or give hospitals the authority to flag the best of the PAC providers in their local markets available at discharge. A quality. While several factors such as available capacity, more prescriptive approach would focus attention on PAC clinical needs, or patient preference could affect where providers that are higher overall performers. However, a beneficiary is served, it is also clear that the current if these providers were not available or unable to take hospital discharge planning process can limit efforts to a patient, the advice a discharge planner could provide refer patients to better performing PAC providers. would be limited. Setting a less restrictive policy that allows hospital discharge planners to recommend the higher performing of available providers could address this Principles for improving hospital issue, but the quality of recommended providers could be discharge planning more variable. CMS has developed a significant quantity of measures Helping beneficiaries to identify better quality PAC for its various quality reporting programs (Table 5-3, providers should be a goal in a reformed discharge p. 122). The selection of a subset of these measures planning process, and authorizing hospital discharge that were of shared importance to beneficiaries and the planners to recommend specific PAC providers would program could serve as criteria for identifying better further this goal. However, several design decisions would PAC providers. These measures would need to minimize need to be resolved. First, a clear approach to identifying bias due to shortcomings in risk adjustment or industry better quality PAC providers would be needed, and quality coding practices. Outcome measures that focused on standards would need to be transparent for PAC providers high-cost events would be appropriate, as would more and beneficiaries. Second, policies would be needed to easily verifiable quality measures such as claims-based safeguard against potential conflicts of interest that could measures of rehospitalization or emergency department ensue from the authority to recommend specific providers. use. Other outcomes such as functional gain are important Finally, the criteria to determine what defined a quality but are more difficult to verify because they rely solely provider would need to account for variations in quality on provider assessment practices. In identifying higher across markets because the number of a market’s higher quality providers, CMS should avoid selecting measures quality providers will depend on how quality is defined. that could be vulnerable to manipulation. Finally, a revised CMS would need to consider whether it should limit the policy could allow hospitals to supplement Medicare’s PAC providers a hospital can recommend to those that core measures with other information. Beneficiaries would June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 121

144 TABLE Selected PAC quality measures available through Medicare quality programs 5–3 Examples of measures available Setting Skilled nursing facilities Share of short-stay residents who: • were rehospitalized after a nursing home admission • had an outpatient emergency department visit • were successfully discharged to the community • received antipsychotic medication for the first time Home health agencies Share of patients experiencing: • acute care hospitalizations • emergency department use without hospitalization rehospitalization during the first 30 days of home health care • • emergency department use without hospital readmission during the first 30 days of home health Inpatient rehabilitation facilities All-cause unplanned 30-day post-IRF discharge readmission measure Discharge to community Cross-sector measures (not yet implemented in all sectors) Medicare spending per beneficiary Potentially preventable 30-day postdischarge readmission measure Note: PAC (post-acute care), IRF (inpatient rehabilitation facility). Includes certain measures from Skilled Nursing Home Compare and Home Health Compare websites. IRF and long-term care hospital measures are from Medicare’s quality reporting programs for these settings. Source: Information on Nursing Home Compare, IRF quality reporting measures, and LTCH quality reporting measures from CMS. be free, but not obligated, to weigh both the core measures requirements encourage (Center for Medicare Advocacy and any supplemental information when selecting their 2016, Coalition to Preserve Rehabilitation 2016). PAC provider. Beneficiaries could have concerns that are not necessarily reflected in standard quality measures, such as language Medicare’s five-star rating systems for SNFs and HHAs competency or proximity to family members. Their reflect its current approach to a composite measure of preferences could lead them to select a PAC provider quality for PAC providers, but it would likely need some that has lower performance on some quality measures, modifications to serve as a measure in a revised discharge but additional quality information would allow them to planning policy. Both systems use a number of process understand the nature of their options and any trade-offs. measures, which may give providers a better rating for measures that do not necessarily improve outcomes. The PAC provider capacity, in addition to patient decision- five-star measures do not focus solely on Medicare PAC making, will also affect the ability of any quality patients. Both systems also include measures of functional information to shift beneficiaries to higher quality PAC improvement that can be sensitive to provider coding providers. The supply of higher quality PAC capacity practices. A revised star-rating system that focuses on is finite. Facilities vary in the services they offer, and, post-acute services and claims-based outcomes measures consequently, beneficiaries requiring specialized or higher would address these shortcomings. cost services may have even fewer options. These factors can limit the ability to shift beneficiaries to PAC providers Beneficiaries must retain their freedom to choose a PAC with higher quality. Optimally, any additional authority for provider under a revised discharge planning process. hospital discharge planners would allow them to identify, Beneficiary preferences would be incorporated in when possible, the higher performing PAC providers the options a discharge planner presented, as current among those with available capacity at discharge. Encouraging Medicare beneficiaries to use higher quality post-acute care providers 122

145 Additional assistance selecting providers could be even Approaches for identifying higher more important if CMS implements a unified payment quality PAC providers system for PAC. Under such a system, providers could have the option to consolidate separate PAC operations Medicare’s options for helping hospitals select appropriate into a single PAC facility. Quality metrics could be used to PAC providers at the point of patient discharge range explain the clinical services and goals of care a patient can from flexible (leaving key decisions about selecting expect from particular PAC providers. Improved quality beneficiaries’ PAC providers to hospital discharge information about the new category of providers, along planners) to prescriptive (setting specific metrics or other with the discharge planner’s ability to highlight the better criteria that define a PAC provider as high quality and performing ones, would make it easier for beneficiaries to limiting a hospital’s selection of PAC providers to those choose among the options in a PAC PPS. meeting this definition) (Table 5-4, p. 124). A hybrid Improving discharge planning should also complement approach could specify certain quality criteria hospitals other efforts to improve value in Medicare. Hospitals have must use while granting hospitals discretion in the use of a financial incentive to encourage beneficiaries to use the these criteria. Table 5-4 illustrates two hypothetical policy PAC providers with which they collaborate under payment options, one more flexible, the other more prescriptive. reforms such as ACOs and bundling programs. However, if the new authority limited the PAC provider options to Illustrative example of a flexible approach only those that met the Medicare-selected quality metrics, Under a flexible approach, hospitals would be responsible hospitals could find that some of their referral partners for defining the criteria they would use to identify higher were not highly rated under these terms. In these instances, quality PAC providers. A hospital would be responsible hospitals would have to weigh how to respond. They could for selecting quality measures, collecting data from PAC encourage these providers to improve quality, provide providers, and setting the performance levels that PAC supplemental information to beneficiaries that emphasizes providers would have to meet to be recommended by the these providers’ other merits (such as meeting other hospital. CMS could require that hospitals establish formal facets of quality not measured by Medicare or providing vetting processes for setting the criteria and reviewing supplemental services like transitional care nurses), or opt PAC provider performance to provide some degree to collaborate with different PAC providers. of transparency for beneficiaries and PAC providers. Hospitals could be required to make their criteria and Developing quality measures that capture the full gamut selection process available for public review. of beneficiaries’ preferences could be challenging. Medicare already has many clinical quality measures, but The advantage of this approach is that it provides hospitals beneficiaries may have other preferences such as facility with the freedom to establish the criteria that they believe condition, staff cultural or linguistic competencies, and best reflect the needs of their patients and to tailor facility amenities such as dining and recreation options. those criteria to the available supply of providers. Some Developing these additional indicators would dilute a hospitals have conducted similar processes to identify focus on clinical outcomes, and, in some cases, it could PAC referral partners for ACOs and bundled payment be impractical or impossible to develop useful measures initiatives, for instance. Metrics could be set to identify for preferences that are more subjective (e.g., facility the best of the local PAC providers, regardless of how they décor or staff demeanor). A more practical approach compared with national levels. As many programs make could be for CMS to focus on a core set of measures hospitals accountable for readmissions, hospitals would that focus on outcomes that matter for the beneficiary have a significant incentive to work with higher quality and the program and allow hospitals to supplement these providers. measures with other information when they deem it relevant to beneficiary preferences. As mentioned earlier, Flexibility would permit hospitals to select the quality many beneficiaries want hospital discharge planners measures they deem appropriate and could include or other clinicians to recommend a facility. Such a compliance history and selected quality measures. If recommendation should respect patient preferences, and some measures did not adequately control for differences a revised discharge planning policy should not overload in patient mix, hospitals could also opt to use judgments beneficiaries with more information than they can process of a PAC provider’s clinical reputation among hospital during an acute health crisis. medical staff. On the one hand, flexibility could permit June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 123

146 TABLE Illustrative examples of policies for revising discharge planning 5–4 Option 2: Medicare sets standards to define higher Option 1: quality PAC providers Hospitals have flexibility to write own standards • Medicare COPs require hospitals to define criteria. • Medicare designates providers that can be Medicare’s role recommended (e.g., must be at least three or four ® score). stars, better CAHPS • Medicare sets hospitals’ selection criteria. Hospitals select measures, allowing for innovation and Use of quality • measures experimentation. Likely, the same safeguards stated in Option 1 • There would need to be: Regulatory would be needed, but standards for recommending safeguards • safeguards to prevent financial conflicts of interest; PAC providers would be clearer. • disclosure of conflict of interest/ownership/ collaboration; and CMS approval of individual hospitals’ criteria and • monitoring of proper application. Beneficiary • Beneficiaries would receive recommendations that • A single set of standards across hospitals would reflect quality of PAC care in the market. implications make reasoning behind selected PAC providers • It could be confusing to have multiple definitions across more transparent to beneficiaries. hospitals. The quality of PAC providers selected would be • more consistent. • PAC provider Providers would have to consider multiple definitions if • A single set of standards would result in consistent working with many hospitals, potentially with different implications designation. measures for each setting. • There would be consistency across markets as to • Designation as a higher quality provider could vary which providers qualify as higher quality. among hospitals and across geographic markets. Flexibility in the definition of quality would allow • Advantages • A single definition of “quality” would provide clear hospitals to develop patient-centered definitions and standards for PAC providers, consistent treatment require them to scrutinize referral partners. under policy. • Approaches could reflect local PAC markets’ capacity • The implementation burden on hospitals would be and scope of offerings. lighter. Enforcement would be less complex. CMS would • need to ensure that hospitals observe sanctioned criteria when recommending PAC providers. Disadvantages There would be a greater burden on hospitals to • • If there were a single standard, the number of implement and maintain standards and on CMS to designated providers would vary across areas. verify and audit standards and their application. Multiple definitions of higher quality providers could be • confusing for beneficiaries and PAC providers. ® PAC (post-acute care), COP (condition of participation), CAHPS® (Consumer Assessment of Healthcare Providers and Systems Note: ). Source: MedPAC analysis. the development of more patient-centered standards based hospitals would have the burden of developing criteria for on a hospital’s clinical expertise. On the other hand, the identifying higher quality PAC providers. quality of PAC providers selected and recommended Both beneficiaries and PAC providers could find this to beneficiaries could vary as a result. In addition, policy confusing since there would be no consistent Encouraging Medicare beneficiaries to use higher quality post-acute care providers 124

147 standards for designating a provider as higher quality. have the same vulnerabilities to fraud, waste, and abuse PAC providers would be subject to different definitions that are present in the flexible approach. of quality among hospitals and could find it difficult to The quality measures available vary among PAC settings, satisfy the multiple and potentially conflicting definitions. but CMS could, in most cases, start with measures of A single PAC provider could have different quality efficiency and quality that are used in the pay-for-reporting designations among the hospitals in the PAC provider’s and value-based purchasing programs for PAC providers. market, qualifying as a higher quality provider with some CMS might focus on hospital readmissions, discharge to hospitals but not others. Medicare has been moving in the community, and other measures that reflect high-cost and opposite direction, toward efforts to develop standardized high-consequence events. CMS is developing cross-sector cross-sector measures of PAC quality that facilitate measures of PAC quality, including readmissions, and comparisons; the use of unique measures by hospitals these measures could be used when they become available. could increase the reporting burden on PAC providers. CMS would have to consider how to set the performance Another disadvantage of this more flexible approach is levels to qualify as a higher quality PAC provider, such that it would be more challenging for CMS to oversee. as setting a benchmark for rehospitalization from a SNF Ensuring that hospitals were not creating inappropriate or HHA to be specified as higher performing. Setting a business or financial relationships that encouraged undue single national benchmark would have the advantage of favoritism or inappropriate PAC volume would require simplicity and consistency, but because the quality of some oversight by CMS. Ensuring that collaboration PAC providers varies across regions, some regions would among hospitals and PAC providers is aimed at improving have more providers that qualified for selection and other outcomes and not cooperating in ways that inefficiently regions would have fewer. increase Medicare spending would be important. A broad range of permissible policies would make it challenging to For example, a national benchmark could be set defining identify when a hospital’s practices created unacceptable higher quality SNFs as those in the bottom third (lowest) risk for fraud, waste, and abuse. CMS might find it on rehospitalization rates. With this benchmark, 114 core- difficult to conduct a uniform and efficient review process based statistical areas (CBSAs) would have only 1 or 2 if each hospital followed a unique approach. SNFs that qualified as higher quality, while 39 CBSAs would have 20 or more SNFs that qualified. A lower Illustrative example of a prescriptive performance benchmark (i.e., a higher rate of readmissions approach as the criteria) could be specified that would increase Under a more prescriptive approach, CMS could establish supply in some markets, but doing so would degrade the quality metrics for designating PAC providers as higher acceptable level of quality in all markets nationwide, even quality. Under this approach, CMS would select the in areas that did not need more providers. measures, set the performance levels, identify and notify Alternatively, a prescriptive approach could establish a hospitals and PAC providers, and update the measures as definition that uses both national and local standards. For new data became available. Hospitals would be required example, the definition could be a two-step test: the first to notify beneficiaries of the PAC providers designated as would designate providers that are in the lowest third of higher quality. the nationwide distribution for readmission rates, and the Establishing a single standard would make the program second would qualify any providers in the lowest third easier for beneficiaries and PAC providers to understand. relative to other providers in their local market area. This Beneficiaries would likely better understand why the combination approach could result in a more even supply recommended providers were selected, which might make of designated higher quality providers across markets them more inclined to use higher quality PAC providers. but would result in designations that varied from region There would be more consistency in the quality of care to region. For example, across urban areas, the average available to beneficiaries from designated providers rate of readmissions for SNFs varied in 2014 from 11 6 because the standards applied by Medicare would be percent to 21 percent. Even if beneficiaries used only identical across markets. The administrative burden on providers deemed “high quality” in their areas, the quality hospitals would be lower relative to the more flexible of care received would vary across markets. Further, option, though CMS would have more responsibility. PAC providers with the same level of performance could Since the standards are set by CMS, this approach does not receive different designations depending on their market. June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 125

148 In a variation of this option, CMS could rate providers require that PAC providers achieve certain performance on a composite measure that captured different aspects of levels (e.g., top third of providers) on selected measures to be designated as a higher quality PAC provider. If PAC quality. Within each market, discharge planners could policymakers favored the more prescriptive approach, highlight the PAC providers that are more highly rated and have available capacity. This approach would account for CMS could provide a standardized definition that includes the variation in quality across markets and provide more a quality rating of PAC in a market. The hospital could observe how the supplemental data revised the rating of flexibility to discharge planners. PAC providers, with the better PAC providers receiving Another approach would be for CMS to create a core the designation as higher performing. Determining the set of metrics but permit hospitals to supplement this appropriate balance would benefit from experimentation, information with their own measures. Medicare’s and CMS could pilot policies that varied the degree of measures could reflect outcomes important to patients and flexibility and regulatory specificity—for example, by the program, such as rates of readmission and discharge geographic region. to the community, and CMS could require that this information be reported to beneficiaries. Hospitals could have the option to include additional information they Conclusion also deem important, and discharge planners could be charged with helping beneficiaries understand the different Medicare policy currently places a premium on protecting indicators. beneficiary choice of PAC provider, but it does not encourage beneficiaries to use higher quality PAC Hybrid approaches combining elements of flexible and prescriptive frameworks providers. Any new policy should seek to ease or simplify the burden on beneficiaries, many of whom already report Policymakers could combine elements of the two that discharge planning can be a difficult and confusing approaches to balance or mitigate the disadvantages of period. Efforts to provide additional information should each approach. For example, policymakers could begin not overwhelm beneficiaries and should ensure that patient with the flexible framework but require hospitals to preferences for PAC are recognized. ■ select quality measures that meet certain standards or are already in use in the program. Alternatively, Medicare could leave the exact measures open for determination but Encouraging Medicare beneficiaries to use higher quality post-acute care providers 126

149 Endnotes 4 The measures for skilled nursing facilities included all- 1 Medicare provides information through Nursing Home cause readmissions during the SNF stay and improvement in Compare, Home Health Compare, Inpatient Rehabilitation mobility; the HHA measures included hospitalization during Facility Compare, and the Long-Term Care Hospital Compare the HHA stay and improvement in walking at discharge. websites available at Medicare.gov. Providers within a 15-mile radius of the beneficiary were 2 IMPACT requires CMS to develop quality measures rated from high to low on these measures, with the two for resource use, hospital readmission, and discharge to measures weighted evenly. community for PAC providers. We included only providers with a complete set of quality 5 3 The measure of distance was based on zip codes. For each measures data in this analysis. beneficiary, we identified the zip codes with a geographic 6 This finding pertains to core-based statistical areas with 10 center within 15 miles of the center of the beneficiary’s or more SNFs that had adequate data for computation of the residential zip code. The five closest providers were identified readmission rate. and rated based on the quality measures. June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 127

150 References Goncalves-Bradley, D. C., N. A. Lannin, L. M. Clemson, et al. Advisory Board Company. 2016. What consumers want: Cochrane Database 2016. Discharge planning from hospital. Understanding post-acute patients and families. Infographic. Systematic Review , no. 1 (Jan 27): CD000313. https://www.advisory.com/research/post-acute-care-collaborative/ members/multimedia/infographics/2016/what-consumers-want- Private sector Hargrave, E., K. Quirk, and D. Weinstein. 2014. from-post-acute-care. . A study conducted by approaches to managing post-acute care staff from NORC at the University of Chicago and Georgetown Baier, R. R., A. Wysocki, S. Gravenstein, et al. 2015. A qualitative University for the Medicare Payment Advisory Commission. study of choosing home health care after hospitalization: The Washington, DC: MedPAC. unintended consequences of ‘patient choice’ requirements. 30, no. 5 (May): 634–640. Journal of General Internal Medicine Harris, K., and M. Beeuwkes-Buntin. 2008. Choosing a health care provider . Synthesis report. Princeton, NJ: Robert Wood BearingPoint. 2003. Consumer testing of Home Health Johnson Foundation. Compare website prototype. Report prepared for the Centers for Medicare & Medicaid Services by BearingPoint. McLean, VA: Horwitz, L. I., J. P. Moriarty, C. Chen, et al. 2013. Quality of BearingPoint. discharge practices and patient understanding at an academic 173, no. 18 (October medical center. JAMA Internal Medicine Burke, R. E., E. Cumbler, E. A. Coleman, et al. 2017. Post-acute 14): 1715–1722. care reform: Implications and opportunities for hospitalists. 12, no. 1 (January): 46–51. Journal of Hospital Medicine Hussey, P. S., H. S. Luft, and P. McNamara. 2014. Public reporting of provider performance at a crossroads in the United Castle, N. G. 2009. The Nursing Home Compare report card: States: Summary of current barriers and recommendations on Consumers’ use and understanding. Journal of Aging & Social how to move forward. Medical Care Research and Review 71, no. Policy 21, no. 2 (April–June): 187–208. 5 supplement (October): 5S–16S. Castle, N. G., J. Engberg, J. Lave, et al. 2009. Factors associated Jung, J. K., B. Wu, H. Kim, et al. 2016. The effect of publicized with increasing nursing home closures. Health Services Research quality information on home health agency choice. Medical Care 44, no. 3 (June): 1088–1109. 73, no. 6 (December): 703–723. Research and Review Center for Medicare Advocacy. 2016. Letter to CMS regarding Kahn, C. 2015. Comment letter by the Federation of American the proposed rule entitled, “Medicare and Medicaid Programs; Hospitals on CMS’s proposed rule entitled, “Medicare Program; Revisions to Requirements for Discharge Planning for Hospitals, Revisions to Requirements for Discharge Planning for Hospitals, Critical Access Hospitals, and Home Health Agencies,” January 4. Critical Access Hospitals, and Home Health Agencies” (CMS Coalition to Preserve Rehabilitation. 2016. Letter to CMS 3317–P). December 21. regarding the proposed rule entitled, “Medicare and Medicaid Lau, C., A. Alpert, P. Huckfeldt, et al. 2014. Post-acute referral Programs; Revisions to Requirements for Discharge Planning for patterns for hospitals and implications for bundled payment Hospitals, Critical Access Hospitals, and Home Health Agencies,” initiatives. 2, no. 3 (September): 190–195. Healthcare January 4. Report to the Medicare Payment Advisory Commission. 2018. Coleman, E. A., E. Mahoney, and C. Parry. 2005. Assessing the . Washington, DC: MedPAC. Congress: Medicare payment policy quality of preparation for posthospital care from the patient’s 43, no. 3 perspective: The care transitions measure. Medical Care Report to the Medicare Payment Advisory Commission. 2017. (March): 246–255. Congress: Medicare payment policy . Washington, DC: MedPAC. ACP Hospitalist Colwell, J. 2017. Perfecting post-acute care. , April. Rahman, M., A. D. Foster, D. C. Grabowski, et al. 2013. Effect https://acphospitalist.org/archives/2016/04/post-acute-care.htm. of hospital-SNF referral linkages on rehospitalization. Health (October 17): 1–22. Services Research Compton-Phillips, A., and N. Mohta. 2016. Strengthening the . http://catalyst. Post-Acute Care Connection. NEJM Catalyst Schoenfeld, A. J., X. Zhang, D. C. Grabowski, et al. 2016. nejm.org/strengthening-post-acute-care-connection/. Hospital-skilled nursing facility referral linkage reduces readmission rates among Medicare patients receiving major surgery. Surgery 159, no. 5 (May): 1461–1468. Encouraging Medicare beneficiaries to use higher quality post-acute care providers 128

151 Tyler, D. A., E. A. Gadbois, J. P. McHugh, et al. 2017. Patients Sefcik, J. S., R. H. Nock, E. J. Flores, et al. 2016. Patient are not given quality-of-care data about skilled nursing facilities preferences for information on post-acute care services. Research when discharged from hospitals. 36, no. 8 (August Health Affairs in Gerontological Nursing 9, no. 4 (July 1): 175–182. 1): 1385–1391. Shugarman, L., and J. Brown. 2006. Nursing home selection: Werner, R. M., R. T. Konetzka, E. A. Stuart, et al. 2011. Changes How do consumers choose? Volume I: Findings from focus groups in patient sorting to nursing homes under public reporting: . Report prepared of consumers and information intermediaries Health Services Improved patient matching or provider gaming? for the Department of Health and Human Services by the RAND 46, no. 2 (April): 555–571. Research Corporation. Washington, DC: Department of Health and Human Services. Werner, R. M., E. C. Norton, R. T. Konetzka, et al. 2012. Do consumers respond to publicly reported quality information? Thompson, A. 2016. Comment letter by the American Hospital 31, Evidence from nursing homes. Journal of Health Economics Association on CMS’s proposed rule entitled, “Medicare no. 1 (January): 50–61. Program; Revisions to Requirements for Discharge Planning for Hospitals, Critical Access Hospitals, and Home Health Agencies” (CMS 3317–P). December 21. June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 129

152

153 CHAPTER 6 Issues in Medicare’s medical device payment policies

154

155 CHAPTER 6 Issues in Medicare’s medical device payment policies In this chapter Chapter summary This chapter explores two distinct topics related to medical devices. First, we Introduction • explore ways to improve Medicare’s payment policies for durable medical • DMEPOS background equipment, prosthetic devices, prosthetics, orthotics, and supplies (DMEPOS). Second, we explore ways to constrain the risks posed by physician-owned Non-CBP DMEPOS • distributors (PODs) and to make them more transparent to beneficiaries, products enforcement agencies, and others. Policy options to improve • Medicare’s DMEPOS payment policies the accuracy of Medicare’s payment rates for non-CBP Medicare beneficiaries rely on DMEPOS products to treat their illness or DMEPOS products and injury and to allow them to remain in their homes, as opposed to seeking care protect beneficiaries in an institutional setting. DMEPOS as a category comprises a large number of products that vary in cost and complexity, ranging from complex power • Physician-owned distributors wheelchairs to diabetes testing supplies to knee braces. • Conclusion Pursuant to a statutory requirement, CMS implemented the DMEPOS Competitive Bidding Program (CBP) to use market competition to set payment rates and limit fraud and abuse while ensuring beneficiaries retain access to needed DMEPOS products. The CBP began in 2011 in nine large urban areas and was focused on the highest cost and highest volume items with the largest potential for savings. Over time, the CBP has added products and expanded geographically. As of 2016, Medicare’s payment rates for DMEPOS products included in the CBP are set either directly through bidding June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 133

156 or indirectly by administratively setting prices at least partially based on CBP information in areas where the CBP has not been implemented (e.g., rural areas). The CBP has successfully driven down the cost of DMEPOS products for Medicare and beneficiaries. Compared with payment rates in the year before the CBP, Medicare’s payment rates for some of the highest expenditure DMEPOS products have fallen by an average of roughly 50 percent. CMS initially estimated that the CBP would save over $42 billion in the first 10 years of the program—$25 billion in savings for the program and $17 billion in savings for beneficiaries. At the same time, Medicare expenditures for DMEPOS products excluded from the CBP have continued to grow. By 2015, nearly half of all Medicare expenditures on DMEPOS products were for products excluded from the CBP. Medicare pays for these products using a fee schedule that is largely based on supplier charges from 1986 to 1987 (updated for inflation) and undiscounted list prices. Medicare’s payment rates for the top 10 non-CBP DMEPOS products in 2015 were a third higher, on average, than private-payer rates for comparable products, and some non-CBP DMEPOS products continue to generate high rates of improper payments, experience high utilization growth, and exhibit patterns of potential fraud and abuse. To address these issues, some additional products that are not currently competitively bid could be moved into the CBP. We also observe that the participation and balance billing rules for DMEPOS products and suppliers could be strengthened to better protect beneficiaries and to better align those policies with many other Part B services. Physician-owned distributors PODs are entities that derive revenue from selling, or arranging for the sale of, devices ordered by their physician-owners for use in procedures the physician- owners perform on their own patients. PODs have the ability to distort the supply chain for medical devices—potentially resulting in an increase in the volume of surgeries performed on beneficiaries, higher costs for hospitals and the Medicare program, and inappropriate care. The Commission questions the value PODs produce for the Medicare program and beneficiaries. We suggest several ways in which Medicare and policymakers can constrain the risks posed by PODs. We discuss two specific options to revise the Stark law, which is intended to prohibit physicians from referring Medicare beneficiaries to certain health care facilities in which they have a financial interest, and several key topics for policymakers to consider if such changes are made. While the options would likely limit the use of PODs, some PODs might continue to operate even if the Stark law was modified. In addition, the Commission supports Issues in Medicare’s medical device payment policies 134

157 increasing the transparency of POD-physician relationships by requiring all PODs to report under the Open Payments program, a program designed to shed light on financial ties between physicians and certain industries. ■ Report to the Congress: Medicare and the Health Care Delivery System | June 2018 135

158

159 DMEPOS spending overview Introduction Medicare sets the payment rates for many DMEPOS products through the CBP. Products excluded from the Medicare beneficiaries rely on durable medical equipment, CBP are primarily paid on a fee schedule basis. The trends prosthetic devices, prosthetics, orthotics, and supplies in Medicare spending for these two broad categories of (DMEPOS) to treat their illness or injury and to allow products substantially diverged over the last several years. them to remain in their homes, as opposed to seeking care in an institutional setting. This chapter provides an Medicare expenditures on DMEPOS products included overview of Medicare’s Competitive Bidding Program in the CBP have decreased considerably over time. (CBP) for DMEPOS products and of Medicare’s payment From 2010 to 2015, Medicare expenditures for products methods for DMEPOS products that are excluded from included in the CBP fell from $7.5 billion to $4.4 billion, the CBP. The chapter describes payment policy changes 1 a decrease of 42 percent. Expenditures for certain types that could be made to improve the accuracy of Medicare’s of products in the CBP declined even faster. For example, payments for DMEPOS products, to protect beneficiaries, between 2010 and 2015, Medicare expenditures on and to enhance program integrity. diabetes testing supplies (e.g., blood glucose test strips) fell from $1.6 billion to $0.3 billion, a decrease of 79 This chapter also includes a discussion of issues percent (Table 6-1, p. 138). surrounding physician-owned distributors (PODs), which allow physicians to profit from the sale of medical devices Over the same time period, Medicare expenditures on they use. PODs, which have historically been concentrated DMEPOS products not included in the CBP continued in the market for implantable medical devices, create an to increase. Between 2010 and 2015, expenditures for incentive for physicians to base their decisions, such as these products grew from $3.3 billion to $4.0 billion, a whether to operate on a patient and which instrumentation 2 total increase of 23 percent. Because of the decrease in to use, on financial rather than clinical considerations. To spending on CBP products and the increase in spending on better protect beneficiaries and the Medicare program, this non-CBP products, the share of total Medicare DMEPOS chapter discusses revisions to the Stark law to limit the use spending attributable to non-CBP products has increased of PODs. rapidly. In 2010, non-CBP products represented about 30 percent of Medicare DMEPOS spending; by 2015, non- CBP products accounted for nearly half (48 percent) of all spending. DMEPOS background At the beginning of the program, CMS expected the CBP’s DMEPOS, as a category, comprises a wide range of overall savings to Medicare and beneficiaries to be more products. Durable medical equipment (DME) comprises than $42 billion over the first 10 years. This estimate products that serve a medical purpose, can withstand included $25 billion in savings for the Medicare program repeated use, are generally not useful in the absence of and $17 billion in savings for beneficiaries, as a result of an illness or injury, and are appropriate for use in the lower coinsurance payments and the downward effect on home (e.g., wheelchairs). Supplies that are necessary premiums (Centers for Medicare & Medicaid Services for the effective use of DME are also covered under the 2012). DME benefit (e.g., oxygen in oxygen tanks). Prosthetic devices replace all or part of an internal body organ or History of DMEPOS payment methods function (e.g., colostomy bags and parenteral and enteral Before implementing the CBP in 2011, CMS paid for nutrition). Prosthetics include artificial legs, arms, and nearly all DMEPOS products on a fee schedule basis. Fee eyes. Orthotic devices are defined as providing rigid or schedule payment rates were largely based on supplier semi-rigid support for weak or deformed body parts or charges from July 1986 through June 1987 and on restricting or eliminating motion in a diseased or injured information such as unadjusted list prices for products part of the body (e.g., leg, arm, back, and neck braces). 3 introduced after this time period. Before 2011, annual Other DMEPOS items include surgical dressings and payment rate adjustments were generally between zero therapeutic shoes and inserts for beneficiaries with percent and the consumer price index for all urban diabetes. consumers (CPI–U). Since 2011, payment rates have June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 137

160 TABLE Medicare expenditures on CBP products fell while expenditures 6–1 on non-CBP products increased, 2010–2015 Total Medicare expenditures (in billions of dollars) 2010 Percent change 2015 CBP products (total) $7.5 $4.4 –42% DMEPOS other than diabetes testing supplies 5.9 –31 4.0 –79 0.3 1.6 Diabetes testing supplies 23 4.0 3.3 Non-CBP products Note: CBP (Competitive Bidding Program), DMEPOS (durable medical equipment, prosthetic devices, prosthetics, orthotics, and supplies). Figures in table are rounded and include beneficiary spending. If a product was included in any CBP round through 2017, it is included in the CBP product categories in both 2010 and 2015. The totals for CBP products include spending in both competitive bidding areas and non–competitive bidding areas. MedPAC analysis of 2010 and 2015 Physician/Supplier Procedure Summary file and Competitive Bidding Implementation Contractor’s Healthcare Common Source: Procedure Coding System code lists. annually been increased by the CPI–U, reduced by the 2010 and 2011 claims for diabetes testing supplies, OIG change in economy-wide productivity (Social Security Act found that $425 million in Medicare-allowed claims had Section 1834 (a)(14)(L)). Historically, fee schedule rates characteristics of questionable billing, such as claims were not updated to reflect technological improvements, billed by suppliers who had an unusually high share of such as efficiency gains in manufacturing, or changes in beneficiaries who received their diabetic testing supplies market conditions. at perfectly regular intervals (which suggests suppliers automatically provided refills as opposed to beneficiaries As a result of setting payment rates based on supplier specifically requesting refills, which is required by charges and largely updating payment rates for inflation Medicare) (Office of Inspector General 2013a). In another over time, many DMEPOS products had become instance, OIG found that 80 percent of claims for power substantially overpriced before the CBP. The Government wheelchairs supplied to beneficiaries in the first half of Accountability Office (GAO) and the Department of 2007 did not meet Medicare requirements (Office of Health and Human Services (HHS) Office of Inspector Inspector General 2011). General (OIG) published numerous reports detailing products for which Medicare’s DMEPOS payment rates The Balanced Budget Act of 1997 instructed the Secretary were higher, often by significant amounts, compared of HHS to conduct a competitive bidding demonstration with what suppliers paid to purchase products from for DMEPOS. CMS conducted demonstrations in Polk manufacturers, what suppliers paid to purchase products County, FL (1999 to 2002), and San Antonio, TX (2000 from wholesalers, list prices on suppliers’ websites, to 2002), that collectively reduced Medicare expenditures payment rates of private payers, and payment rates for the subject DMEPOS products by 19 percent, or of other government purchasers (Office of Inspector $9.4 million—$7.5 million in savings for the Medicare General 2009, Office of Inspector General 2005, Office program and $1.9 million in savings for beneficiaries. The of Inspector General 2004, Government Accountability demonstrations had little overall impact on beneficiary Office 1997). For example, based on the 2006 median access (Karon et al. 2003). Medicare fee schedule amount, a 2006 OIG report found After the successful demonstrations, the Medicare that Medicare paid $7,215 for 36 months’ rental of oxygen Prescription Drug, Improvement, and Modernization concentrators that cost $587, on average, to purchase Act of 2003 (MMA) required the Secretary to establish (Office of Inspector General 2006). competitive bidding for certain DMEPOS products. The MMA also expressly prohibited certain DMEPOS Excessively high payment rates increased expenditures and likely encouraged inappropriate utilization. After analyzing Issues in Medicare’s medical device payment policies 138

161 products such as Class III devices from being included in by the Patient Protection and Affordable Care Act of 2010, 4 competitive bidding. CMS began in 2016 to use pricing information from the The law required CMS to implement CBP to adjust the fee schedule payment rates in non-CBAs the CBP in 10 of the largest metropolitan statistical areas for DMEPOS items included in the CBP. DMEPOS items (MSAs) initially and expand to additional areas thereafter. that are not included in the CBP, regardless of whether The law also gave the Secretary the authority to phase in a beneficiary lives in a CBA or non-CBA, are still paid competitive bidding among the highest cost and highest largely on a fee schedule basis. volume items or those with the largest savings potential. Suppliers who furnish DMEPOS products included in CMS implemented CBP Round 1 in 2008, but the the CBP must accept assignment (42 CFR § 414.408 Medicare Improvements for Patients and Providers Act (c)). For DMEPOS products not included in the CBP and of 2008 (MIPPA) canceled all the contracts two weeks CBP products used by beneficiaries who live outside a after the program began and instructed CMS to rebid the CBA, assignment is generally not mandatory. As a result, round. Because the CBP was expected to produce savings DMEPOS suppliers do not have to accept Medicare’s for Medicare and beneficiaries, the DMEPOS industry fee schedule rate as payment in full and may balance bill agreed to a 9.5 percent payment reduction for all items beneficiaries (i.e., bill beneficiaries for the difference that were to be included in the CBP in exchange for between the fee schedule rate and what the supplier delaying the CBP. decides to charge for a given product). In contrast to other In 2011, CMS implemented CBP Round 1 rebid for Part B services, there is currently no limit on balance nine product categories in nine MSAs, referred to billing for DMEPOS products. For example, physicians as competitive bidding areas (CBAs). This round may balance bill only up to 115 percent of the allowed of the CBP was referred to as a “rebid” because it amount under the physician fee schedule. largely covered the same areas and products as the 5 Further, Medicare’s current payment policies do not original Round 1 that was canceled by MIPPA. Since encourage DMEPOS suppliers to enroll as participating 2011, CMS has conducted two additional rounds of suppliers. Participating suppliers accept assignment on all competitions (i.e., “recompetes”) in the same nine Round Medicare claims during the year, whereas nonparticipating 1 MSAs. These rounds are referred to as “Round 1 suppliers are able to accept or reject assignment on a recompete” and “Round 1 2017.” As required by statute, claim-by-claim basis. Under the physician fee schedule, CMS also conducted competitions in 90 additional Medicare reduces the allowed amount to 95 percent of MSAs beginning in July 2013, referred to as “Round 2” the fee schedule rate for all nonparticipating providers, and “Round 2 recompete.” Finally, CMS implemented even if a particular claim is paid on an assignment the National Mail-Order Program for diabetes testing 6 basis. In contrast, no such payment reduction exists for supplies (e.g., blood glucose test strips) in July 2013. nonparticipating DMEPOS suppliers. As the name implies, this competition covers the entire country, including both urban and rural areas, but applies CBP structure only to diabetes testing supplies purchased on a mail- order basis (which include supplies shipped or delivered Suppliers are required to meet certain eligibility to a beneficiary’s home, regardless of the method of requirements to be considered for a contract under the delivery). As of 2018, two CBP rounds are active (Round CBP. For example, eligible suppliers are required to: 1 2017 and Round 2 recompete) that together operate in be enrolled in Medicare and in good standing; • 99 large MSAs, and the National Mail-Order Program recompete for diabetes testing supplies is also active be accredited by a CMS-approved accrediting • (Figure 6-1, p. 140). organization; CMS also uses pricing information from the CBP to adjust • meet applicable state licensing requirements; and fee schedule payment rates for areas and channels not directly covered by the CBP. Pursuant to the American submit certain financial documents, including • Taxpayer Relief Act of 2012, CMS sets the payment rates the suppliers’ most recent tax return, financial for non-mail-order diabetes testing supplies equal to the statements, and credit report (Competitive Bidding payment rate determined through the National Mail-Order Implementation Contractor 2014b). Program beginning July 2013. Additionally, as required June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 139

162 FIGURE XXXX X-X FIGURE Time line of DMEPOS Competitive Bidding Program rounds, 2008–2018 6–1 Round 1 Round 1 recompete Round 1 rebid Round 1 2017 • January 2014 to • January 2017 to • January 2011 to • July 1–15, 2008 December 2018 December 2013 December 2016 • 9 MSAs • 9 MSAs • 9 MSAs • 7 product categories • 6 product categories • 9 product categories Round 2 Round 2 recompete • July 2013 to June 2016 • July 2016 to December 2018 • 90 MSAs • 91 MSAs • 8 product categories • 7 product categories National Mail-Order Program National Mail-Order Program recompete • July 2013 to June 2016 • July 2016 to December 2018 • Diabetes testing supplies • Diabetes testing supplies 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 DMEPOS (durable medical equipment, prosthetic devices, prosthetics, orthotics, and supplies), MSA (metropolitan statistical area). Round 2 recompete covers the Note: same geographic areas that were included in Round 2. However, as a result of the Office of Management and Budget’s updates to the original 91 Round 2 MSAs, there are 90 MSAs for Round 2 recompete. The specific DMEPOS items included in a product category may change between rounds. Source: Government Accountability Office and CMS. Note and Source in InDesign Eligible suppliers submit bids for one or more product negative pressure wound therapy pump product category categories in one or more CBAs. For example, a supplier includes only three HCPCS codes (Competitive Bidding Implementation Contractor 2017). could bid on the standard mobility product category in the Pittsburgh, PA, CBA. Product categories can CMS requires bids to be bona fide. To meet this criterion, comprise a number of individual products and can vary suppliers should include in their bid the cost to purchase greatly in scope. For example, the standard mobility the item, overhead, and profit. Suppliers may be asked product category in CBP Round 1 2017 includes over 150 to submit a rationale and documentation to verify that different Healthcare Common Procedure Coding System they can furnish an item for the bid amount. For example, (HCPCS) codes, ranging from walkers to power and to prove that their bids are bona fide and that they can manual wheelchairs (Competitive Bidding Implementation supply the products at the price stipulated in their bid, Contractor 2017). Other product categories include suppliers may be required to submit manufacturer fewer products. For example, in the same round, the Issues in Medicare’s medical device payment policies 140

163 invoices, receipts (including retail sales receipts), awards at least five contracts per product category and manufacturer price lists, and signed written quotes. If CBA (42 CFR § 414.414 (h)). Accordingly, CMS caps an amount for any one of a bid’s products is determined the share of the product category the agency expects a bidder to supply at a maximum of 20 percent of a given not to be bona fide, then the supplier’s entire bid for market’s potential demand. For example, if a supplier’s the product category and CBA is rejected (Competitive bid indicated that it could supply 70 percent of the demand Bidding Implementation Contractor 2014a). for a given product category, CMS disregards the 70 In their bids, suppliers indicate the volume of a product percent and assumes that the supplier can supply only they can provide in a given CBA and the price at which 20 percent of the market for the purposes of establishing they are willing to supply the product. To select winning the pivotal bid. Once contracts are awarded and suppliers bids, composite bids are first constructed for each product begin serving beneficiaries, suppliers are not limited to category. To construct a composite bid, the price that each any specific market share—that is, suppliers are free to supplier provides in its bid is multiplied by a weight for compete with other suppliers that won contracts to supply each product. The weight for a product is based on the as much of the market as possible. utilization of that item compared with other items within CMS is also required by statute to ensure that small the product category based on historic Medicare claims suppliers have an opportunity to participate in the CBP. (Centers for Medicare & Medicaid Services 2007). To that end, CMS set a target for 30 percent of suppliers Once the suppliers’ composite bids are calculated, they are under the CBP to be small suppliers. CMS defines small arrayed from least to most expensive. Winning suppliers suppliers as those with annual gross revenues of $3.5 are then selected, starting with the lowest cost bid, until million or less, including Medicare and non-Medicare the “pivotal bid” is reached. The pivotal bid is the lowest revenue (42 CFR § 414.402). If fewer than 30 percent of composite bid for a product category that includes a suppliers at or below the pivotal bid are small suppliers, sufficient number of suppliers to meet beneficiary demand then CMS offers contracts to small suppliers whose for the items in that product category (42 CFR § 414.402). composite bids were above the pivotal bid in ascending All suppliers with composite bids at or below the pivotal order based on the proximity of each small supplier’s bid are offered contracts. composite bid to the pivotal bid. CMS continues making these offers until 30 percent of the suppliers are small After the winning composite bids are selected, payment suppliers or until there are no more small suppliers who rates are determined from among those bids. While submitted composite bids for the product category (42 winning bids are selected on a composite basis, payment CFR § 414.414 (g)(1)). rates are set at the individual HCPCS code level. Specifically, the payment rate for each HCPCS code— Subsequent to the awarding of contracts, CMS also has referred to as the single payment amount (SPA)—is the discretion to award additional contracts if the agency derived from the median of all winning suppliers’ bids determines that more suppliers are needed to meet for that specific item. The CBP ensures savings to the beneficiary demand. To do so, CMS refers to the original Medicare program and beneficiaries by requiring that the arrayed list of composite bids for a product category and SPA for any product cannot exceed the fee schedule rate offers contracts to suppliers whose composite bids were for the same product. closest to the pivotal bid. These additional contracts are offered on the same terms and conditions as those awarded After CMS selects the winning composite bids and to other winning suppliers (42 CFR § 414.414 (i)(1)). calculates SPAs, the agency offers contracts to the winning suppliers. Suppliers are not required to accept contract Health status monitoring 7 offers; that is, the bids are nonbinding. If suppliers accept Concurrent with the implementation of the CBP, CMS a contract, they are referred to as contract suppliers. instituted a real-time claims monitoring system that is Beneficiaries living in CBAs must get DMEPOS products designed to analyze changes in several key secondary included in the CBP through contract suppliers, with a few 8 indicators of beneficiary access to medically necessary exceptions. DMEPOS products—mortality rates, monthly hospital Except for the National Mail-Order Program and cases admission rates, monthly emergency room rates, monthly without a sufficient number of eligible suppliers, CMS physician visit rates, monthly skilled nursing facility June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 141

164 FIGURE Emergency department use among beneficiaries likely to 6–2 FIGURE Title here... need home oxygen was lower in Round 2 competitive bidding areas 1-XX compared with non–competitive bidding areas (West region) 12 10 8 6 CBP Round 2 began CBP Round 2 recompete began 4 2 Non–competitive bidding areas emergency department per month Share of beneficiaries who visited an Round 2 competitive bidding areas 0 April April July October January July April July October October January April July October January January 2016 2016 2015 2016 2014 2017 2013 2013 2014 2015 2014 2016 2014 2015 2013 2015 Note: CBP (Competitive Bidding Program). Source: CMS health status monitoring data, March 2017 update. admission rates, average monthly days in a hospital, every case, the beneficiary reported having more than 9 and average monthly days in a skilled nursing facility. enough supplies on hand, often multiple months’ worth, Note: Note and Source are in InDesign. which suggests that beneficiaries had historically received CMS analyzes these data for each product category Source: excessive replacement supplies before they were medically and CBA for multiple cohorts of beneficiaries—all fee- necessary (Wilson 2012). Based on the results of the for-service (FFS) beneficiaries, beneficiaries who have monitoring system, CMS has said that no negative changes a claim for one of the CBP products in a given time Notes about this graph: in beneficiary health outcomes have resulted from the CBP period, and beneficiaries who are likely to use one of the • Data is in the datasheet. Make updates in the datasheet. (Centers for Medicare & Medicaid Services 2017b). CBP products on the basis of related health conditions (Centers for Medicare & Medicaid Services 2017b). These • I deleted the years from the x-axis and put in my own. CMS publicly posts aggregated data from its health status data are analyzed multiple times each month using an • I had to manually draw tick marks and axis lines because they kept resetting when I changed any data. monitoring program. In the public data, the results are algorithm designed to identify potential changes in health aggregated by region—Midwest, Northeast, South, and • The dashed line looked ok here, so I didn’t hand draw it. outcomes (Government Accountability Office 2016). If West. The data are also stratified by whether a beneficiary potential problems are identified in utilization or outcome • I can’t delete the legend, so I’ll just have to crop it out in InDesign. lives in one of the CBP Round 1 areas, Round 2 areas, or changes, CMS discusses them internally and has the • Use direct selection tool to select items for modification. Otherwise if you use the black selection tool, they will reset to graph a non-CBA. For example, Figure 6-2, using the publicly ability to follow up to determine the specific cause. For available data, shows the trend in the share of Medicare default when you change the data. example, CMS’s monitoring revealed declines in the use FFS beneficiaries who visited an emergency department of mail-order blood glucose test strips and continuous • Use paragraph styles (and object styles) to format. in each month from April 2013 through March 2017 and positive airway pressure (CPAP) device supplies in had a diagnosis in claims data indicating a potential need certain Round 1 CBAs, so CMS conducted 300 calls to for home oxygen (e.g., chronic obstructive pulmonary beneficiaries who stopped using the supplies after the disease). The data in the figure are limited to beneficiaries CBP was implemented. CMS found that, in virtually Issues in Medicare’s medical device payment policies 142

165 who lived in the West region and are stratified by whether number of beneficiaries receiving hospital beds declined a beneficiary lived in a Round 2 CBA or a non-CBA. The 37 percent for CBAs and 28 percent for non-CBAs figure reveals several patterns. First, the use of health care after CBP Round 2 was implemented (Government services varies across geographic areas, likely for reasons Accountability Office 2016). For one product category— beyond the CBP. In this case, emergency department use CPAPs—both the number of beneficiaries and items was actually lower in CBAs compared with non-CBAs, a received increased in both CBAs and non-CBAs after 10 trend that also held in the other three geographic regions. implementation of CBP Round 2. Specifically, after CBP Round 2 was implemented, the number of CPAP items Second, there appeared to be a secular trend of higher received in CBAs increased by 25 percent compared emergency department use; that is, emergency department with a 17 percent increase in non-CBAs (Government use appeared to be increasing for all beneficiaries during Accountability Office 2016). the period from 2013 to 2017. In fact, the Commission has documented that emergency department use had Critiques of the CBP been growing for the Medicare population even before the implementation of the CBP (See Chapter 1 of this The DMEPOS industry, economists, and others have report). Given these observations, Figure 6-2 does not criticized the CBP. The criticisms generally fall into suggest that a major increase in emergency department three categories—criticisms of the CBP’s structure, how utilization occurred among beneficiaries likely to need CBP information is used to adjust fee schedule payment home oxygen in the months after either of the CBP Round rates in non-CBAs, and the structure of the health status 2 competitions began. monitoring program. Regarding the CBP’s structure, the four main critiques are that: Price and utilization changes under the CBP • the bids are nonbinding (i.e., a supplier can win a bid The payment rates for DMEPOS products have declined and then reject the contract); substantially since the CBP’s implementation. Among the 25 highest expenditure DMEPOS products included SPAs are set using the median price of all winning • in the CBP (based on 2015 Medicare expenditures), the bids as opposed to the price of the pivotal bid (i.e., the median payment rate decrease was 53 percent from 2010 market-clearing price); (the year before the CBP began) to the most current CBP • composite bids are used; and round, which is CBP Round 1 2017 for most products. Among these 25 products, price declines ranged from 25 the program lacks transparency (167 Concerned • percent for certain standard power wheelchairs (HCPCS Auction Experts on Medicare Competitive Bidding code K0823) to 75 percent for blood glucose test strips Program 2010). (HCPCS code A4253) (Table 6-2, p. 144). Critics of the CBP contend that these issues, especially Utilization of DMEPOS products included in the CBP the first two, will have several negative consequences. declined more in CBAs compared with non-CBAs after First, they suggest that using nonbinding bids encourages the implementation of competitive bidding. In a 2016 “low-ball” bids whereby suppliers bid at unreasonably low report, GAO analyzed the change in the number of rates to ensure that they are offered a contract. Then, after beneficiaries utilizing a particular product and number the SPAs are announced, the low-ball bidders can decline of items received in the year before and after the the contract. Second, using the median of winning bids to implementation of CBP Round 2 in July 2013. GAO set SPAs results in half of winning bidders being offered found that the number of beneficiaries receiving a product contracts at prices less than their bids, which could result included in CBP Round 2 declined by 17 percent in CBAs in many suppliers rejecting contracts or supplying products compared with 6 percent in non-CBAs (Government 12 at a loss. In addition, critics suggest that using the 11 Accountability Office 2016). The utilization changes median of winning bids further encourages low-ball bids, varied substantially among the eight product categories since a low bid increases the chances of a supplier being included in CBP Round 2. Seven of eight product offered a contract but has a modest effect on the SPA. In categories saw declines in the number of beneficiaries total, critics of the CBP believe that these design issues receiving products after the CBP was implemented, will lead to supply shortages, as suppliers refuse to offer and most of the declines were larger than the declines unprofitable products, and a deterioration in the quality for the same products in non-CBAs. For example, the June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 143

166 TABLE Change in Medicare payment rates from 2010 to current round of CBP 6–2 for 25 highest expenditure DMEPOS products included in the CBP Median Median Medicare competitive Total fee bidding Percent Medicare schedule single change expenditures payment payment in HCPCS (in millions) rate amount payment code Description (2015) (2010) (2017) rate Oxygen concentrator $1,216 % –55 $79 $173 E1390 A4253 Blood glucose test strips 311 33 8 –75 E0601 Continuous positive airway pressure (CPAP) device 205 101 42 –58 151 A7030 Full face mask used with positive airway pressure device 171 90 –47 A7034 128 106 56 –47 Nasal interface used with positive airway pressure device –58 Negative pressure wound therapy electrical pump 112 1,553 659 E2402 63 94 Face mask interface, replacement for full face mask A7031 –46 34 E0431 17 29 91 Portable gaseous oxygen system –40 127 Hospital bed, semi-electric, with any type side rails, with mattress 89 –53 60 E0260 –53 E0470 Respiratory assist device, bi-level pressure capability, without backup 85 232 109 rate feature, used with noninvasive interface (e.g., facial mask) A7032 82 37 19 –47 Cushion for use on nasal mask interface, replacement only –53 Enteral feeding supply kit 78 11 5 B4035 76 E0562 Humidifier, heated, used with positive airway pressure device 273 140 –49 E0471 Respiratory assist device, bi-level pressure capability, with back-up rate 66 581 276 –53 feature, used with noninvasive interface (e.g., facial mask) 273 –25 K0823 Power wheelchair, group 2 standard, captain’s chair, patient weight 364 57 capacity up to and including 300 pounds 0.30 53 Enteral formula, nutritionally complete, calorically dense B4152 –44 0.54 E0143 109 48 –56 52 Walker, folding, wheeled, adjustable or fixed height Pillow for use on nasal cannula type interface, replacement only 50 26 16 –39 A7033 36 Headgear used with positive airway pressure device 48 –50 18 A7035 A7037 46 37 12 –68 Tubing used with positive airway pressure device –54 Standard wheelchair 45 56 26 K0001 7 E0570 Nebulizer, with compressor 43 17 –56 –42 0.68 1.18 43 Enteral formula, nutritionally complete, for special metabolic needs B4154 Enteral formula, nutritionally complete –43 0.37 0.65 42 B4150 4.83 Filter, disposable, used with positive airway pressure device 2.00 –59 38 A7038 CBP (Competitive Bidding Program), DMEPOS (durable medical equipment, prosthetic devices, prosthetics, orthotics, and supplies), HCPCS (Healthcare Common Note: Procedure Coding System). Numbers may be rounded. The unit of payment for the payment rates listed in the table varies (e.g., per month, per device, etc.). Some HCPCS code descriptions are shortened for brevity. All CBP prices were based on Round 1 2017 single payment amounts except A4253, which was based on the National Mail-Order Program recompete. Fee schedule rates for 2010 were calculated as a median of the state-level payment amounts except enteral nutrition codes, which were based on a national fee schedule. HCPCS codes E1007, A4221, and E0784 were excluded from this table because they were excluded from the current rounds of competitive bidding (Round 2 recompete and Round 1 2017). MedPAC analysis of CBP single payment amounts, 2015 Physician/Supplier Procedure Summary file, and DMEPOS and parenteral and enteral nutrition fee schedules. Source: of products, as suppliers engage in a “race to the bottom” needed DMEPOS products, CBP critics contend that to offer only the cheapest products to beneficiaries (167 Medicare costs might actually increase as beneficiaries Concerned Auction Experts on Medicare Competitive seek care in more expensive settings (e.g., hospitals) 13 Bidding Program 2010). (Crampton et al. 2015). If beneficiaries cannot access Issues in Medicare’s medical device payment policies 144

167 The DMEPOS industry has also criticized the use of use had been growing for the Medicare population before information from the CBP to set prices in non-CBAs. the implementation of the CBP, so increasing emergency Non-CBAs generally consist of small and moderate-size department use appears to be a secular trend with many urban areas and rural areas. The primary criticism is that likely contributing factors beyond the CBP. applying CBP rates to non-CBAs is inappropriate because the CBP’s design flaws result in prices that are artificially low. Critics also contend that suppliers in non-CBAs Non-CBP DMEPOS products cannot accept CBP payment rates because they cannot serve the volume of beneficiaries that suppliers in CBAs In 2015, non-CBP products represented $4 billion in do because CBAs have higher populations and the number Medicare spending, nearly half of all Medicare spending of suppliers in CBAs is limited based on the number of on DMEPOS. Unlike products under the CBP, payment contracts awarded. Finally, critics suggest that the cost to rates for non-CBP products are not routinely evaluated supply DMEPOS products can be higher in rural areas for accuracy, and the payment rate for many products (e.g., higher costs to deliver products in more remote continues to be based on historical supplier charges. As locations) (American Association for Homecare 2017). a result, some non-CBP products are likely mispriced. As was seen before CMS instituted the CBP in 2011, Critics of the CBP have alternately criticized CMS’s mispriced DMEPOS products can lead to rapid growth health status monitoring program but then also used the in expenditures, inappropriately high utilization, and program’s data to suggest that beneficiaries living in potential fraud and abuse. CBAs are negatively affected by the CBP. One criticism is that not all beneficiaries who might need DMEPOS There are a large number of non-CBP DMEPOS products, products are tracked because of relatively short look- but spending is concentrated among relatively few of back periods used to identify beneficiaries as having them. While the number of products varies over time, a specific diagnosis (Lewis 2012). For example, CMS Medicare paid suppliers for roughly 1,500 non-CBP tracks outcomes for beneficiaries with diabetes to ensure DMEPOS products in each year from 2010 through diabetics have sufficient access to diabetes testing 2015, compared with about 400 DMEPOS products that supplies, which are included in the National Mail-Order have ever been included in the CBP. Average spending Program. CMS defines diabetics by searching through per product is lower for non-CBP DMEPOS products FFS claims for four months—the month for which compared with CBP products, reflecting the fact that CMS the outcome is measured and three previous months. included higher expenditure DMEPOS products in the Critics contend that this four-month look-back period CBP first. Notwithstanding the lower average, a relatively is insufficient because many diabetics might not have small number of non-CBP products have substantial generated a claim in the previous four months. Other expenditures associated with them and account for a criticisms of the health status monitoring program include disproportionate share of the total non-CBP DMEPOS the lack of transparency, unsteady cohorts (i.e., the spending. For example, the top 25 products in spending beneficiaries tracked by CMS change over time), and lack represented about half of the $4 billion in non-CBP of a matched control group (National Minority Quality DMEPOS spending in 2015 (Table 6-3, p. 146). Forum 2015). While some stakeholders have criticized CMS’s health status monitoring program as inadequate, Rapid growth in expenditures for non-CBP other industry representatives have asserted that these DMEPOS products same data contradict the agency’s claims of no negative In contrast to the rapid decline in spending for products health outcomes related to the CBP. For example, industry included in the CBP, Medicare spending on non-CBP representatives have pointed to the increase in emergency products has grown. Since the implementation of department use among diabetics to suggest that diabetics competitive bidding, non-CBP DMEPOS products do not have sufficient access to diabetes testing supplies. have more commonly experienced rapid growth in However, we have seen emergency department use expenditures compared with CBP products. For example, increase among both beneficiaries with diabetes and those among all DMEPOS products with at least $10 million without diabetes. Also, as we note in the readmissions in expenditures in 2015, 9 of the 10 products with the chapter (Chapter 1) in this report, emergency department fastest growth in expenditures from 2014 to 2015 were June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 145

168 TABLE The 25 highest expenditure non–competitively bid DMEPOS products, 2015 6–3 Total Medicare expenditures HCPCS (in millions) Product description code $343 Pressure support ventilator used with non-invasive interface (e.g., mask) E0464 K0606 179 Automatic external defibrillator, with integrated electrocardiogram analysis, garment type Intermittent urinary catheter, straight tip A4351 133 Lumbar-sacral orthosis, off-the-shelf 114 L0650 105 Knee orthosis, off-the-shelf L1833 A4352 103 Intermittent urinary catheter, curved tip 97 E0748 Osteogenesis stimulator, electrical, non-invasive, spinal applications B4197 90 Parenteral nutrition solution, 74 to 100 grams of protein A5500 76 For diabetics only, fitting, custom preparation and supply of off-the-shelf depth-inlay shoe 69 E0463 Pressure support ventilator used with invasive interface (e.g., tracheostomy tube) L0648 63 Lumbar-sacral orthosis, off-the-shelf A5513 54 For diabetics only, multiple density insert, custom fabricated 54 Intermittent urinary catheter, with insertion supplies A4353 L5673 53 Addition to lower extremity, below knee/above knee, custom fabricated from existing mold or prefabricated 50 For diabetics only, multiple density insert, direct formed, molded to foot after external heat A5512 source of 230 degrees Fahrenheit or higher, prefabricated 47 Below knee, molded socket, shin, SACH foot, endoskeletal system L5301 K0861 43 Power wheelchair, group 3 standard, multiple power option, sling/solid seat/back, patient weight capacity up to and including 300 pounds B4199 42 Parenteral nutrition solution, over 100 grams of protein L0637 41 Lumbar-sacral orthosis, prefabricated item that has been customized to fit a specific patient by an individual with expertise Addition to lower extremity prosthesis, endoskeletal knee-shin system, microprocessor control 40 L5856 feature, swing and stance phase, includes electronic sensor(s) 39 Collagen dressing, sterile, size 16 sq. in. or less A6021 L5700 38 Replacement, socket, below knee, molded to patient model 38 Parenteral nutrition solution, 52 to 73 grams of protein B4193 Ostomy skin barrier, with flange, extended wear, with built-in convexity, 4x4 inches or smaller 35 A4407 31 E0483 High frequency chest wall oscillation air-pulse generator system DMEPOS (durable medical equipment, prosthetic devices, prosthetics, orthotics, and supplies), HCPCS (Healthcare Common Procedure Coding System), SACH Note: (solid ankle cushion heel). Expenditures are rounded and include beneficiary cost sharing. Some HCPCS code descriptions are shortened for brevity. Source: MedPAC analysis of 2015 Physician/Supplier Procedure Summary file. non-CBP products, with the lone CBP product being The growth in expenditures for these products is largely tubing commonly used in conjunction with CPAP devices due to growth in utilization; the increases in payment rates (A4604). Among the 25 highest expenditure non-CBP and number of Part B FFS beneficiaries between 2014 15 DMEPOS products, Medicare spending from 2014 to and 2015 were modest. These large, one-year growth 2015 grew 21 percent. Several non-CBP products grew rates were also not likely driven by changes in beneficiary even faster than this average, such as back braces (see text health, given that the relative health status of the Medicare 14 box on off-the-shelf orthotics). population is unlikely to change substantially over such a Issues in Medicare’s medical device payment policies 146

169 Rapid growth and potentially inappropriate utilization of off-the-shelf orthotics Physicians ordered braces for beneficiaries • roadly, the orthotics market can be separated into without billing Medicare for other services. three segments—off-the-shelf, custom-fitted, The 25 top-ordering physicians ordered back and custom-fabricated products. Off-the-shelf B braces for roughly 38,000 FFS beneficiaries in orthotics are prefabricated products that require minimal 2016. These physicians billed Medicare for other self-adjustment for appropriate use (42 CFR § 414.402). physician services, such as an office visit or Custom-fitted orthotics are also prefabricated but surgical procedure, for less than 1 percent of these require substantial modification by a certified orthotist beneficiaries. In contrast, we randomly sampled or someone with equivalent training. Custom-fabricated roughly 500 physicians who ordered at least one orthotics are the most individualized type of orthotic and back brace in 2016 but were not among the top are individually fabricated for the patient. 100 physicians in terms of back braces ordered and Medicare spending on off-the-shelf orthotics has found that the physician who ordered the brace also grown rapidly in the last several years. From 2014 to billed a physician service for the same beneficiary 2016, Medicare expenditures on off-the-shelf orthotics over 80 percent of the time. roughly doubled, from $255 million to $547 million. • Physicians ordered braces for beneficiaries from There are currently over 50 off-the-shelf products Many top-ordering physicians across the country. payable by Medicare, but spending is concentrated ordered back braces for beneficiaries from across on relatively few products. For example, in 2016, the country. For example, in 2016, one physician spending for one back brace product (Healthcare ordered at least 100 of the back braces we studied Common Procedure Coding System code L0650) was for beneficiaries who resided in 9 geographically $190 million and for one knee brace product (L1833) distant states—California, Connecticut, Florida, was $107 million. Expenditures for these two codes Indiana, Maryland, Massachusetts, New York, also grew rapidly. From 2014 to 2016, Medicare Ohio, and Virginia. expenditures for the back brace product grew by 311 percent (from $46 million to $190 million), while • Top-ordering physicians have a history of expenditures for the knee brace product grew by 81 Of the 12 physicians who disciplinary actions. percent (from $59 million to $107 million). ordered the highest number of back braces in 2016, we identified 9, or 75 percent, who had previously Given the rapid growth in expenditures for off-the- been disciplined by at least one state medical board shelf orthotics, we examined in greater depth one or were under investigation when their medical type of prefabricated back brace with high Medicare 16 license expired. In contrast, in 2015, less than 0.5 spending for signs of inappropriate utilization. We percent of the general population of physicians was identified several patterns involving physicians and 17 sanctioned by a state medical board. Among the suppliers suggesting that a meaningful portion of the top-ordering physicians, the severity of the actions increased use of off-the-shelf orthotics since 2014 that triggered state medical boards to act ranged could represent supplier-induced demand or even from submitting false or misleading information on potential fraud and abuse. their medical license applications to participating A limited number of physicians ordered a • in inappropriate referral schemes. For example, disproportionate share of back braces. In 2016, one top-ordering physician was put on probation over 50,000 physicians ordered at least one of for participating in a referral scheme in which she the back braces we examined for a Medicare fee- was paid $30 per patient to speak with patients for-service (FFS) beneficiary. However, only 25 over the phone and then write prescriptions for physicians ordered 20 percent of all such braces in pharmaceuticals. 2016. (continued next page) June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 147

170 Rapid growth and potentially inappropriate utilization of off-the-shelf orthotics (cont.) relationship (based on their lack of Medicare claims Roughly • Suppliers were concentrated in Florida. and the geographic distance between the physicians 7 percent of Medicare FFS beneficiaries reside and beneficiaries) from suppliers who often ship their in Florida. However, roughly 30 percent of the products to beneficiaries (based on the geographic spending increase from 2014 to 2016 on the back distance between suppliers and beneficiaries). brace product we studied was attributable to Based on a review of several telehealth companies’ suppliers located in Florida. Suppliers located in websites and other public documents, we found that Florida have a history of elevated rates of fraud and several of the top back brace–ordering physicians abuse. were employed by telehealth companies. All of this • Suppliers—especially new ones—drove the information appears to be consistent with the existence increase in expenditures. In 2016, suppliers of of supplier-funded telehealth arrangements that some durable medical equipment, prosthetic devices, industry analysts have warned could violate the anti- prosthetics, orthotics, and supplies (DMEPOS) kickback statute (Baird 2016). Under one type of furnished two-thirds of the back braces we studied, such arrangement, a supplier pays a lead-generation while physicians, physical therapists, and orthotists company to recruit Medicare beneficiaries who might furnished most of the remaining third. From 2014 want a back brace (e.g., through television advertising); to 2016, DMEPOS suppliers accounted for over the lead-generation company pays a telehealth 80 percent of the growth in Medicare expenditures company; the telehealth company pays a physician on the back brace products we studied, while to conduct a telehealth visit with beneficiaries; the the growth attributable to physicians, physical physician orders back braces; and suppliers ship the therapists, and orthotists was much smaller. braces to beneficiaries and bill Medicare. This nexus Among the 25 suppliers with the highest Medicare of relationships between certain physicians, telehealth expenditures for the back brace product we studied companies, lead-generation companies, and suppliers in 2016, 18 of them did not bill Medicare for those who predominantly mail orthoses to their customers products in 2014. appears to be driven more by financial considerations than by clinical ones. Independent of including The physicians who are driving the increasing orthoses in the Competitive Bidding Program, utilization appear to be ordering braces for policymakers may want to consider policies designed beneficiaries with whom they have a limited to limit such practices. ■ short time period. Given these facts, some of the growth payments are fraudulent. In fact, improper payments in utilization could be supplier induced and represent typically do not involve fraud. Rather, insufficient potentially inappropriate utilization. documentation errors caused the vast majority (80.4 percent) of improper payments for DMEPOS in Improper payment rates and potential fraud 2016 (Centers for Medicare & Medicaid Services and abuse 2016d). Claims are placed into this category when the documentation submitted is inadequate to support payment In addition to rapid expenditure growth, many non-CBP for the services billed. For example, a few of the more DMEPOS products tend to have high improper payment common missing pieces of documentation for DMEPOS rates, and some have been involved in cases of fraud and products include an order form for the product, a abuse over the last several years. certificate of medical necessity, and a physician evaluation While all payments made as a result of fraud are (Centers for Medicare & Medicaid Services 2016d). Even considered “improper payments,” not all improper Issues in Medicare’s medical device payment policies 148

171 though improper payments are predominantly not related (Department of Justice 2012). The case also saw several to fraud, such high rates of improper payments make it company employees (including company officers ranking difficult to determine whether all DMEPOS utilization is as high as a vice president of sales) and providers plead appropriate. guilty to or be convicted of charges including paying kickbacks to induce providers to prescribe the company’s Compared with other Part B services, DMEPOS products products, falsifying beneficiary medical records to 18 are prone to high improper payment rates. As part of its fraudulently induce Medicare to pay for the company’s Comprehensive Error Rate Testing (CERT), CMS found bone growth stimulators, and making a false statement to the improper payment rate for all DMEPOS products to a grand jury (Department of Justice 2014, Department of be 46.3 percent compared with 11.7 percent for all other Justice 2012). Part B services in 2016 (Centers for Medicare & Medicaid Services 2016d). Several categories of non-CBP DMEPOS Potentially excessive payment rates products had improper payment rates above the already Excessive payment rates can lead to inappropriately high DMEPOS average. For example, shoes designed high utilization and expenditure growth and encourage to be worn by diabetics had an improper payment rate potential fraud and abuse. To examine whether any of of 64.0 percent, and surgical dressings had an improper the highest expenditure non-CBP DMEPOS products payment rate of 84.3 percent (Centers for Medicare & had excessive payment rates, we evaluated Medicare’s Medicaid Services 2016d). In addition to the CERT report, payment rates for the 10 highest expenditure non-CBP DME Medicare administrative contractors (MACs) have DMEPOS products in 2015. To do so, we reviewed also initiated targeted service-specific prepayment reviews CMS’s payment policy changes since 2015 (if any) that (Centers for Medicare & Medicaid Services 2017d). were made to address overpriced products and compared The results of these service-specific reviews generally Medicare’s payment rates with private-payer rates and substantiate the CERT findings that DMEPOS products direct-purchase prices for two orthoses. The results are prone to high improper payment rates. For example, suggest that Medicare is substantially overpaying for from January through April 2017, one MAC found that many non-CBP DMEPOS products. the potential improper payment rate was 89 percent or higher for several non-CBP DMEPOS products— Comparison to private-payer rates parenteral nutrition (Healthcare Common Procedure To compare Medicare rates with private-payer rates, we Coding System code B4197), diabetic shoes (A5500), first determined the median Medicare payment rate for off-the-shelf back braces (L0650), and off-the-shelf knee each non-CBP DMEPOS product because payment rates braces (L1833) (Noridian Healthcare Solutions 2017a, can vary by state. We then calculated the median payment Noridian Healthcare Solutions 2017b, Noridian Healthcare 19 rate from a private-payer database. Finally, we compared Solutions 2017c, Noridian Healthcare Solutions 2017d). these two rates to determine the difference and the amount The text box (p. 151) describes some policy options, Medicare and beneficiaries would have saved if Medicare beyond competitive bidding, to reduce potentially had paid for the DMEPOS product at the median private- inappropriate utilization of DMEPOS products. payer rate in 2015. While documented cases of fraud are far less common The median Medicare payment rate was higher than the than improper payments, there have been several comparable private-payer rate in 2015 for 9 of the top 10 documented fraud cases involving non-CBP DMEPOS non-CBP DMEPOS products. For those nine products, products in recent years. One high-profile case of fraud we found Medicare’s median payment rates were 18 and abuse involved bone growth stimulators. Bone percent to 57 percent higher than median private-payer growth stimulators, or osteogenesis stimulators, are used rates. In dollars, Medicare’s median payment rates ranged to promote bone healing in difficult-to-heal fractures or from $0.60 higher per item for one type of catheter to fusions by applying electrical or ultrasonic current to over $1,100 higher per item for one type of bone growth the site of the fracture or fusion. As part of a settlement stimulator (Table 6-4, p. 150). announced in December 2012, the government detailed how one large manufacturer of bone growth stimulators For two ventilator products, we found Medicare’s obstructed a federal audit and manipulated certificates payment rates were higher than private-payer rates in of medical necessity, including having its employees 2015, but CMS lowered the payment rates in 2016. (For fill out the entire form and forging physician signatures June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 149

172 TABLE Comparison of private-payer rates to Medicare rates for 6–4 the 10 highest expenditure non-CBP DMEPOS products, 2015 Median Potential savings Percentage more if Medicare (or less) Medicare Median paid median Medicare paid fee private private-payer rate schedule relative to HCPCS payer rate rate private-payer rate (in millions) code Product description $1,561 35% $1,153 Pressure support ventilator used with non- $89 E0464 invasive interface (e.g., mask) 2,945 2,795 (5) N/A Automatic external defibrillator, with K0606 integrated electrocardiogram analysis, garment type 41 A4351 Intermittent urinary catheter, straight tip 1.33 1.93 45 29 1,130 877 Lumbar-sacral orthosis, off-the-shelf L0650 25 49 34 L1833 Knee orthosis, off-the-shelf 436 650 37 4.55 7.13 57 A4352 Intermittent urinary catheter, curved tip E0748 35 25 Osteogenesis stimulator, electrical, non- 4,318 3,191 invasive, spinal applications B4197 Parenteral nutrition solution, 74 to 100 260 322 24 17 grams of protein—premix 18 A5500 For diabetics only, fitting, custom preparation 60 71 11 and supply of off-the-shelf depth-inlay shoe 1,561 39 19 E0463 1,125 Pressure support ventilator used with invasive interface (e.g., tracheostomy tube) Note: CBP (Competitive Bidding Program), DMEPOS (durable medical equipment, prosthetic devices, prosthetics, orthotics, and supplies), HCPCS (Healthcare Common Procedure Coding System), N/A (not applicable). Some of the figures are rounded. Because of data limitations, we were unable to determine the specific month of the capped rental period for K0606 in the private-payer data, which can affect the payment rate. Given this limitation and the fact that most Medicare beneficiaries use K0606 for three or fewer months, all private claims for K0606 were assumed to be from the first three months, which means that the private-payer rate in the above table is likely a lower bound in terms of comparing the rate to the Medicare payment rate for the first three months. MedPAC analysis of 2015 MarketScan Commercial Claims and Encounters Database; 2015 Medicare durable medical equipment and parenteral and enteral Source: nutrition fee schedules; and 2015 Physician/Supplier Procedure Summary File. more information, see the text box on payment rates for Medicare could likely achieve a lower payment rate ventilators, p. 152.) For one product, a wearable automatic compared with private-payer rates; that is, private-payer external defibrillator (AED), the Medicare and private- rates likely represent an upper bound on appropriate payer rates were relatively comparable. For the remaining Medicare DMEPOS payment rates. seven products, roughly $192 million dollars would have Medicare’s payment rates for some non-CBP DMEPOS been saved in 2015 if Medicare paid the median private- products outside the top 10 highest expenditure products payer rate for all such products—approximately $154 are higher than private-payer rates. For example, the million in savings for the Medicare program and $38 two off-the-shelf orthotic codes included in Table 6-4 million in savings for beneficiaries. represented approximately $218 million of the $433 Medicare could likely save substantially more than $192 million in Medicare expenditures on off-the-shelf orthotics million per year if non-CBP DMEPOS products’ payment in 2015. For the remaining off-the-shelf orthotic codes rates were set more appropriately, for two reasons. First, with at least $1 million in Medicare expenditures in 2015, private-payer rates for some products outside the top 10 we found that Medicare’s payment rates ranged from 20 non-CBP DMEPOS products are lower compared with percent to 50 percent higher compared with private-payer Medicare’s payment rates. Second, in some instances, rates and that Medicare would have saved an additional Issues in Medicare’s medical device payment policies 150

173 Efforts to reduce potentially inappropriate utilization Demonstration since 2012. For the original seven n addition to implementing the competitive bidding states included in the demonstration, Medicare program (CBP), CMS over the last several years has expenditures fell from roughly $12 million per implemented broader initiatives that could reduce I month to $3 million per month one year after the rate of potentially inappropriate utilization, such as implementation and remained relatively steady taking additional steps to identify aberrant or suspicious thereafter (Centers for Medicare & Medicaid billing patterns among all Medicare fee-for-service Services 2015c). Because prior authorization claims before making payments and implementing new disproportionally affects suppliers who furnish safeguards to better screen existing and new Medicare products inappropriately, such a process could help suppliers (Government Accountability Office 2016). reduce improper payment rates. Some have suggested expanding certain efforts to cover a broader range of products. Three examples of CMS requires face-to-face Face-to-face visits. • initiatives that could be expanded include: visits for some DMEPOS items, such as certain hospital beds, but not for others (e.g., knee or back • Prior authorization. Prior authorization is 21 braces). To meet the requirement, a physician, a process through which suppliers request a physician assistant, nurse practitioner, or a clinical preliminary determination from CMS that a product nurse specialist must have had a face-to-face is covered before submitting an actual claim. One encounter with the beneficiary on the date the advantage of prior authorization is that it stops DMEPOS item was ordered or within six months many improper payments before they are made, before such date (42 CFR § 410.38(g)). The intent instead of trying to recoup payments after they are of requiring a face-to-face visit for certain items made. CMS currently maintains a list of durable is to ensure that a beneficiary needs a particular medical equipment, prosthetic devices, prosthetics, DMEPOS product, based on a needs assessment orthotics, and supplies (DMEPOS) products that conducted by a physician or other practitioner, could be subject to prior authorization, referred before a product is dispensed. to as the “master list.” To be added to the master list, products must have an average fee schedule Pricing, Data Analysis, and Coding (PDAC) • purchase price of $1,000 or greater, or an average contractor letters. Among other duties, the rental fee schedule of $100 or greater (adjusted PDAC contractor provides coding guidance to annually for inflation), and have been identified manufacturers on the proper use of HCPCS codes. by the Office of Inspector General, Government Manufacturers submit a product to the PDAC Accountability Office, or CMS as susceptible to contractor, and within 90 days the contractor issues high rates of fraud, unnecessary utilization, or a coding verification letter that delineates the 20 improper payments (42 CFR § 414.234). From HCPCS code(s) under which a product is billable. among the products on the master list, CMS has Some DMEPOS items already require a PDAC required prior authorization nationally for two letter before suppliers can bill for them while others power wheelchair products (Healthcare Common do not. Requiring PDAC letters for a broader array Procedure Coding System (HCPCS) codes of items could represent a modest step to help K0856 and K0861) since July 2017. Separate limit “upcoding”—that is, suppliers furnishing a from the national prior authorization process relatively inexpensive product and then submitting for these two codes, CMS has been running the a claim for a more expensive product. ■ Prior Authorization of Power Mobility Devices $55 million in 2015 if Medicare’s payment rates were diabetic shoes/inserts, also have products not included equal to the median private-payer rate of the comparable in Table 6-4 for which Medicare could have achieved 22 product. additional savings if Medicare’s payment rates were Other families of products, including bone lowered to private-payer rates. growth stimulators, catheters, parenteral nutrition, and June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 151

174 CMS revised the payment rates for ventilators in 2016 methodologies. Since being added to the fee schedule n 2015, 2 of the 10 highest expenditure durable in 2005, the payment rates for E0464 and E0463 were medical equipment, prosthetic devices, prosthetics, based on manufacturer suggested retail prices; these orthotics, and supplies (DMEPOS) products I codes were intended to represent specific types of excluded from Medicare’s Competitive Bidding ventilators, such as those used by pediatric patients. In Program (CBP) were ventilators (Healthcare Common contrast, the payment rates for the older ventilator codes Procedure Coding System (HCPCS) codes E0464 were based on supplier charges from 1986 to 1987. and E0463). Billing for these products grew rapidly The latter payment method resulted in substantially from 2010 through 2015. For example, Medicare lower payment rates. As evidence of abuse related to the expenditures for noninvasive pressure support newer, higher paid codes (E0464 and E0463) mounted, ventilators (E0464) grew from $9 million to $343 CMS, beginning in 2016, used its authority to base million over that time period, an average annual growth DMEPOS payment rates on 1986–1987 supplier charges rate of 107 percent. In a 2016 report, the Office of for all ventilators. Between 2015 and 2016, this change Inspector General noted that the rise in ventilator billing resulted in the median monthly rental rate for products was related to a change in technology that allowed the historically billed under E0464 and E0463 going from same machine to function as a ventilator, continuous $1,561 to $1,055, a reduction of 32 percent. positive airway pressure (CPAP) device, or respiratory assist device (RAD) (Office of Inspector General 2016). While the change reduced overpayments, it is unclear Compared with ventilators, CPAPs and RADs are used whether the new payment rates represent appropriate to treat lower acuity patients. prices. Specifically, the payment rates are still based on supplier charges that are 30 years old, updated over Beginning in 2016, CMS changed the way it paid for time for inflation. CMS proposed including noninvasive ventilators by collapsing five ventilator HCPCS codes pressure support ventilators in CBP Round 1 2017 but into two codes (Centers for Medicare & Medicaid removed the product before the round began. ■ Services 2015b). Specifically, in 2015, CMS was paying for five ventilator HCPCS codes using two different While private payers might have negotiated payment rates codes—L0650 (back brace) and L1833 (knee brace)—we that were lower than Medicare for products excluded identified what products were certified as payable under from the CBP, private-payer rates might not represent the those two HCPCS codes through CMS’s Pricing, Data best price that Medicare could achieve. Recent research Analysis, and Coding contractor. We then selected several suggests that average CBP Round 1 rebid payment rates approved products and conducted an internet search to were 8.1 percent lower than commercial prices for several determine the prices at which these products could be common DMEPOS products (Newman et al. 2017). directly purchased. For the off-the-shelf back brace, the Further, Medicare’s payment rates generally continued median private-payer rate in 2015 was $877; we found to fall in subsequent CBP rounds. To further illustrate multiple products eligible to be billed under that HCPCS the point that private-payer rates likely represent an code that could be purchased for less than $250. For the upper bound on Medicare rates, we looked at the direct- off-the-shelf knee brace, the median private-payer rate in purchase price—that is, the price at which beneficiaries 2015 was $436; we found multiple products eligible to be could purchase a DMEPOS product outside of insurance billed under that HCPCS code that could be purchased for coverage—for two off-the-shelf orthotic codes. less than $150. A large number of braces can be billed under each Direct-purchase price for off-the-shelf orthotic of these HCPCS codes we examined. The limited codes number of examples we examined were not designed To identify specific products (e.g., manufacturer and to be statistically representative, and other braces model) that could be billed under the off-the-shelf orthotic Issues in Medicare’s medical device payment policies 152

175 However, CMS is statutorily prohibited from including that are billable under these HCPCS codes could be substantially more expensive. However, previous OIG other groups of products in the CBP. Many of these work substantiates our finding. Specifically, in 2012, products are likely good candidates for the CBP because OIG reported that, for one type of back brace, Medicare multiple suppliers furnish the products, and Medicare’s paid an average of $919 compared with an average of payment rates appear to be substantially higher than private-payer rates. For example, CMS is statutorily $191 paid by suppliers to acquire the braces (Office of prohibited from including parenteral nutrition in the CBP, Inspector General 2012). Furthermore, the magnitude of the differences between the private-payer rates and the despite the fact that we found Medicare’s payment rate for the highest expenditure parenteral nutrition product was direct-purchase prices suggest that the private-payer rates, 24 percent higher compared with private-payer rates in while already below Medicare’s rates, do not necessarily represent the lowest payment rates that Medicare could 2015, and the agency already has substantial experience potentially obtain. successfully bidding out a similar product—enteral nutrition. In another case, Medicare’s payment rate for the highest expenditure bone growth stimulator product is even higher relative to private payers—roughly 35 percent Policy options to improve the accuracy higher—but CMS is prohibited from including such 23 of Medicare’s payment rates for non- products in the CBP because they are Class III devices. CBP DMEPOS products and protect For a third group of products, CMS’s authority is beneficiaries unclear or additional legislative authority would likely be beneficial. In the case of ostomy, tracheostomy, Shifting additional products into the CBP and urological supplies (e.g., catheters), we found two The Commission supports shifting additional DMEPOS products for which Medicare’s payment rates were 45 products from being paid on a fee schedule basis to being percent and 57 percent higher than private-payer rates. included in the CBP. Medicare’s reliance on outdated and CMS has stated that it has the authority to include certain inflated pricing information (e.g., 30-year-old supplier medical supplies in the CBP (Centers for Medicare & charges and unadjusted list prices) to set payment rates Medicaid Services 2007). However, compared with other for non–competitively bid DMEPOS products results in products, the legal authority to do so appears to be less excessive payment rates. Setting payment rates too high clear. An explicit grant of authority could accelerate the also creates incentives for higher volume, financially inclusion of these products into the CBP and protect burdens beneficiaries and taxpayers, and encourages the agency from potential legal challenges. In the case fraud and abuse. Shifting more products into the CBP of orthotics, CMS has the authority to include only off- is consistent with the Commission’s long-held support the-shelf products in the CBP. Including only off-the- of payment accuracy in FFS payment systems. Payment shelf orthotics in the CBP would likely lower costs and rates should be high enough to ensure beneficiary access reduce inappropriate unitization. However, including to needed products and low enough to encourage efficient a broader array of orthoses in the CBP would likely provision of those products. better protect Medicare by eliminating the incentive that suppliers would have to shift utilization from off- The CBP has been operating for over seven years and the-shelf products to more customizable products. For has effectively reduced excessive payment rates, reduced example, if only off-the-shelf orthotics were included in the financial burden on beneficiaries and taxpayers, and the CBP, some suppliers who did not win a contract might been an important tool to combat fraud and abuse. CMS’s simply switch to billing for more custom-fitted braces, health status monitoring program has helped ensure which are prefabricated products that require substantial beneficiaries maintain access to needed DMEPOS items modification by a trained practitioner. This behavior and is more advanced than outcomes monitoring in many would be especially likely, given that many prefabricated other sectors. products are approved to be billed under two codes—an off-the-shelf code if no customization is done and a CMS currently has the authority to include some custom-fitted code if the device is customized. We have additional products in the CBP. Examples of such products found that, in the past, suppliers have rapidly shifted the include chest wall oscillation devices, ventilators, and off- types of products they bill for based on the incentives they the-shelf orthotics. face (see text box on back braces, p. 154). June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 153

176 Suppliers can rapidly shift utilization between off-the-shelf and custom-fitted back braces Therefore, suppliers currently have an incentive to n 2014, CMS split many orthotic Healthcare furnish off-the-shelf instead of custom-fitted products. Common Procedure Coding System (HCPCS) codes into two separate codes—one for off-the- I Suppliers quickly responded to this incentive. In the three shelf products (prefabricated products that require years following this coding change, Medicare spending minimal self-adjustment) and another for custom- for the off-the-shelf back brace increased rapidly, while fitted products (prefabricated products that require spending for the custom-fitted brace decreased rapidly. substantial modification by a trained practitioner). For Specifically, from 2014 to 2016, Medicare’s expenditures example, CMS split a back brace product into L0650 for the off-the-shelf back brace increased by over 300 (an off-the-shelf product) and L0637 (a custom-fitted percent ($46 million to $190 million) compared with a product). The payment rates for the new codes are the decrease of nearly 50 percent for the custom-fitted back same, but suppliers that bill for custom-fitted products brace ($62 million to $34 million) over the same time are subjected to additional quality requirements (e.g., period (Figure 6-3). This example suggests that suppliers Appendix C of the DMEPOS Quality Standards) can rapidly shift utilization between off-the-shelf and (Centers for Medicare & Medicaid Services 2016c). custom-fitted orthoses. ■ FIGURE Cumulative change... x-x FIGURE Suppliers rapidly shifted to billing for off-the-shelf 6–3 back braces following a 2014 coding change 250 Total Custom-fitted back brace (HCPCS code L0637) 200 Off-the-shelf back brace (HCPCS code L0650) 150 100 50 Medicare expenditures (in millions of dollars) 0 2011 2013 2012 2015 2014 2016 HCPCS (Healthcare Common Procedure Coding System). Note: Source: MedPAC analysis of Medicare fee-for-service durable medical equipment claims. If the Congress grants CMS additional authority, incorporated is likely important. In the past, the Congress then requiring a date by which the products must has mandated that CMS make changes to the CBP by be incorporated into the CBP could be helpful, but certain dates, which, to some extent, protects the agency Note: Note and Source are in InDesign. flexibility regarding the manner in which the products are from industry pressure to delay the program. The deadline Source: Issues in Medicare’s medical device payment policies 154 Notes about this graph: • Data is in the datasheet. Make updates in the datasheet. • I deleted the years from the x-axis and put in my own. • I had to manually draw tick marks and axis lines because they kept resetting when I changed any data. • The dashed line looked ok here, so I didn’t hand draw it. • I can’t delete the legend, so I’ll just have to crop it out in InDesign. • Use direct selection tool to select items for modification. Otherwise if you use the black selection tool, they will reset to graph default when you change the data. • Use paragraph styles (and object styles) to format.

177 should reflect the level of effort required by CMS. For competition exists, the CBP will not produce savings, but instance, the agency would need to design any special CMS will still incur the administrative costs of including rules for the new product categories, solicit industry such products in the CBP. Table 6-5 (p. 156) provides feedback, and incorporate the new products into its health some basic information on the competitiveness of the 25 status monitoring program. To expedite the inclusion of highest expenditure non-CBP DMEPOS products in 2015, new products, the agency could be given the flexibility to using Medicare FFS claims. phase in bidding in a small number of areas or bid out the As the results in Table 6-5 indicate, in 2015, wearable new products only in a limited number of areas and use AEDs (HCPCS code K0606) did not have sufficient that information to adjust the fee schedule in the rest of the competition to include them in the CBP. From 2010 to country. 2016, Medicare FFS expenditures on wearable AEDs totaled $760 million and grew at an average annual rate of As the agency has done in the past, CMS could consider 42 percent per year, reaching $204 million in 2016 alone. allowing physicians and other providers, such as Medicare’s payment rate for wearable AEDs is likely hospitals, to furnish CBP products to their own patients excessive as a result of basing the rate on the undiscounted at the single payment amount without bidding or being manufacturer suggested retail price of the only company contract suppliers. To further encourage continuity of who manufactured the product (see text box on wearable care, policymakers could also consider allowing hospitals AEDs, p. 157, for more information). to furnish certain products to their patients without undergoing a DMEPOS accreditation process, similar to the Allowing manufacturers or suppliers, and especially accreditation exemptions currently allowed for physicians those who face little competition, to functionally set 24 and other suppliers. While allowing noncontract suppliers Medicare’s payment rates for their own products and to provide DMEPOS products could drive down the value then largely increase those rates by inflation over time of winning a contract and result in higher single payment leads to excessive payment rates. Given the fact that large amounts, they could also allow for greater convenience payment declines have occurred when products are added and continuity of care for beneficiaries. We found that to the CBP, policymakers could consider directing CMS physicians, hospitals, physical therapists, and orthotists to reduce the payment rates for wearable AEDs and other furnished a minority of the off-the-shelf back brace product products that are excluded from the CBP and that meet we studied and are not driving the increase in utilization certain other criteria, such as rapid utilization growth, and expenditures for such products. Therefore, for the back that indicate a potentially mispriced product. Future braces we examined, exempting such providers would Commission work could also further examine how to more likely increase continuity of care without substantially rationally set fee schedule rates for DMEPOS products affecting the operation of the CBP. CMS could also monitor when including them in the CBP is not practical. the implementation of such policies to make sure that the exceptions were not abused. Limiting balance billing and encouraging supplier participation to protect beneficiaries DMEPOS products that are not good Another policy option for policymakers to consider is candidates for the CBP changing Medicare’s assignment and participation rules Regardless of CMS’s authority to add certain products to for DMEPOS products and suppliers to better protect the CBP, some DMEPOS products are not good candidates beneficiaries. Unlike many other suppliers, DMEPOS for inclusion in the CBP. Two such types of products are suppliers are generally not required to accept Medicare’s those with small Medicare FFS markets and those without payment rate as payment in full (i.e., assignment is a sufficient number of suppliers to produce lower prices not mandatory) outside of CBP products furnished to 25 through competition. beneficiaries who reside in CBAs, and there is no limit on balance billing (i.e., billing beneficiaries beyond the First, even if there is sufficient competition, DMEPOS standard 20 percent coinsurance) when assignment is not products with a small Medicare FFS market could be 27 mandatory. excluded from the CBP. The principle underlying this Also, DMEPOS suppliers do not face the 5 notion is that the administrative costs of incorporating percent payment reduction that physicians do when they products into the CBP should not exceed the potential enroll as nonparticipating (a status that allows physicians 26 savings. and other suppliers to bill unassigned on a claim-by-claim Second, the CBP relies on competition among 28 basis). suppliers to produce lower payment rates. If insufficient As a result, DMEPOS suppliers are far more June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 155

178 TABLE Number of companies supplying products varied among the 6–5 25 highest expenditure non-CBP DMEPOS products in 2015 Number of companies Share of With at product’s least 1 allowed percent of charges product’s accounted for HCPCS allowed by top three All code Product description charges companies 633 11 44% E0464 Pressure support ventilator used with non-invasive interface (e.g., mask) 1 Automatic external defibrillator, with integrated electrocardiogram analysis, 100 K0606 1 garment type 3,086 15 43 A4351 Intermittent urinary catheter, straight tip Lumbar-sacral orthosis, off-the-shelf 1,073 23 15 L0650 Knee orthosis, off-the-shelf L1833 33 1,402 19 1,492 16 Intermittent urinary catheter, curved tip A4352 45 E0748 Osteogenesis stimulator, electrical, non-invasive, spinal applications 137 4 87 Parenteral nutrition solution, 74 to 100 grams of protein 345 19 27 B4197 For diabetics only, fitting, custom preparation and supply of off-the-shelf depth- 8,861 4 6 A5500 inlay shoe Pressure support ventilator used with invasive interface (e.g., tracheostomy tube) 402 15 20 E0463 17 Lumbar-sacral orthosis, off-the-shelf 1,307 16 L0648 A5513 For diabetics only, multiple density insert, custom fabricated 7 9 5,413 Intermittent urinary catheter, with insertion supplies A4353 31 937 17 Addition to lower extremity, below knee/above knee, custom fabricated from 1,267 4 23 L5673 existing mold or prefabricated For diabetics only, multiple density insert, direct formed, molded to foot after 2 6,816 A5512 4 external heat source of 230 degrees fahrenheit or higher, prefabricated L5301 1,176 3 23 Below knee, molded socket, shin, SACH foot, endoskeletal system Power wheelchair, group 3 standard, multiple power option, sling/solid seat/ 501 K0861 15 29 back, patient weight capacity up to and including 300 pounds Parenteral nutrition solution, over 100 grams of protein 249 21 27 B4199 L0637 Lumbar-sacral orthosis, prefabricated item that has been customized to fit a 6 1,913 10 specific patient by an individual with expertise L5856 Addition to lower extremity prosthesis, endoskeletal knee-shin system, 444 8 27 microprocessor control feature, swing and stance phase, includes electronic sensor(s) A6021 Collagen dressing, sterile, size 16 sq. in. or less 656 11 48 L5700 Replacement, socket, below knee, molded to patient model 1,102 4 23 B4193 268 13 32 Parenteral nutrition solution, 52 to 73 grams of protein A4407 Ostomy skin barrier, with flange, extended wear, with built-in convexity, 4x4 49 2,114 10 inches or smaller E0483 High frequency chest wall oscillation air-pulse generator system 96 3 93 CBP (Competitive Bidding Program), DMEPOS (durable medical equipment, prosthetic devices, prosthetics, orthotics, and supplies), HCPCS (Healthcare Common Note: Procedure Coding System), SACH (solid ankle cushion heel). We define “companies” as unique tax ID numbers. Source: 2015 durable medical equipment 100 percent standard analytic file. Issues in Medicare’s medical device payment policies 156

179 Rapid expenditure growth and high Medicare payment rates for wearable AEDs The implied purchase price for wearable AEDs is he wearable automatic external defibrillator substantially higher compared with direct-purchase (AED) was approved by the Food and Drug prices of nonwearable AEDs. Specifically, nonwearable Administration in 2001 and is designed for T AEDs can commonly be purchased directly for $1,500 patients at risk of sudden cardiac death who are not to $2,000 (American Heart Association 2017). Thus, immediate candidates for an implantable cardioverter- Medicare’s implied purchase price for a wearable AED defibrillator (ICD), such as patients at risk of sudden is roughly 15 times higher than the purchase price of a cardiac death but who have an active infection or whose nonwearable AED. clinical condition continues to improve (and therefore might not need an ICD) (Piccini et al. 2016). While While a reasonable payment rate for wearable AEDs technologically similar to nonwearable AEDs, wearable is likely based on a price somewhat higher than the AEDs have the clinical advantage of not needing purchase price of nonwearable AEDs (e.g., to account another individual present to initiate defibrillation. for the additional functionality, the cost of refurbishing the device between beneficiary rentals, etc.), several Between 2010 and 2016, Medicare expenditures for facts—beyond the magnitude of the price difference wearable AEDs increased from approximately $25 between wearable and nonwearable AEDs—suggest million to $204 million, an average annual growth that Medicare’s payment rate is potentially excessive. rate of 42 percent. Wearable AEDs are capped rental First, Medicare’s payment rate is based on the items, meaning that Medicare pays a monthly fee for undiscounted manufacturer suggested retail price beneficiaries to rent the product from a supplier for of the only company that manufactured the product up to 13 months. If the beneficiary uses the device (Centers for Medicare & Medicaid Services 2006). for less than 13 months, the device is returned to the The lack of competition means the sole manufacturer supplier; if the beneficiary uses the device for 13 had an opportunity to set a price as high as possible. months, ownership is transferred to the beneficiary. In Second, the manufacturer’s own data, submitted as 2018, Medicare’s payment rate for a wearable AED is part of the Healthcare Common Procedure Coding about $2,800 per month for the first 3 months and about System code assignment process, suggested the median $2,100 for months 4 through 13. Given Medicare’s manufacturing cost was under $8,000 in 2003 (Centers formula for determining the monthly payment rates for for Medicare & Medicaid Services 2006). Medicare’s capped rental items (i.e., the payment rate for the first $28,000 implied purchase price far exceeds that figure month is 10 percent of the purchase price), Medicare’s and leads to high gross profit margins. For example, for implied purchase price for a wearable AED is over the fiscal year ending in October 2011, the gross profit $28,000 in 2018. margin for wearable AEDs appears to be greater than 29 50 percent (Zoll Medical Corporation 2011). ■ likely to enroll as nonparticipating suppliers compared explanation could be that payment rates have generally with other providers. For example, in 2016, more than been adequate or excessive, so suppliers that routinely 60 percent of DMEPOS claim lines were submitted by balance billed beneficiaries would have likely lost business nonparticipating suppliers. In contrast, less than 5 percent to other DMEPOS suppliers that could profitably furnish of physicians generally enroll as nonparticipating (Boccuti the products on an assignment basis. As payment rates 2016). for DMEPOS products are reduced to more appropriate levels and less efficient suppliers drop out of the market, Historically, DMEPOS assignment rates have remained the remaining DMEPOS suppliers could try to account high despite the fact that suppliers have commonly for some of their lost revenues by balance billing enrolled as nonparticipating suppliers (and therefore beneficiaries. have the ability to bill on an unassigned basis). One June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 157

180 • Distributor model. IMD manufacturers traditionally Therefore, while nonparticipating suppliers have largely sell and distribute their products directly to hospitals. not exercised their ability to bill on a nonassigned basis, the large pool of nonparticipating suppliers poses a risk Under the distributor model, PODs operate as to Medicare beneficiaries should these suppliers begin intermediaries between device manufacturers and hospitals; that is, a device manufacturer sells a device to balance bill in response to falling payment rates. To to a POD, and the POD resells the device to a hospital mitigate that risk and to better align DMEPOS policies with the rest of Medicare, policymakers could consider at a higher price. capping balance billing and reducing the allowed fee Under the manufacturer • Manufacturer model. schedule amount by 5 percent for nonparticipating model, PODs typically sell devices that another DMEPOS suppliers. The balance billing cap could be set company manufactures on their behalf. For example, equal to the physician fee schedule cap—115 percent— a manufacturer POD might obtain a Food and Drug or somewhat higher (e.g., 125 percent) to the extent Administration clearance to market a relatively simple policymakers want to allow for added flexibility. device, such as a surgical screw, and outsource its production to a contract manufacturer. Under this model, physicians reportedly • GPO model. Physician-owned distributors form a POD to aggregate their purchasing power and get bulk discounts from manufacturers. However, Physician-owned distributors (PODs) allow physicians given the small size of PODs, it is unclear the amount to profit from the sale of medical devices they use. of negotiating leverage such entities would have with Specifically, PODs are entities that derive revenue from manufacturers relative to the hospital itself or other, selling, or arranging for the sale of, devices ordered by larger GPOs. their physician-owners for use in procedures the physician- owners perform on their own patients. The primary Prevalence of PODs and their impact on concern with PODs is that such entities create an incentive Medicare for physicians to base their preferences, such as whether to Relatively little is known about the current prevalence of operate on a patient and which instrumentation to use, on PODs. OIG found that PODs supplied spinal devices for financial rather than clinical considerations. nearly one in five spinal fusion surgeries billed to Medicare PODs have historically been concentrated in the market in 2011 and that roughly a third of hospitals purchased such for implantable medical devices (IMDs), and the spinal devices from PODs in the same year (Office of Inspector 30 31,32 implant market in particular. The IMD market is General 2013c). While these data suggest that the use of particularly fertile ground for PODs for several reasons. PODs was relatively widespread as of 2011, OIG released First, hospitals typically purchase IMDs, so any higher a special fraud alert in 2013, calling PODs “inherently costs associated with POD-supplied devices are not suspect” under the anti-kickback statute (AKS) (Office of borne by the physician-owners. Second, physicians Inspector General 2013b). The special fraud alert caused have traditionally had significant influence on hospitals’ some hospitals to reevaluate whether purchasing devices purchasing decisions, so they can help channel hospitals’ from PODs was worth the legal risk, and some ceased device purchases to their PODs. According to a 2013 doing business with PODs. However, industry stakeholders OIG report, 94 percent of hospitals that purchased from have suggested that, while the special fraud alert slowed PODs reported that surgeon preference influenced their the spread of PODs, many PODs continue to operate, and decision to purchase from PODs (Office of Inspector a 2016 report from the Senate Finance Committee found General 2013c). Hospitals have historically been willing to PODs were operating in 43 states as of December 2015 accommodate such preferences due to physicians’ ability (U.S. Senate Committee on Finance 2016). to control patient referrals and the profitability of surgical Even though Medicare does not directly pay for most lines of business. IMDs, PODs raise several concerns for the Medicare program and beneficiaries: Types of PODs PODs are commonly structured using one of three Physicians who own PODs have Increased volume. • models—a distributor, manufacturer, or group purchasing an incentive to refer more patients for surgery because organization (GPO) model: more surgeries result in more devices used. For some Issues in Medicare’s medical device payment policies 158

181 spinal conditions, appropriate treatments can range POD in which he invested over the course of more than two years (United States of America vs. Reliance Medical from physical therapy to intensive surgical procedures, Systems et al. 2014). Three other physician-owners so physician-induced demand could be a larger issue in this area compared with areas in which clinical are alleged to have received similar or higher monthly guidelines are more prescriptive. payments from their PODs (United States of America vs. Reliance Medical Systems et al. 2014). In one of • Increased intensity. Physicians who own PODs have these cases, Sabit pled guilty and was sentenced in 2017 an incentive to use more devices in a given case or (Department of Justice 2017). In connection with his refer patients for more intense procedures that require guilty plea, Sabit admitted the following: more devices. The financial incentives provided to him by his POD • PODs’ financial incentives could • Inappropriate care. caused him to use more spinal implant devices than encourage physicians to refer patients for surgery were medically necessary to treat his patients in order inappropriately, and, because they have a financial to generate more sales revenue for his POD, which interest in choosing devices that their PODs sell, to resulted in serious bodily injury to his patients. use devices of inferior quality or that are not best suited for a procedure (U.S. Senate Committee on The money he made from using his POD’s spinal • Finance 2016). implant devices motivated him either to refer patients for unnecessary spine surgeries or for more complex Higher device costs. • PODs profit from selling procedures that they did not need (Department of or arranging for the sale of devices at the highest Justice 2017). possible price. Higher device prices put pressure on hospital margins and can contribute to calls for higher Application of the anti-kickback statute and reimbursements from Medicare. Stark law to PODs Two federal laws are critical to determine the legality of Data from OIG and an example from a POD prosecuted a POD—the AKS and the Stark law. The AKS generally by the Department of Justice substantiate some of the makes it a criminal offense to knowingly and willfully concerns about PODs. Specifically, OIG found the offer, pay, solicit, or receive any remuneration to induce following: referrals of federal health care program enrollees for the The rate of spinal surgery grew faster among • furnishing or arranging for the furnishing of items or hospitals that began purchasing devices from PODs services reimbursable by federal health care programs. compared with all hospitals (16 percent vs. 5 percent, In the case of PODs, the kickback would be the payment respectively). physicians receive from their POD for arranging for the furnishing of the POD’s devices purchased by hospitals • The rate of spinal fusions—a subset of spinal surgeries for use on the physician’s patients. To violate the AKS, that are more likely to use devices—grew faster a person or entity must offer, pay, solicit, or receive among hospitals that acquired devices from PODs remuneration to induce the referral with knowledge that compared with all hospitals (21 percent vs. 9 percent, the conduct is wrongful—that is, the government must respectively). prove intent. None of the six types of spinal devices examined was • OIG has suggested that PODs are “inherently suspect” less costly per unit when purchased through a POD, under the AKS, and some industry stakeholders echo that and one—spinal plates—cost $845 more on average sentiment. However, other industry stakeholders suggest when supplied by a POD ($2,475 vs. $1,630) (Office that PODs may be structured to avoid violating the AKS. 33 of Inspector General 2013c). In practice, government prosecutions of PODs on AKS grounds have been limited. Government enforcement actions One example of a POD’s financial incentives warping against PODs may be rare at least partly because the AKS clinical judgment involves a series of cases brought by the requires proof of intent, which can be difficult to prove in Department of Justice against Dr. Aria Sabit, a POD in court. The limited number of prosecutions and the difficulty which Sabit was an investor (Apex Medical Technologies), in proving AKS cases suggest that the Stark law may need to and others (e.g., Reliance Medical Systems). Sabit was be revised to more effectively limit the use of PODs. allegedly paid an average of $17,000 per month by the June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 159

182 The Stark law is intended to prohibit physicians from some intervening entity between the DHS entity and the referring Medicare beneficiaries to certain health physician. Establishing that a financial relationship exists care facilities in which they have a financial interest. and the type of relationship is important in applying the Specifically, the Stark law (1) prohibits a physician from Stark law and determining whether an exception applies making referrals for designated health services (DHS) because some exceptions apply to only one type of payable by Medicare to an entity with which he or she (or financial relationship. an immediate family member) has a financial relationship, An ownership relationship means that a physician has an unless an exception applies; and (2) prohibits the entity ownership or investment interest in a DHS entity (e.g., from filing claims with Medicare for those referred a physician who owns a clinical laboratory). There are DHS, unless an exception applies. This prohibition is relatively few ownership exceptions, and some believe that based on the premise that physicians have a conflict of the application of the Stark law to ownership/investment interest in such situations because they have significant relationships has been relatively effective in reducing influence over patient referrals and directly profit from physician investment in DHS entities and straightforward referring their patients to facilities in which they have to regulate compared with compensation arrangements. a financial interest. Opponents of PODs suggest that However, PODs are not DHS entities, so the Stark law the incentives inherent in PODs violate the intent of the does not prohibit physician ownership or investment in Stark law and may also often violate the letter of the law PODs. (AdvaMed 2016). CMS has also said that PODs may run afoul of the Stark law (Centers for Medicare & Medicaid The second type of financial relationship is a Services 2008b). However, others believe that PODs can compensation arrangement between a DHS entity and a be structured to comply with the Stark law, and, to our referring physician. Again, compensation arrangements knowledge, no POD has yet been prosecuted based on a can be either direct or indirect. Because PODs are not violation of the Stark law. DHS entities, financial arrangements between PODs and physicians do not typically create direct compensation The principal sanction for violating the Stark law is denial arrangements. of payment for any claims involving DHS arising from a prohibited referral. (Knowing violations of the Stark law The inclusion of indirect compensation arrangements can also trigger civil monetary penalties and False Claims in the Stark law is intended to prevent DHS entities Act liabilities.) Unlike the AKS, the government does not and physicians from circumventing the Stark law need to prove intent; instead, parties are strictly liable for by channeling an otherwise prohibited arrangement Stark law violations, even inadvertent ones. through other entities. To be categorized as an indirect compensation arrangement for the purposes of the Stark A wide range of services are considered DHS, including law, three conditions must be met: clinical laboratory services, radiology services, and physical therapy services. Importantly for the application There must be an unbroken chain of financial • of the Stark law to PODs, IMDs are not DHS, but hospital arrangements between a DHS entity and the referring inpatient and outpatient services are. Generally, a “DHS physician. entity” is any person or entity that performs DHS or bills Medicare for DHS. For example, in the case of a physician The referring physician receives aggregate • who refers his or her patient to receive spinal fusion as compensation from the person or entity in the chain a hospital inpatient procedure, the DHS is the inpatient with which the physician has a direct financial facility service, and the DHS entity is the hospital. Even if relationship (e.g., the POD) that varies with the a POD sold the devices used in the fusion to the hospital, volume or value of referrals generated by the referring the POD is not a DHS entity because it neither performs physician for the entity furnishing the DHS (e.g., the nor bills Medicare for the DHS. hospital). Broadly, the Stark law defines two types of financial • The entity furnishing the DHS (e.g., the hospital) relationships—ownership/investment arrangements and knows or recklessly disregards evidence that the compensation arrangements. Either type of relationship referring physician receives aggregate compensation may be direct, meaning the relationship is between the that varies with the volume or value of referrals to the DHS entity and physician, or indirect, meaning there is DHS entity. Issues in Medicare’s medical device payment policies 160

183 FIGURE XXXX X-X FIGURE Illustrative example of an indirect compensation relationship 6–4 between a hospital and a physician-owner of a POD Payment for device Distribution of profits POD Physician Hospital Patient referral POD (physician-owned distributor). Note: MedPAC analysis of Stark law and POD–physician relationships. Source: For PODs, the unbroken chain often consists of the the compensation arrangement does not violate the • AKS; physician’s ownership interest in the POD and the POD’s sale of devices to a hospital (Figure 6-4). In general, the compensation received by the referring physician • the referring physician’s aggregate compensation from the entity with which he or she has a direct from a POD should vary with the volume or value of financial relationship must be fair market value; and referrals generated. For example, a physician’s return on investment is often a portion of the POD profits, • the compensation received by the physician from which in turn takes into account sales of devices used the entity with which he or she has a direct financial by the physician in procedures he or she referred to relationship does not take into account the volume or the hospital. Given that devices often cost hospitals value of referrals by the referring physician for the thousands of dollars per case, hospitals should be aware entity furnishing the DHS. that referring physicians who own PODs increase Meeting the first requirement appears to be perfunctory. their payments from PODs as the number of referrals Note and Source in InDesign As for the second, most PODs that avoid suspect increase. Therefore, PODs selling medical devices to a characteristics appear to not violate the AKS or, at least, hospital where physician-owners use the devices in their have not been prosecuted for doing so. With respect inpatient or outpatient surgeries appears to create an to the third element, the compensation received by the indirect compensation arrangement between the referring referring physician from a POD will generally be at fair physician and the hospital. market value if the devices sold by the POD are sold at Once a financial relationship between a physician and competitive prices. While this provision might prevent a DHS entity is established, that physician is prohibited substantially aberrant pricing, the price paid for the same from referring Medicare beneficiaries to the DHS entity device often varies substantially from one hospital to unless an exception applies. While there are many another, so there is likely substantial leeway in how PODs exceptions for direct compensation arrangements, there price their devices while still meeting the fair market value is only one for indirect compensation arrangements. The test. Regarding the last element, the payments physicians indirect compensation exception has the following key receive from their PODs do vary based on their referrals to elements: the hospital. PODs would therefore appear to fail the last criterion needed to qualify for the indirect compensation the compensation arrangement is set out in writing, • exception. However, the compensation can be deemed not signed by the parties, and specifies the services to take into account referrals so long as it complies with covered by the arrangement; the “per unit of service” rule. June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 161

184 The “per unit of service” rule states that unit-based There is a precedent for making such a change. In compensation is deemed not to take into account the 2008, CMS revisited the “per unit of service” rule as it volume or value of referrals if the compensation per unit applied to space and equipment leases. The revised rule is fair market value and does not vary during the course prohibited physicians from renting an imaging machine, of the arrangement in any manner that takes into account for instance, on a per unit or “per click” basis to a hospital 34 referrals of DHS. (i.e., the physician gets paid every time the machine is For example, if a hospital agrees to pay used) and then referring their patients to use that imaging a POD $1,000 per pedicle screw over the course of a year, machine. CMS said that such arrangements create the such an arrangement should meet the “per unit of service” incentive for overutilization; provide the incentive for rule so long as $1,000 is a fair market price for a pedicle the physician lessor to refer patients to the lessee of the screw and the $1,000 price does not increase or decrease physician’s space or equipment (rather than to entities that based on referral patterns. may employ a different, and possibly more appropriate, Potential revisions to the Stark law treatment modality); and may foster anticompetitive behavior because entities (e.g., hospitals) may enter into The Commission questions the value PODs produce for such agreements due to fears of losing the physician the Medicare program and beneficiaries. The conflict of lessor’s referrals (Centers for Medicare & Medicaid interest that PODs create is the type of problem the Stark Services 2008a). law was designed to solve—providers’ self-interest unduly influencing medical decisions. Unlike the AKS (which has In defending its proposal to no longer allow “per click” proved ill equipped to limit the use of PODs), the Stark equipment and space leases, CMS said that the agency law does not require the government to prove intent for a monitors financial arrangements in the health care industry violation to have occurred. The goal of any change to the and revises its regulatory decisions as evidence of abuse Stark law would not be to ban PODs per se, but rather to or overutilization changes. Therefore, eliminating the prohibit physician self-referral involving PODs (i.e., to application of the “per unit of service” rule to PODs could limit the use of PODs). be seen as a logical extension of CMS’s regulatory history of modifying the application of the rule as evidence of While there are several ways the Stark law could be potential abuse presents. Also, as was the case for the 2008 revised to limit the use of PODs, the Commission has revision, CMS could possibly make such a change without discussed two specific revisions: (1) eliminating the any new legislative authority. application of the “per unit of service” rule to PODs and (2) making PODs DHS entities. The second potential revision to the Stark law entails classifying PODs as DHS entities. Under such a change, The “per unit of service” rule appears to be key in allowing physicians who have an ownership stake in PODs would self-referral involving PODs that would otherwise have an ownership stake in a DHS entity and would violate the Stark law. Referring physicians commonly therefore be prohibited from referring their patients for compensation from their PODs that receive aggregate services that use devices supplied by their PODs, unless varies with the volume or value of referrals to hospitals. another exception applied. For example, there is an Such compensation creates an indirect compensation ownership exception for an entity that furnishes at least 75 arrangement for the purposes of the Stark law and would percent of its DHS to residents of rural areas. Therefore, normally result in a prohibition of POD owners referring if PODs were reclassified as DHS entities, the rural patients for surgeries in which their PODs supplied the exception would need to be amended to limit the use of devices. However, the “per unit of service” rule deems PODs in rural areas. such arrangements to not take into account the volume or value of physician referrals if the per unit compensation Reclassifying PODs as DHS entities would be a departure is fair market value and does not vary during the course from how CMS currently defines a DHS entity and would, of the arrangement based on referral patterns. Therefore, therefore, require some additional accommodations. For the only reason referrals in such arrangements appear example, the principal penalty for a Stark law violation to be legal under the Stark law is due to the “per unit is nonpayment of a claim, and given that PODs do not of service” rule, and, as a consequence, eliminating the submit claims to Medicare, specific rules stipulating rule’s application to PODs would prohibit physicians from how PODs, hospitals, or both would be held accountable referring their patients for surgeries in which their PODs for Stark law violations involving PODs would likely supplied the devices, unless another exception applied. Issues in Medicare’s medical device payment policies 162

185 be needed. Furthermore, CMS will likely require new not allow an entity to avoid being categorized as a POD, legislative authority to classify PODs as DHS entities. so long as the entity’s fundamental structure remains unchanged. If the Stark law is amended, policymakers would face several decisions to adapt the law to limit the use of PODs, To avoid being classified as a POD or being regulated by including defining a POD, considering whether additional the Stark law, some physician-owners could try to channel exceptions to protect device innovation are warranted, and money through immediate family members, become POD implementing any changes. employees, or engage in other referral schemes. For the purposes of defining a POD-owner, an immediate family Defining a POD member of the physician-owner should be included in the 35 definition of a physician-owner. To prevent PODs from The Stark law currently does not define PODs. Therefore, converting their physician-owners to employees to avoid a definition of PODs would need to be added to the Stark regulation, language could be added to the POD definition law. To ensure that the definition of PODs captures as to clarify that PODs include entities that generate revenue many PODs as possible (and as few non-POD entities as from selling medical devices ordered by a physician who possible), the definition should include characteristics that is an owner, employee, or contractor for use in procedures are common to all PODs, include characteristics as distinct performed by such physician. To prevent referral as possible from non-POD entities, cover all three types of schemes that might be designed to circumvent any POD known POD models (distributor, GPO, and manufacturer), restrictions, language could also be added to the POD and be flexible enough to cover idiosyncratic design definition, although the legality of some of these schemes features that do not alter the basic incentives of PODs. 36 is likely already questionable under current law. The core of any POD definition should be an entity that receives revenue from selling medical devices ordered by Device innovation a physician-owner for use in procedures performed by a While some believe that limiting the use of PODs could physician-owner. To ensure that the definition applies to all inhibit medical device innovation, the Commission known POD models, language could be explicitly added to concludes that innovation in the medical device market include PODs that do not directly sell devices or that do so would be largely unaffected by such changes. through contractual relationships. Using these two criteria, a basic definition of a POD could be an entity that receives Limiting the use of PODs through the Stark law would not any of its revenue from selling or arranging for the sale prohibit physician investment in companies developing (including through contractual arrangements such as group new medical devices. Rather, limiting the use of PODs purchasing organization contracts) of medical devices would prohibit Medicare payment for cases where a ordered by a physician-owner for use in procedures physician performs surgery using a device supplied by a performed by a physician-owner. company in which the referring physician has a financial interest. Some stakeholders believe that this limitation In response to prior legislative changes such as the reduces the ability of physicians to profit from their establishment of the Open Payments program, PODs have inventions, and, therefore, additional exceptions should be reportedly changed their structure while maintaining the added to the Stark law preserving physicians’ abilities to fundamental incentives embodied in PODs (U.S. Senate 37 self-refer. Committee on Finance 2016). Language could be added to the definition of a POD to ensure that superficial The Commission concludes that no additional exceptions variations in ownership and payment structures do not are needed to protect innovation in the medical device preclude a POD from being characterized as such. To market for several reasons. First, current Stark regulations that end, a POD owner could be defined as a physician protect investment interests in companies that are listed who has an ownership or investment interest in a POD, on public exchanges and that have a net value of over including ownership or investment through agents, trusts, $75 million. This provision recognizes that physician partnerships, limited liability companies, corporations, ownership in large entities is unlikely to create an unincorporated associations, or any other entity. inappropriate incentive to refer patients for services Further, the type of payment a POD owner receives—a because the physician’s impact is likely to be attenuated. commission, return on investment, profit sharing, profit A similar clause could be added to any new POD 38 distribution, or any other type of remuneration—should provisions. Second, the Commission believes physicians June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 163

186 would still be able to profit from contributing significant Absent changes in the Stark law, additional transparency intellectual capital to the development of medical devices could still help beneficiaries make informed decisions if the use of PODs were limited. The Commission argues and help enforcement agencies, payers, and others that a device is unlikely to be innovative if the only better understand the effect of PODs. Also, enhanced manner in which physicians profit from it is through transparency could be useful even if Stark law changes are using it themselves. If a device does represent an actual made, given that some PODs could continue to exist. advancement, other providers will use the device, and the Under the Open Payments program, manufacturers physician who contributed to the invention of the device of drugs, devices, biologics, and supplies are required would continue to profit. Finally, the Commission notes to annually report to CMS information about certain that physicians contributed to medical device innovation payments and other transfers of value to physicians and before the proliferation of PODs (and will continue to do teaching hospitals. GPOs must also report payments and so if the use of PODs is limited) and that physicians have transfers of value to physicians who have an ownership or many nonfinancial incentives to continue innovating. investment interest. In addition, manufacturers and GPOs are required to report ownership or investment interests Implementation issues that physicians or their immediate family members The Stark law is intended to be self-implementing to have in their companies (Medicare Payment Advisory a large degree. The potential for significant Medicare Commission 2017). The intent of the Open Payments disallowances provides a strong incentive for hospitals program is to shed light on industry ties to providers. to police their arrangements with physicians. As a consequence, many hospitals have implemented conflict- The statute that forms the basis of the Open Payments of-interest policies, especially with regard to physician program does not explicitly mention PODs. However, relationships with hospital vendors. If the Stark law were PODs that fall within the definition of an applicable changed to limit the use of PODs, hospitals would likely manufacturer or GPO must report. In its 2013 final rule adopt similar policies to protect against Stark law and establishing the Open Payments program, CMS stated additional False Claims Act liabilities by demonstrating that it intended to capture as many PODs as possible in they took reasonable measures to comply. To the extent the Open Payments program, but not every POD model active enforcement is needed, most Stark law cases that are may be covered by the program (Centers for Medicare & brought by the government are initiated by whistleblowers. Medicaid Services 2013). For example, PODs that sell or arrange for the sale of devices to only one hospital may not Even if the Stark law is changed to limit the use of PODs, fit the definition of an applicable GPO and may therefore some PODs could continue to exist. First, the Stark law not be required to report. predominantly applies to FFS Medicare, so any new restrictions would not apply to all payers. For example, In addition, some PODs that are likely covered by the the Stark law contains an exception for services provided program are failing to report. For example, a 2016 report to Medicare Advantage enrollees (42 CFR § 411.355 from the Senate Finance Committee found that many (c)). Second, while most PODs sell to hospitals, others PODs identified by the Committee staff did not appear in may sell to non-DHS entities (e.g., ambulatory surgical the Open Payments data. The report concluded that there 39 centers). Such sales are not encumbered by the Stark were serious gaps in the reporting of POD arrangements law. Finally, PODs could adapt to the new regulations under the Open Payments program (U.S. Senate 40 in some unforeseen manner that would allow them to Committee on Finance 2016). continue operating. For example, after CMS prohibited Likely as a result of the incomplete requirement for per click arrangements for space and equipment leases, PODs to report under the Open Payments program and some entities began leasing based on a block of time (e.g., underreporting by covered PODs, very few PODs appear renting an MRI machine for a day per week) rather than in Open Payments data. For example, using the 2015 Open per use. Payments data (which were released in January 2017), the Commission found that only 8 PODs reported general Improving transparency of POD–physician relationships payments to physicians, and only 16 PODs reported physician ownership (Medicare Payment Advisory The Commission maintains that the financial relationships Commission 2017). between physicians and PODs should be more transparent. Issues in Medicare’s medical device payment policies 164

187 To address this lack of reporting, the Commission supports time continuing to ensure beneficiaries maintain access to needed products. In addition, policymakers should requiring all PODs to report under the Open Payments also consider making Medicare’s DMEPOS payment program. When reporting under the Open Payments policies consistent with those of other Part B suppliers and program, PODs should identify as a POD, as opposed to clinicians by capping balance billing and giving suppliers another type of entity that is required to report. Improving the specificity of the data could improve their utility to an incentive to enroll as participating suppliers. policymakers, oversight agencies, researchers, hospitals, Because Medicare does not directly pay for most IMDs, and others. the Commission focused on policy changes to better align the incentives between physicians (who refer beneficiaries for procedures in which IMDs are used) and hospitals Conclusion (who predominantly pay for IMDs). The Commission supports limiting the use of PODs because they encourage The Commission believes that Medicare can improve its physicians to use more and more-expensive devices payment policies for both DMEPOS products and IMDs. without providing countervailing benefits. The Stark For DMEPOS products, the CBP has effectively used law could be modified to achieve that goal, and the market competition to reduce payment rates and limit Commission discussed two such options, although other fraud and abuse for over seven years. Medicare could viable approaches likely exist. ■ include additional products in the CBP, while at the same June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 165

188 Endnotes 9 CMS also employs other tools to ensure beneficiary access to 1 In this report, we define DMEPOS using Berenson-Eggers needed DMEPOS items under the CBP, including monitoring Type of Service categories D1A, D1B, D1C, D1D, D1E, D1F, inquiries to 1-800-MEDICARE, conducting secret shopping and O1C, with certain exclusions. These categories exclude calls to DMEPOS suppliers, and conducting beneficiary drugs used in conjunction with DME; we excluded such drugs satisfaction surveys. because their payment rates are set in a manner different from other DMEPOS items. For other product categories or outcome measures, the 10 differences across geographic areas varies. For example, 2 Over the same time period, the number of Medicare FFS hospital admission rates among beneficiaries with a potential beneficiaries enrolled in Part B increased by roughly 3 percent need for home oxygen tended to be higher in CBAs than non- (Boards of Trustees 2017). CBAs, both before and after CBP Round 2 was implemented. For more information on how supplier charges were used 3 11 Of the 15 areas with the largest declines in utilization to set fee schedule rates, see 42 CFR § 405.502. The time after CBP Round 2 was implemented, 12 were in Texas or period from which supplier charges were used to set payment California. CMS officials have said that the relatively large rates may vary by payment class. Payment rates for products decreases in California and Texas were likely because these introduced after that initial time period are set using a gap- states historically had high rates of potential fraud and abuse filling process that relies on, among other sources, unadjusted (Government Accountability Office 2016). list prices. 12 In practice, CMS has reported high contract acceptance rates. 4 The Food and Drug Administration classifies medical devices For example, suppliers accepted 92 percent of contracts based on the risks associated with the device. Devices are offered in CBP Round 1 2017 (Centers for Medicare & classified into one of three categories—Class I, Class II, Medicaid Services 2016b). and Class III. Class III devices are generally the highest risk devices and are therefore subject to the highest level of As part of its ongoing work to evaluate the CBP, OIG found 13 regulatory control. that CBP Round 2 did not appear to disrupt beneficiary access to CPAP/respiratory assist devices (RADs) (Office of 5 There were some differences between the CBP Round 1 Inspector General 2017). The report was inconclusive about and Round 1 rebid. For example, a CBA in Puerto Rico was whether access to CPAP/RAD supplies was disrupted. excluded from the CBP Round 1 rebid. 14 Some of the growth in off-the-shelf orthotic codes appears 6 Mail-order diabetes testing supplies were originally included to be attributable to CMS splitting existing codes into two in the Round 1 rebid. However, Round 1 rebid contracts for in 2014 (one for the off-the-shelf version and another for the mail-order diabetes testing supplies ended on December 31, custom-fitted version). 2012, and the supplies were included in the National Mail- Order Program as of July 2013. In 2015, the payment rate increase was 1.5 percent (Centers 15 for Medicare & Medicaid Services 2015a). From 2014 to 7 For future CBP rounds, suppliers will have to obtain bid 2015, the number of Part B FFS beneficiaries increased from surety bonds of $50,000 for each CBA. If a supplier rejects a roughly 33.2 million to 33.3 million, or 0.25 percent (Boards contract and its composite bid for the product category was of Trustees 2017). at or below the median composite bid rate for all suppliers included in the calculation of the single payment amounts, 16 Specifically, we examined the utilization and expenditures then the supplier will forfeit the bid surety bond (42 CFR for a prefabricated back brace when it was dispensed as an § 414.412(h)). This provision was intended to prevent “low off-the-shelf brace (L0650) or a custom-fitted brace (L0637). ball” bidders who bid unreasonably low (to ensure they are Analyzing the combined figures allowed us to determine net offered a contract) and then accept or reject the contract after increases in utilization and spending, as many suppliers began the payment rates are known. billing for L0650 instead of L0637 beginning in 2014. 8 One exception is that beneficiaries may continue to receive 17 Specifically, the Federation of State Medical Boards reported certain products from grandfathered suppliers. that only 4,091 out of 931,921 licensed physicians in the United States were disciplined by a state medical board in 2015 (Federation of State Medical Boards 2016). Issues in Medicare’s medical device payment policies 166

189 28 Assignment is mandatory for many Medicare providers. For 18 An improper payment is any payment made in error or in example, clinical diagnostic laboratory services, services of an incorrect amount; to an ineligible recipient; for ineligible nurse practitioners, ambulatory surgical center services, and goods or services; for goods or services not received; that several other categories of services are required to be billed duplicates a payment; that does not account for credit for on an assignment basis under current Medicare payment rules applicable discounts; without supporting documentation; or (Centers for Medicare & Medicaid Services 2017c). for services where documentation is missing or not available (Centers for Medicare & Medicaid Services 2016d). 29 The company that manufactures wearable AEDs (among other products) reported an increase in gross margins between 19 The MarketScan Commercial Claims and Encounters 2010 and 2011 from 54 percent to 57 percent. Part of this Database captures person-specific utilization and expenditures increase was attributable to the higher margin wearable-AED in the outpatient and other settings for active employees, early business being a larger share of the company’s overall sales retirees, COBRA continuees, and dependents insured by in 2011 compared with 2010 (Zoll Medical Corporation employer-sponsored plans. 2011). Therefore, to contribute to increasing the overall gross Implementing prior authorization involves added 20 margins up to 57 percent, the gross margin for wearable administrative costs. Therefore, limiting prior authorization AEDs was likely above 50 percent. In 2012, the company to DMEPOS products above a certain dollar value could that manufactures wearable AEDs was acquired by the Asahi help ensure the process results in savings for the Medicare Kasei Corporation, making access to more recent financial program. information regarding wearable AEDs more difficult to ascertain (Zoll Medical Corporation 2012). 21 CMS has largely suspended enforcement of this requirement even for many DMEPOS products that are required to have a 30 Some are concerned that PODs could spread to other types of face-to-face visit. implants, prosthetics, or orthotics (U.S. Senate Committee on Finance 2016). For this analysis, we examined an additional 20 HCPCS 22 codes. (One code with over $1 million in Medicare 31 Surgeries can involve multiple devices. If at least one POD- expenditures was excluded because of an insufficient number supplied device was used in a surgery, OIG counted that of private-payer claims.) surgery as using a POD-supplied device. CMS is also statutorily prohibited from including inhalation 23 32 Among hospitals that purchased spinal devices from PODs, drugs in the CBP and, per the 21st Century Cures legislation, OIG found that approximately three-quarters purchased spinal infusion drugs used in conjunction with DME. devices from PODs that manufacture their own devices: 40 percent of hospitals bought only from PODs that manufacture Such an exemption would apply to hospitals, not hospital- 24 their own devices, 19 percent of hospitals bought only from owned DMEPOS suppliers or DMEPOS suppliers that are PODs that buy devices from other entities, 36 percent of only affiliated with a hospital. hospitals bought from both types of PODs, and 5 percent of hospitals were unclear whether PODs they bought from Other products beyond these two categories might also not 25 manufactured their own devices (Office of Inspector General be ideal candidates for inclusion in the CBP. For example, 2013c). many industry representatives have suggested that highly customized products should not be included in the CBP. The 33 The OIG study did not substantiate all the concerns that have Commission could consider this topic in the future. been expressed regarding PODs. For example, the study found that surgeries in which devices were acquired through PODs At the HCPCS level, there are many non-CBP DMEPOS 26 involved fewer devices on average (12.3 vs. 14.2 when not products with relatively low expenditures. In determining acquired through PODs). Also, OIG’s findings were mixed whether a market is large enough to justify inclusion in the with regard to the complexity of surgeries at hospitals that CBP, families of HCPCS codes should be considered because acquired devices through PODs and those that did not. any given HCPCS code might have low expenditures, but a related family of products that suppliers often provide See 42 CFR § 411.354(d)(2) and (d)(3) for a description of 34 together could be large enough to justify inclusion. the unit-based special rules on compensation. Similar to other suppliers, DMEPOS suppliers are prohibited 27 from balance billing beneficiaries dually eligible for Medicare and Medicaid. June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 167

190 39 To the extent a physician has an ownership stake in an 35 The concept of a physician’s immediate family member ambulatory surgical center (ASC), his or her incentive to is used throughout the Stark law and means husband or use POD-supplied devices may be attenuated. Physicians wife; birth or adoptive parent, child, or sibling; stepparent, that have an ownership stake in ASCs have an incentive to stepchild, stepbrother, or stepsister; father-in-law, mother-in- negotiate the lowest price for their devices because the ASC’s law, son-in-law, daughter-in-law, brother-in-law, or sister-in- profits are the difference between the ASC facility payment law; grandparent or grandchild; and spouse of a grandparent and the costs (including device costs) to perform the surgery. or grandchild (42 CFR § 411.351). Applicable manufacturers and GPOs that fail to report 40 referrals For example, the Stark definition of 36 reaches referrals required information are subject to civil monetary penalties by others at a physician’s direction or control and could of up to $1,150,000 annually—up to $10,000 per instance of encompass such arrangements. In addition, there is a civil nonreporting (up to an annual maximum of $150,000) and monetary penalty for circumvention schemes that could apply. up to $100,000 per knowing instance of nonreporting (up to 37 For example, self-referral could be allowed if PODs generated an annual max of $1,000,000) (42 CFR § 403.912). In the a certain share of their business (e.g., 60 percent) from non- agency’s 2016 and 2017 annual reports to the Congress on self-referrals or for products for which a physician holds a the Open Payments program, CMS said it did not impose patent. any civil monetary penalties in program years 2014 or 2015 (Centers for Medicare & Medicaid Services 2017a, Centers 38 Under such a provision, physicians would be allowed to refer for Medicare & Medicaid Services 2016a). their patients for surgery in which their POD supplied the devices so long as the net value of the POD was $75 million or more. We believe that few, if any, PODs would currently meet this threshold, based on conversations with industry. Issues in Medicare’s medical device payment policies 168

191 References Centers for Medicare & Medicaid Services, Department of Health 167 Concerned Auction Experts on Medicare Competitive Annual report to Congress on the and Human Services. 2016a. Bidding Program. 2010. Letter to the Honorable Pete Stark. Open Payments program . Baltimore, MD: CMS. September 26. Centers for Medicare & Medicaid Services, Department of AdvaMed. 2016. Letter to the Honorable Orrin Hatch and Health and Human Services. 2016b. CMS awards contracts for Honorable Kevin Brady. Potential revisions to the physician self- the DMEPOS Competitive Bidding Program Round 1 2017. referral law (“Stark law”). February 5. Fact sheet. November 1. https://www.cms.gov/Newsroom/ American Association for Homecare. 2017. DMEPOS MediaReleaseDatabase/Fact-sheets/2016-Fact-sheets- adjusted fee methodology for non-bid areas: Stakeholder items/2016-11-01-2.html. input on Section 16008 of the 21st Century Cures Act. Centers for Medicare & Medicaid Services, Department of Health https://www.aahomecare.org/uploads/Document-Library/ and Human Services. 2016c. DMEPOS quality standards. https:// bf4ad1bc565732d3d6ccbca989cdfe09.pdf. www.cms.gov/Outreach-and-Education/Medicare-Learning- American Heart Association. 2017. CPR & first aid emergency Network-MLN/MLNProducts/Downloads/DMEPOS-Qual-Stand- cardiovascular care: AED programs Q and A. http://cpr.heart. Booklet-ICN905709TextOnly.pdf. org/AHAECC/CPRAndECC/Programs/AEDImplementation/ Centers for Medicare & Medicaid Services, Department of UCM_475254_AED-Programs-Q-A.jsp#How%20much%20 Health and Human Services. 2016d. Medicare fee-for-service does%20an%20AED%20cost. 2016 improper payments report. https://www.cms.gov/Research- Baird, J. 2016. Avoid sham telehealth arrangements. Medtrade. Statistics-Data-and-Systems/Monitoring-Programs/Medicare- https://www.medtrade.com/news/general-healthcare/Avoid-Sham- FFS-Compliance-Programs/CERT/CERT-Reports.html. Telehealth-Arrangements/. Centers for Medicare & Medicaid Services, Department of Health Boards of Trustees, Federal Hospital Insurance and Federal and Human Services. 2015a. Calendar year (CY) 2015 update for 2017 Supplementary Medical Insurance Trust Funds. 2017. durable medical equipment, prosthetics, orthotics and supplies annual report of the Boards of Trustees of the Federal Hospital MLN Matters no. MM8999. Baltimore, (DMEPOS) fee schedule. Insurance and Federal Supplementary Insurance Trust Funds. MD: CMS. Washington, DC: Boards of Trustees. Centers for Medicare & Medicaid Services, Department of Paying a visit to the doctor: Current financial Boccuti, C. 2016. Health and Human Services. 2015b. Internal Healthcare Common protections for Medicare patients when receiving physician Procedure Coding System (HCPCS) decision regarding codes Issue brief. Menlo Park, CA: The Kaiser Family services. for ventilators. https://web.archive.org/web/20150610042556/ Foundation. http:/www.cms.gov/Medicare/Coding/MedHCPCSGenInfo/ Downloads/Internal-Coding-Decisions.pdf. Centers for Medicare & Medicaid Services, Department of Health and Human Services. 2017a. Annual report to Congress on the Centers for Medicare & Medicaid Services, Department of Health . Baltimore, MD: CMS. Open Payments program and Human Services. 2015c. Medicare prior authorization of power mobility devices demonstration status update. https://www. Centers for Medicare & Medicaid Services, Department of cms.gov/Research-Statistics-Data-and-Systems/Monitoring- Health and Human Services. 2017b. Durable Medical Equipment, Programs/Medicare-FFS-Compliance-Programs/Medical-Review/ Prosthetics, Orthotics, and Supplies Competitive Bidding Downloads/PMDDemoOctoberStatusUpdate10142015.pdf. Program health status monitoring: Summary of findings thru the second quarter of 2017 . Baltimore, MD: CMS. Centers for Medicare & Medicaid Services, Department of Health and Human Services. 2013. Medicare, Medicaid, Children’s Centers for Medicare & Medicaid Services, Department of Health Health Insurance Programs; transparency reports and reporting of Medicare claims processing manual . and Human Services. 2017c. physician ownership or investment interests. Final rule. Federal Baltimore, MD: CMS. 78, no. 27 (February 8): 9458–9528. Register Centers for Medicare & Medicaid Services, Department of Health and Human Services. 2017d. Verifying potential errors and . Medicare Program Integrity Manual taking corrective actions. In Baltimore, MD: CMS. June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 169

192 Centers for Medicare & Medicaid Services, Department of Crampton, P., S. Ellermeyer, and B. Katzman. 2015. Designed Health and Human Services. 2012. Next steps for the Medicare to fail: The Medicare auction for durable medical equipment. Durable Medical Equipment, Prosthetics, Orthotics, and Supplies Economic Inquiry 36, no. 1 (January): 469–485. Competitive Bidding Program. Fact sheet. April 17. https://www. Department of Justice. 2017. Detroit-area neurosurgeon sentenced cms.gov/Newsroom/MediaReleaseDatabase/Fact-sheets/2012- to 235 months in prison for role in $2.8 million health care fraud Fact-sheets-items/2012-04-17.html. scheme. News release. January 9. https://www.justice.gov/opa/pr/ Centers for Medicare & Medicaid Services, Department of detroit-area-neurosurgeon-sentenced-235-months-prison-role-28- Health and Human Services. 2008a. Medicare program; changes million-health-care-fraud. to the hospital inpatient prospective payment systems and fiscal Department of Justice. 2014. Virginia doctor convicted on perjury year 2009 rates; payments for graduate medical education in charges in Orthofix investigation. News release. January 15. certain emergency situations; changes to disclosure of physician https://www.justice.gov/usao-ma/pr/virginia-doctor-convicted- ownership in hospitals and physician self-referral rules; updates to perjury-charges-orthofix-investigation. the long-term care prospective payment system; updates to certain IPPS-excluded hospitals; and collection of information regarding Department of Justice. 2012. Orthofix, Inc. sentenced for illegal financial relationships between hospitals. Final rule. Federal promotion of bone growth stimulators: Company to pay almost Register 73, no. 161 (August 19): 48433–49084. $8 million in criminal penalties; $42 million recovered in joint criminal/civil resolution. News release. December 20. https:// Centers for Medicare & Medicaid Services, Department of www.justice.gov/archive/usao/ma/news/2012/December/ Health and Human Services. 2008b. Medicare program; proposed OrthofixsentPR.html. changes to the hospital inpatient prospective payment systems and fiscal year 2009 rates; proposed changes to disclosure of Federation of State Medical Boards. 2016. U.S. medical physician ownership in hospitals and physician self-referral regulatory trends and actions. https://www.fsmb.org/Media/ rules; proposed collection of information regarding financial Default/PDF/FSMB/Publications/us_medical_regulatory_trends_ Federal Register relationships between hospitals and physicians. actions.pdf. 73, no. 84 (April 30): 23528–23938. Government Accountability Office. 2016. CMS’s Round 2 Centers for Medicare & Medicaid Services, Department of Health durable medical equipment and national mail-order diabetes and Human Services. 2007. Medicare program; competitive testing supplies competitive bidding programs . Washington, DC: acquisition for certain durable medical equipment, prosthetics, GAO. orthotics, and supplies (DMEPOS) and other issues. Final rule. Federal Register 72, no. 68 (April 10): 17991–18090. Government Accountability Office. 1997. Medicare: Comparison . Washington, of Medicare and VA payment rates for home oxygen Centers for Medicare & Medicaid Services, Department of DC: GAO. Health and Human Services. 2006. Pricing workbook for K0606 . Baltimore, MD: automatic external defibrillator—garment type Evaluation of Karon, S., T. Hoerger, R. Lindrooth, et al. 2003. CMS. Medicare’s competitive bidding demonstration for DMEPOS: Final evaluation report . Prepared under contract to the Centers Competitive Bidding Implementation Contractor. 2017. for Medicare & Medicaid Services. Report prepared for the Product categories: Round 1 2017 HCPCS codes. https:// Medicare Payment Advisory Commission. Research Triangle dmecompetitivebid.com/Palmetto/Cbicrd12017.Nsf/files/14_ Park, NC: RTI International. June. HCPCS_Excel.xls/$File/14_HCPCS_Excel.xls. Lewis, A. 2012. Letter from the Disease Management Purchasing Competitive Bidding Implementation Contractor. 2014a. Consortium International to Secretary Sebelius regarding the Requirement to submit a bona fide bid. https://www. methodology of the Health Status Monitoring data used by CMS. dmecompetitivebid.com/Palmetto/Cbicrd2Recompete.Nsf/ August 6. http://www.cramton.umd.edu/papers2010-2014/lewis- files/23_Fact_Sheet_Requirement_to_Submit_a_Bona_Fide_Bid. letter-to-secretary-sebelius-6-aug-2012.pdf. pdf/$File/23_Fact_Sheet_Requirement_to_Submit_a_Bona_ Fide_Bid.pdf. Report to Medicare Payment Advisory Commission. 2017. . the Congress: Medicare and the health care delivery system Competitive Bidding Implementation Contractor. 2014b. Washington, DC: MedPAC. Round 2 recompete & national mail-order recompete request for bids (RFB). http://www.dmecompetitivebid.com/Palmetto/ Cbicrd2Recompete.Nsf/files/R2RC_Request_for_Bids_RFB. pdf/$File/R2RC_Request_for_Bids_RFB.pdf. Issues in Medicare’s medical device payment policies 170

193 Office of Inspector General, Department of Health and Human National Minority Quality Forum. 2015. Centers for Medicare & Services. 2013b. Special fraud alert: Physician-owned entities. Medicaid Services Competitive Bidding Program: Assessment of impact on beneficiary acquisition of diabetes-testing supplies and March 26. https://oig.hhs.gov/fraud/docs/alertsandbulletins/2013/ pod_special_fraud_alert.pdf. durable medical equipment prosthetics orthotics and supplies– . Washington, DC: NMQF. http://www. associated health outcomes Office of Inspector General, Department of Health and Human nmqf.org/wp-content/uploads/2015/11/National-Minority-Quality- Services. 2013c. Spinal devices supplied by physician-owned Forum-Report-on-CMS-Competitive-Bidding-Program.pdf. . Washington, DC: distributors: Overview of prevalence and use Newman, D., E. Barrette, and K. McGraves-Lloyd. 2017. OIG. Medicare Competitive Bidding Program realized price savings Office of Inspector General, Department of Health and Human 36, no. 8 Health Affairs for durable medical equipment purchases. Services. 2012. Medicare supplier acquisition costs for L0631 (August 1): 1367–1375. back orthoses . Washington, DC: OIG. Noridian Healthcare Solutions. 2017a. Knee orthoses (HCPCS Office of Inspector General, Department of Health and Human L1832, L1833) quarterly results of service specific prepayment Most power wheelchairs in the Medicare program Services. 2011. review. https://med.noridianmedicare.com/web/jadme/cert- did not meet medical necessity guidelines . Washington, DC: OIG. review/mr/complex-notifications-results/knee-orthoses-hcpcs- l1832-l1833-quarterly-results-of-service-specific-prepayment- Office of Inspector General, Department of Health and Human review. Comparison of prices for negative pressure Services. 2009. wound therapy pumps . Washington, DC: OIG. Noridian Healthcare Solutions. 2017b. Parenteral nutrition (HCPCS B4197, B4199) quarterly results of service specific Office of Inspector General, Department of Health and Human prepayment review. https://med.noridianmedicare.com/web/ Services. 2006. Medicare home oxygen equipment: Cost and jadme/cert-review/mr/complex-notifications-results/parenteral- servicing . Washington, DC: OIG. nutrition-hcpcs-b4197-b4199-quarterly-results-of-service- specific-prepayment-review. Office of Inspector General, Department of Health and Human Medicare and FEHB payment rates for home Services. 2005. Noridian Healthcare Solutions. 2017c. Spinal orthoses (HCPCS oxygen equipment . Washington, DC: OIG. L0648, L0650) quarterly results of service specific prepayment review. https://med.noridianmedicare.com/web/jadme/cert- Office of Inspector General, Department of Health and Human review/mr/complex-notifications-results/spinal-orthoses-l0648- Services. 2004. A comparison of prices for power wheelchairs in l0650-quarterly-results-of-service-specific-prepayment-review. the Medicare program . Washington, DC: OIG. Noridian Healthcare Solutions. 2017d. Therapeutic shoes Piccini, J. P., L. A. Allen, P. J. Kudenchuk, et al. 2016. Wearable (HCPCS A5500) quarterly results of service specific prepayment cardioverter-defibrillator therapy for the prevention of sudden review. https://med.noridianmedicare.com/web/jadme/cert- cardiac death: A science advisory from the American Heart review/mr/complex-notifications-results/therapeutic-shoes-hcpcs- Circulation 133, no. 17 (April 26): 1715–1727. Association. a5500-quarterly-results-of-service-specific-prepayment-review. U.S. House of Representatives. 2013. American Taxpayer Relief Office of Inspector General, Department of Health and Human Act of 2012. 112th Cong., 2nd sess., H.R. 8. Services. 2017. Round 2 competitive bidding for CPAP/RAD: . Disrupted access unlikely for devices, inconclusive for supplies U.S. House of Representatives. 2010. Patient Protection and Washington, DC: OIG. Affordable Care Act. 111th Cong., 2nd sess., H.R. 3590. Office of Inspector General, Department of Health and Human U.S. House of Representatives. 2003. Medicare Prescription Services. 2016. Escalating Medicare billing for ventilators raises Drug, Improvement, and Modernization Act of 2003. 108th concerns . Washington, DC: OIG. Cong., 1st sess., H.R. 1. Office of Inspector General, Department of Health and Human U.S. House of Representatives. 1997. Balanced Budget Act of Inappropriate and questionable Medicare billing Services. 2013a. 1997. 105th Cong., 1st sess., H.R. 2015. for diabetes test strips . Washington, DC: OIG. U.S. Senate Committee on Finance. 2016. Physician owned distributorships: An update on key issues and areas of congressional concern . A Senate Finance Committee majority staff report. Washington, DC: Government Printing Office. June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 171

194 United States of America vs. Reliance Medical Systems et al. Zoll Medical Corporation. 2012. Asahi Kasei completes 2014. United States of America vs. Reliance Medical Systems acquisition of Zoll. https://www.zoll.com/news- LLC; Apex Medical Technologies LLC; Kronos Spinal releases/2012/04/26/asahi-kasei-completes-zoll-acquisition. Technologies LLC; Bret Berry; John Hoffman; Adam Pike; and Zoll Medical Corporation. 2011. Form 10–K submitted to Aria O. Sabit, M.D. United States District Court (Central District the Securities and Exchange Commission. https://www.sec. of California) Civ. Action No. 14–6979. gov/Archives/edgar/data/887568/000119312511321014/ Wilson, L. 2012. Testimony by Laurence Wilson, Director of the d221219d10k.htm#tx221219_6. Chronic Care Policy Group, Centers for Medicare & Medicaid Services. Medicare Durable Medical Equipment Competitive Bidding Program, before the Committee on Ways and Means Subcommittee on Health, U.S. House. 112th Congress, 2nd sess. May 9. Issues in Medicare’s medical device payment policies 172

195 CHAPTER 7 Applying the Commission’s principles for measuring quality: Population-based measures and hospital quality incentives

196

197 CHAPTER 7 Applying the Commission’s principles for measuring quality: Population-based measures and hospital quality incentives Chapter summary In this chapter The Commission has recommended that Medicare link payment to quality • Introduction of care to reward accountable entities and providers for offering high-quality care to beneficiaries. The Commission has recently formalized a set of • Applying the Commission’s principles for measuring principles for measuring quality in the Medicare program. Overall, quality quality to population-based measurement should be patient oriented, encourage coordination, and promote measures delivery system change. Medicare quality incentive programs should use a small set of outcomes, patient experience, and value measures to assess the • Applying the Commission’s quality of care across different populations, such as beneficiaries enrolled in principles for measuring Medicare Advantage (MA) plans, accountable care organizations (ACOs), quality to hospital quality incentives and fee-for-service (FFS) in defined market areas, as well as those cared for by specified hospitals, groups of clinicians, and other providers. Applying the Commission’s principles, Medicare quality incentive programs should score these risk-adjusted, population-based measure results against absolute performance thresholds and then use peer grouping to determine payment adjustments based on the provider’s quality performance. In this chapter, we first apply the Commission’s principles to two population-based outcome measures (potentially preventable admissions and home and community days (formerly known as “healthy days at home”)) that may be used to evaluate quality of care for different populations. Next, we apply the principles to the design of a new hospital quality incentive program that combines measures of hospital outcomes, patient experience, and Medicare spending per beneficiary. June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 175

198 Applying the Commission’s principles for measuring quality to population-based measures We analyzed the utility of two population-based measure concepts to assess the quality of FFS care at market-area levels (e.g., geographic areas representing local health care market areas) and whether there is enough variation in performance to allow comparisons of FFS quality of care across market areas. Potentially preventable admissions Potentially preventable admissions (PPAs) constitute an important quality measure because hospitalizations for conditions such as diabetes and pneumonia can potentially be preventable if ambulatory care is provided in a timely and effective manner. To build on the Commission’s work testing the measurement of PPAs in FFS Medicare and across Medicare payment models, we applied a quality measure developed for MA to FFS administrative claims data. We calculated the observed rate of PPAs per 1,000 FFS beneficiaries for both chronic (e.g., diabetes) and acute (e.g., bacterial pneumonia) conditions. We found that observed (that is, not risk-adjusted) PPA rates varied across population groups (e.g., age, sex, Medicaid eligibility) and across two different definitions of market areas. This variation signals opportunities to improve the quality of care within areas and the potential to use the measure to compare quality across local health care markets. However, more development is needed to incorporate risk adjustment based on FFS data in the analysis. Home and community days The Commission tested a prototype home and community days (HCDs) measure to assess how well health care markets and organizations that take responsibility for a population keep people alive and out of health care institutions. The HCD measure is defined as 365 days minus the sum of (1) the number of days in the year that a beneficiary spends in certain institutional (e.g., hospital, skilled nursing facility) and ambulatory (e.g., emergency department) health care settings and (2) the number of mortality days (i.e., the number of days in the year that a beneficiary was not living, if any). We calculated risk-adjusted HCDs from 2013 to 2015 for two populations of FFS beneficiaries (all beneficiaries 65 years and older and beneficiaries 65 years and older with two or more chronic conditions). In 2015, the adjusted HCD rate for beneficiaries 65 years and older was 351 days compared with 328 days for beneficiaries 65 years and older with 2 or more chronic conditions. We also compared the distribution of mean, risk-adjusted HCDs by MedPAC-defined market areas and hospital service areas. For the group of all beneficiaries 65 Applying the Commission’s principles for measuring quality: Population-based measures and hospital quality incentives 176

199 years and older, the difference in HCDs was only 3 days between high- and low- performing market areas; the difference was only 9 days for the group with 2 or more chronic conditions. However, because of the limited variation in HCDs over market areas and the challenges posed by the need to develop appropriate weights for constructing the composite measure, the Commission questions the immediate utility of the HCD measure in its current form to assess market-level FFS performance. The Commission has continued interest in developing claims-calculated, population-based outcome measures. Ideas for population-based measures include “mean time between failure” (e.g., mean time between hospitalizations), successful community discharge, home-to-home transition time, end-of-life care and burdensome transitions, and low-value care. Applying the Commission’s principles for measuring quality to hospital quality incentives We also examined the potential to create a single quality-based payment program for hospitals in light of Medicare’s experience with four hospital payment incentive programs: the Hospital Inpatient Quality Reporting Program, Hospital Readmissions Reduction Program, Hospital-Acquired Condition Reduction Program (HACRP), and Hospital Value-based Purchasing (VBP) Program. The Commission’s and others’ main concerns about these programs are that (1) there are too many overlapping hospital quality payment and reporting programs, which creates unneeded complexity in the Medicare program; (2) all-condition measures are more appropriate to measure the performance of hospitals rather than the condition-specific readmissions and mortality measures currently used; (3) the existing programs include process measures and measures not consistently reported by providers; and (4) some of the programs score hospitals using “tournament models” (providers are scored relative to one another) rather than on clear, absolute, and prospectively set performance targets. Ideally, the Congress could redesign the multiple hospital quality payment programs under a single hospital value incentive program (HVIP) that would be patient oriented, encourage coordination across providers and time, and promote change in the delivery system. Although CMS likely has the authority to make some of our suggested changes to hospital quality payment without congressional action (e.g., improving public reporting), other key reforms would require statutory changes. The Commission asserts that the Medicare program should consider differences in providers’ patient populations—which affect providers’ performance on quality measures, including social risk factors—and that Medicare should account for June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 177

200 social risk factors in quality programs by adjusting payment through peer grouping. Applying these principles, we modeled an HVIP in which quality-based payments are distributed to hospitals organized under 10 peer groups based on the share of fully dual-eligible beneficiaries treated. (Fully dual-eligible beneficiaries are covered by both Medicare and Medicaid, and so we use this population category as a proxy for low income as a social risk factor.) In our model, the HVIP is budget neutral, with awards funded by a payment withhold from all hospitals. Our HVIP model uses a 2 percent withhold, which is the same as the existing VBP program uses, but policymakers could raise or lower the withhold amount. Under our HVIP model, relative to the 2 percent withhold, about half of hospitals would receive a negative payment adjustment, and about half would receive a positive adjustment. Most hospitals rewarded under the existing programs would continue to receive rewards, and hospitals currently incurring penalties would continue to do so. Our peer grouping of hospitals allowed us to examine how hospitals serving large shares of low-income patients perform. We found that, compared with the existing quality payment programs, the HVIP approach makes payment adjustments among hospitals that serve different populations more equitable. Over the next year, the Commission plans to continue to refine a design for an HVIP that conforms with our principles for quality measurement. Some topics the Commission will further explore include weighting of measures, withhold values, patient experience measures, and patient safety measures. ■ Applying the Commission’s principles for measuring quality: Population-based measures and hospital quality incentives 178

201 to beneficiaries. Based on the Commission’s principles, Introduction Medicare quality incentive programs for these accountable entities should score risk-adjusted measure results against The Commission contends that Medicare payments should absolute performance thresholds and then use peer grouping not be made without consideration of the quality of care to adjust payment based on performance. Medicare’s delivered to beneficiaries. The Congress has enacted use of the same set of measures and scoring framework quality reporting programs for almost all of the major across populations could also promote other payers (e.g., fee-for-service (FFS) provider types and for Medicare Medicaid and commercial) using the same systems, which Advantage (MA) and Part D plans, and it has mandated could reduce the burden providers face in tracking a diverse pay-for-performance (which Medicare refers to as number of quality measures and methodologies across value-based purchasing) for hospitals, dialysis facilities, payers. physicians, accountable care organizations (ACOs), and In this chapter, we first apply the Commission’s alternative skilled nursing facilities. Over the past several years, policy and principles to test the use of two population-based the Commission has expressed concern that Medicare’s outcome measures (potentially preventable admissions quality measurement programs are “overbuilt,” relying (PPAs) and home and community days (HCDs) (formerly on too many clinical process measures that are, at best, known as “healthy days at home”)) to evaluate FFS quality weakly correlated with health outcomes of importance to of care and beneficiary access to health care in local beneficiaries and the program. Relying on a large number health care market areas. We wanted to test the use of the of process measures can reinforce payment incentives measures for the FFS population in health care markets in FFS to overprovide and overuse measured services. before applying the measures to other populations. Next, we Process measures are also burdensome for providers to apply the Commission’s principles to the design of a new report, while yielding limited information to support hospital quality payment program that uses current hospital clinical improvement. Although CMS has been shifting outcome, patient experience, and Medicare spending per away from process to outcome measures in some of the beneficiary measures. Medicare quality programs, more work is needed to align the quality measurement systems with the Commission’s principles for measuring quality (see text box, p. 180). Applying the Commission’s principles for Applying quality measurement principles measuring quality to population-based across populations measures In the June 2014 and 2015 reports to the Congress, the Commission put forth a concept for an alternative to This chapter presents our analysis of two claims-based, Medicare’s current system for measuring the quality population-based measures: PPAs and HCDs. Our of care provided to beneficiaries (Medicare Payment analyses are meant to test whether the two measures can Advisory Commission 2015a, Medicare Payment be used to evaluate quality of care for FFS beneficiaries Advisory Commission 2014). This alternative led to the and compare performance across local health care development of the Commission’s principles on quality market areas, before applying the measures to other measurement—in particular, encouraging providers to work populations. across the delivery system. Under this alternative policy, Medicare quality incentive programs would use a small Potentially preventable admissions set of outcomes, patient experience, and value measures Hospital stays can pose risks to patients, particularly to assess the quality of care across different populations, the elderly. Adverse events represent a prominent risk, such as beneficiaries enrolled in Medicare Advantage (MA) including iatrogenic infections, medication errors, plans, accountable care organizations (ACOs), and fee- device failures, and pressure injuries such as decubitus for-service (FFS) in defined market areas, as well as those ulcers. According to researchers at the Centers for cared for by specified hospitals, groups of clinicians, and Disease Control and Prevention (CDC), on any given other providers. Medicare can link quality performance day, approximately 1 in 25 U.S. patients contracts at to payment using such measures to create incentives for least one infection during the course of hospital care MA plans, ACOs, and providers to offer high-quality care June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 179

202 The Commission’s principles for measuring quality in the Medicare program • The Medicare program should take into account, he Commission has recently formalized a set of as necessary, differences in a provider’s patient principles for measuring quality in the Medicare population, including social risk factors. Because program, principles that we apply in developing T adjusting measure results for social risk factors can quality measures, modeling the design or redesign of mask disparities in clinical performance, Medicare quality incentive (or value-based purchasing) programs, should account for social risk factors by directly and commenting on CMS proposals for quality adjusting payment through peer grouping. measurement. Over recent years, the Commission has articulated elements of these principles in its policy Medicare should target technical assistance • development process, but we now present them in resources to low-performing providers. a complete framework for evaluating Medicare’s approaches to assessing quality of care. The Medicare should support research and data • Commission’s principles are as follows: collection to reduce measurement bias, including, for example, the effects of social risk factors. • Quality measurement should be patient oriented, encourage coordination across providers and time, The Commission also maintains that the goal of and promote relevant change in the nature of the improved care should extend to all patients, regardless delivery system. of health status, income, and race. Recognizing that those expectations are more likely to be met if they Quality measurement should not be unduly • are combined with additional resources to accelerate burdensome for providers. a provider’s ability to address particularly challenging care delivery environments, the Commission • Medicare quality programs should include recommended in June 2011 that the Quality population-based measures such as outcomes, Improvement Organization Program be fundamentally patient experience, and value (e.g., Medicare restructured and that funding be reprogrammed to give spending per beneficiary, measures of services that providers and communities more choices in who assists have little or no clinical benefit). Providers may them in quality improvement activities and flexibility choose to use more granular measures to manage in how resources can be used. (Medicare Payment their own quality improvement. Advisory Commission 2011). The Commission also • Medicare quality programs should give rewards recommended that Medicare make technical assistance based on clear, absolute, and prospectively set to low-performing providers and community initiatives performance targets (as opposed to “tournament a high priority as a strategy to complement payment models,” under which providers are scored relative policy and address persistent health care disparities. ■ to one another). 1 (Centers for Disease Control and Prevention 2016b). In care is provided in a timely and effective manner. addition, the inpatient environment itself can lead to a PPAs can fall into five categories: system related (e.g., reduction in elderly patients’ independence as they cope unavailability of services), physician related (e.g., with functional loss that can stem from extended bed suboptimal monitoring), medical (e.g., medication side rest (Covinsky et al. 2011). Furthermore, the hospital effects), patient related (e.g., delay in seeking help), and environment often hinders discussion about treatment social (e.g., lack of social support) (Freund et al. 2013). options. Evidence also suggests that effective primary care is associated with lower PPAs (Gao et al. 2014). The patient Hospitalizations due to conditions such as diabetes and may have required acute-level services at the time he or pneumonia are potentially preventable if ambulatory Applying the Commission’s principles for measuring quality: Population-based measures and hospital quality incentives 180

203 she sought care, but the need for the admission might Calculating potentially preventable admissions in the FFS population have been avoided with appropriate ambulatory care and coordination activities. To further test the concept of measuring PPAs for FFS beneficiaries and to compare performance across market Rates of PPAs calculated through administrative claims areas, we used a 2018 measure specification developed data can reflect the quality of the care provided under by AHRQ and adopted with permission by the National payment models and by providers in a local market area Committee for Quality Assurance (NCQA). The measure (that is, a defined population). High-quality MA plans in specification is publicly available as part of NCQA’s a local market area should be able to manage beneficiary, ® Healthcare Effectiveness Data and Information Set hospital, and physician relations to coordinate care and ® (HEDIS ), and the measure is written for the Medicare provide appropriate access (Wholey et al. 2003). High- 2 population, specifically for MA plans to report. In the quality ACOs should also be able to manage relationships summer of 2018, MA plans will report measure results to to improve care. For example, ACOs can provide tools and CMS using the 2018 measure specification (collected by data to clinicians about patients with chronic ambulatory NCQA), along with other quality measures that are used to care–sensitive conditions (such as diabetes and asthma) 3 calculate star ratings. Thus, in the future, we may have the so they can appropriately monitor, coordinate, and follow ability to use one PPA measure specification to compare up with patients and reduce avoidable hospitalizations. performance across MA, FFS, and ACOs nationally and FFS clinicians can also play a role in affecting admissions within markets. in the ambulatory care area they serve by effectively coordinating with other providers and offering adequate The HEDIS (MA) PPA measure represents the observed access to beneficiaries. For example, a clinician’s rate of PPAs and the risk-adjusted ratio of observed-to- availability for appointments can affect how well a expected potentially preventable admissions. PPAs are patient’s chronic conditions are managed and whether calculated for chronic conditions (e.g., diabetes) and acute 4,5 a patient’s acute conditions (such as pneumonia) can be conditions (e.g., pneumonia). (Although we chose to identified and treated outside of the hospital in a timely analyze this measure specification for FFS, we are not manner (Davies et al. 2009). endorsing any approach to measuring PPAs. We are simply exploring the use of PPAs as a population-based measure The Commission’s prior work on measuring PPAs of ambulatory care.) Comparing FFS and MA plan quality In our June 2014 and 2015 reports to the Congress, the performance in a local area is a future goal of this work, Commission included PPAs for inpatient hospital care as a so we did not make changes to the HEDIS specification population-based measure concept for evaluating quality in in order to permit “apples-to-apples” comparisons among a market area (Medicare Payment Advisory Commission Medicare payment models. Our analysis examines PPAs 2015a, Medicare Payment Advisory Commission 2014). for Medicare FFS beneficiaries ages 67 and older because Over the years, we have used two existing measure the HEDIS specification requires two years of beneficiary specifications to define the concept of “potentially enrollment in the MA plan. In future analyses, we could preventable” and measure FFS rates accordingly. In the apply the measure to different populations, including March 2017 report to the Congress, we presented national the under–age 67 population. We did not calculate risk- rates and variation by market areas using a definition of adjusted numbers of expected discharges because the PPAs developed by 3M (Medicare Payment Advisory regression model NCQA uses to calculate the expected Commission 2017). During the Commission’s November results is based on the risk profiles of a sample of MA 2016 meeting, the Commission expressed concern that the beneficiaries. Since MA plan populations and the coding TM 3M measure was not available in the public domain and intensity of diagnoses differ from FFS, we would need that providers could find the measure definitions overly to develop FFS-based risk weights to calculate expected complicated. In recent March reports to the Congress, we results. We therefore focused our analysis on the observed also published CMS-reported rates of hospitalizations rate of unadjusted PPAs per 1,000 beneficiaries ages 67 based on the Agency for Healthcare Research and Quality and older. We also focused on national results and not (AHRQ) Prevention Quality Indicator (PQI) specifications results at the market level because unadjusted results for three individual ambulatory care–sensitive conditions would not capture any underlying differences in market- (diabetes, congestive heart failure (CHF), and bacterial area population characteristics. pneumonia). June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 181

204 TABLE PPAs per 1,000 FFS beneficiaries ages 67 and older vary by population group, 2016 7–1 PPA rate per 1,000 FFS beneficiaries Total Chronic conditions Acute conditions Age groups 8.4 12.0 20.3 67–74 20.0 17.0 37.0 75–84 85+ 34.3 31.3 65.6 Sex 17.6 Male 13.2 30.9 17.8 17.0 Female 34.8 Medicaid eligibility Fully dual eligible 32.2 34.0 66.1 Partially dual eligible 22.4 33.3 55.7 Non–dual eligible (Medicare only) 13.3 15.5 28.7 15.3 17.7 Total 33.1 Note: PPA (potentially preventable admission), FFS (fee-for-service). To evaluate the utility of measuring PPAs for FFS beneficiaries, we calculated the observed (not risk- adjusted) rates of admissions tied to acute (e.g., pneumonia), chronic (e.g., diabetes) and total (acute plus chronic) conditions. Rates presented are the number of PPAs divided by the number of beneficiaries in the qualifying population, multiplied by 1,000. The qualifying population is the same across the acute and chronic categories. Beneficiaries who died in the measurement year are excluded. Fully dual-eligible beneficiaries qualify for a full range of Medicaid benefits, and partially dual-eligible beneficiaries qualify for Medicaid payment of the Medicare premium and perhaps the cost sharing for Medicare services. Source: MedPAC analysis of 2016 Medicare claims data. We found that it is feasible to calculate unadjusted, severity based on CMS’s hierarchical condition categories observed PPAs for FFS beneficiaries nationally and for (CMS–HCCs) because we have access to these FFS two different geographic area levels representing local data. In the future, if the PPA measure is considered for health care markets (MedPAC-defined market areas a Medicare quality payment program, we can test the use designed to match insurance markets served by private of peer grouping to account for differences in the social plans and Dartmouth-defined hospital service areas risk factors of populations. The Commission continues to (HSAs), which are collections of zip codes that represent encourage CMS to support research and data collection to a local market area whose residents receive most of their improve our ability to take into account social risk factors. inpatient care from the hospitals in that area). We also Qualifying population The qualifying population for found variation by population groups (e.g., age, sex, the PPA measure is all FFS beneficiaries who meet the Medicaid eligibility) and by market area, which signals following criteria: are ages 67 years and older at the both opportunities to improve quality performance within end of the measurement year, are alive at the end of the areas and the measure’s potential for comparing quality measurement year, are continuously enrolled in Part A across local health care markets. and Part B for the measurement year and the previous In the future, the Commission could develop a risk year with no months of MA enrollment, and have used no adjustment model to calculate FFS and ACO expected hospice services in the measurement year. For the 2016 PPA rates and compare market-area risk-adjusted PPAs. measurement year, the population of FFS beneficiaries The risk adjustment model would need to ensure that the who met those criteria was about 22.5 million nationwide. PPA measure primarily reflects an organization’s or area’s Beneficiaries with three or more discharges in the quality of care rather than underlying differences in patient measurement year were considered outliers and removed severity. Using the MA PPA measure as an example, from the qualifying population and observed event we can test risk adjustment using age, sex, and disease Applying the Commission’s principles for measuring quality: Population-based measures and hospital quality incentives 182

205 TABLE Distribution of PPAs per 1,000 FFS beneficiaries ages 67 7–2 and older varies by local health care market area, 2016 PPA rate per 1,000 FFS beneficiaries Acute conditions Chronic conditions Total National mean 15.3 17.7 33.1 11.1 22.4 10th percentile (highest performing) 10.4 17.8 16.2 50th (median) 34.9 90th (lowest performing) 24.3 48.7 24.9 2.3 Ratio of 90th to 10th percentile 2.2 2.2 Note: PPA (potentially preventable admission), FFS (fee-for-service). To evaluate the utility of measuring PPAs for FFS beneficiaries, we calculated the observed (not risk- adjusted) rates of admissions tied to acute (e.g., pneumonia), chronic (e.g., diabetes) and total (acute plus chronic) conditions. Rates presented are the number of PPAs divided by the number of beneficiaries in the qualifying population, multiplied by 1,000. The qualifying population is the same across the acute and chronic categories. Beneficiaries who died in the measurement year are excluded. There are over 1,200 MedPAC-defined market areas designed to match insurance markets served by private plans. The average qualifying population in each market area is about 19,000 beneficiaries. Source: MedPAC analysis of 2016 fee-for-service Medicare claims data. counts. Almost 57,000 outlier beneficiaries were removed beneficiaries had higher PPA rates than men for acute from our measure calculation (about 75 percent had conditions and about the same PPA rate as men for chronic hospitalizations tied to chronic conditions vs. acute conditions. Both fully (i.e., receive full range of Medicaid primary diagnoses). benefits) and partially (i.e., Medicaid pays Medicare premium and may also pay the cost sharing for Medicare As with the MA PPAs by chronic and acute conditions services) dual-eligible beneficiaries had higher PPA rates plan PPA measure, we calculated the number of inpatient for both acute and chronic conditions compared with non- admissions and observation stays tied to the beneficiaries dual-eligible (Medicare-only) beneficiaries. These patterns in the qualifying population (both are observed events). are consistent with CMS-produced results using selected The observed events include admissions with the primary AHRQ PQIs and with our prior work using the 3M PPA chronic conditions: diabetes diagnosis of the following measure. (short-term and long-term complications, uncontrolled The pattern of higher PPA rates for the dual-eligible diabetes, lower extremity amputation among patients with population is also expected when comparing admission diabetes); chronic obstructive pulmonary disease (COPD); rates that are not risk adjusted for population asthma; hypertension; and heart failure. Observed events characteristics. For example, the fully dual-eligible also include admissions tied to beneficiaries with the population is older than the partially dual-eligible acute conditions: primary diagnosis of the following population, which may explain the fully dual-eligible bacterial pneumonia, urinary tract infection, cellulitis, and population’s higher rate of PPAs. In future analyses of pressure ulcers. We calculated a total number of PPAs PPA rates, we will consider the effect of dual eligibility on (chronic plus acute). the PPA results. In 2016, PPAs accounted for about 8 National PPAs results 6 percent of FFS Medicare hospital admissions. Nationally, Distribution of PPAs in local health care market areas there were 15.3 acute-condition-related PPAs per 1,000 Differences in PPA results across local health care markets can help distinguish differences in quality compared with FFS beneficiaries and 17.7 chronic-condition-related PPAs a national mean and across market areas. We calculated per 1,000 FFS beneficiaries, for a total of 33.1 PPAs per 1,000 FFS beneficiaries (Table 7-1). PPA rates for acute and chronic conditions and total PPAs per 1,000 FFS beneficiaries in the 1,200 MedPAC In 2016, older Medicare beneficiaries had higher PPA market areas that the Commission recommends for MA rates for both acute and chronic conditions. Female payment and quality reporting (Table 7-2). We found that June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 183

206 total observed (not risk-adjusted) PPA rates varied across Commission staff worked with a team from the Harvard market areas, with the market area in the 90th percentile School of Public Health to develop a prototype HCD of PPA rates having a rate that was 2.2 times the market measure. As described in the June 2015 report to the area in the10th percentile. The magnitude of difference Congress, an HCD measure using Medicare claims data between the market areas in the 90th and 10th percentiles may be a meaningful gauge for comparing differences in of observed PPA rates for acute conditions and chronic health status across populations and be less complicated conditions individually was similar to that for the total than other measures for beneficiaries, policymakers, and PPA rate. other stakeholders to understand (Medicare Payment Advisory Commission 2015a). To model rates at a more narrowly defined health care market level (that is, the Dartmouth-defined hospital CMS is actively developing a quality measure for service areas (HSAs)), we calculated PPA rates for acute Medicare and Medicaid health plans and long- and chronic conditions and total PPA rates per 1,000 term services and support populations based on the FFS beneficiaries in the roughly 3,400 HSAs. An HSA Commission’s HCD measure. CMS may submit the is a collection of zip codes that represents a local market measure, currently named “days in the community,” area whose residents receive most of their inpatient for endorsement by the National Quality Forum. Also, care from the hospitals in that area. As with the larger in 2016, the National Bureau of Economic Research MedPAC market areas presented in Table 7-2, PPA rates released a working paper, Healthy-time Measures of varied across HSAs, with HSAs in the 90th percentile of , that describes Health Outcomes and Healthcare Quality PPA rates exceeding HSAs in the 10th percentile of PPA some conceptual and empirical foundations of “healthy- rates by 2.1 times (data not shown). PPA rates for acute time” measures of health care quality. Their analysis conditions had slightly more variation compared with PPA features the Commission’s developing HCD measure and 8 rates for chronic conditions. similar measures from other organizations. The authors concluded that “the basic premises underlying this [the Home and community days measure Commission’s] measure’s definition are conceptually sound and intuitively appealing; its use as a patient- The Commission tested a “home and community days” centered outcome or care-quality indicator holds promise” (HCDs) quality measure to assess how well health care (Burns and Mullahy 2016). organizations keep people healthy and out of health care 7 institutions. We chose to focus on the number of days Calculating home and community days per year that beneficiaries did not receive institutionalized The Commission’s HCD measure, for the purposes of this medical care (such as days during which a beneficiary chapter, pertains to FFS Medicare beneficiaries 65 years and did not have an inpatient stay) and mortality days. An older, excluding those enrolled in MA for any part of the alternative to the measure could include days in which year and those not enrolled in Medicare FFS continuously beneficiaries had any interaction with the health system throughout the year. For the HCD measure we modeled, (i.e., days in which Medicare covered any medically we focused on beneficiary interactions with more serious necessary service such as a physician office visit or an health care that is covered by Medicare and on mortality. We inpatient stay (Medicare Part A and Part B)). defined this measure algorithmically as follows: High-quality MA plans and ACOs are designed to manage HCDs = 365 days – (days in short-term acute care beneficiary, hospital, and physician relations to coordinate hospital + days in inpatient rehabilitation facility + care and provide appropriate access to keep people days in long-term care hospital + days in inpatient out of health care institutions. For example, ACOs can psychiatric facility + days in skilled nursing facility provide tools and data to physicians about patients with + days in observation status + days of emergency ambulatory care–sensitive conditions (such as diabetes and department use + mortality days) asthma) so that they can appropriately monitor, coordinate, and follow up with patients and reduce inpatient stays. For each FFS beneficiary, we calculated his or her total FFS clinicians can also play a role in affecting HCDs in number of mortality days, which is defined by the number their ambulatory care area by effectively coordinating of days remaining in the calendar year after the date of with other providers and offering adequate access to death. For example, a beneficiary who did not die during beneficiaries. Applying the Commission’s principles for measuring quality: Population-based measures and hospital quality incentives 184

207 the year would have no mortality days. A beneficiary who on HCDs were respiratory arrest, nephritis, extensive died on December 28 would have three mortality days for third-degree burns, seizure disorders and convulsions, and the year. Inpatient, observation, skilled nursing facility coma/brain compression/anoxic damage (all statistically (SNF), inpatient psychiatry, inpatient rehabilitation, and significant). Our analysis found that HCDs decrease with long-term care hospital days were defined as the total age. Men had slightly more HCDs than women. number of days per year the beneficiary spent in each of Effect of dual-eligibility status We also tested the effects these respective settings. For the purposes of this analysis, that social risk factors could have on the risk adjustment we weighted HCD components equally, but policymakers model. In a separate regression model, we included race interested in developing this measure further could and dual eligibility (defined by a beneficiary having both give the components different weights based on some Medicare and Medicaid coverage for at least one month of prioritization that takes into account interests shared by the a year). (Dual eligibility may be a proxy for low income.) Medicare program and its beneficiaries. When included as a variable, dual-eligibility status had We did not subtract home health visit days in calculating some impact on HCDs (regression coefficient = –7.76) a beneficiary’s HCDs. Home health represents a midpoint (i.e., dual-eligible status corresponds with fewer HCDs). at which the patient is at home but is still in need of Coefficients for race were not significant. When we 2 health care services. In some health care markets, home R compared the explanatory power ( ) of a risk adjustment health visits are used to prevent or limit use of other, more model with age, sex, disease burden, and market-fixed expensive services—in particular, inpatient and SNF care. effects with a model that included those variables plus Subtracting home health visit days from the HCD measure dual-eligibility status, there was no difference in the 2 could therefore penalize these markets and providers explanatory power of the models (both R = 0.32). unfairly. Documented overuse of home health care could Since dual eligibility seemed to have some impact for make a case for subtracting home health visits from individual beneficiaries but not on the overall model’s the HCD measure. For instance, the Office of Inspector explanatory power, we investigated the impact of dual- General (OIG) has recently identified 27 geographic 10 eligibility on market-area performance. We examined areas as “hotspots” for characteristics commonly found how market performance varies among high-share versus in OIG-investigated cases of home health fraud, so, in low-share dual-eligible markets and found that mean some markets, penalizing home health use could be an HCDs decline with increasing deciles of dual-eligible appropriate approach (Office of Inspector General 2016). share, although the relationship is not constant (Figure Yet even with these potential differences in home health 7-1, p. 186). Among all beneficiaries ages 65 and older, use by market area, from the beneficiary’s perspective, markets in the top decile of dual-eligible share—in which home health visits are likely more desirable than the use more than 37 percent of beneficiaries were Medicaid of other health care services that would lower HCDs, a eligible—had, on average, about 4 fewer HCDs compared circumstance that argues for not subtracting home health with markets in the bottom decile of dual-eligible share visit days from the HCD measure. (in which over 9 percent of beneficiaries were dual eligible). Among beneficiaries ages 65 and older with 2 Risk adjustment modeling or more chronic conditions, markets in the top decile of A critical step in the development of the HCD measure is dual-eligible share had, on average, about 6 fewer HCDs to test appropriate risk adjustment models. Such models compared with markets in the bottom decile of dual- should ensure that the HCD measure primarily reflects 11 eligible shares. an organization’s or area’s quality of care rather than underlying differences in patient severity. Using linear We found that mortality days tended to be somewhat regression, we developed a model that included variables higher in markets with high dual-eligibility shares, readily available in FFS claims data and used in other resulting in a lower average number of HCDs, although quality measures: age, sex, disease burden determined the differences were small. Inpatient and SNF days were from HCCs, and market-fixed effects (e.g., local stable across the deciles of markets. 9 characteristics). If CMS opted to use HCDs to compare quality across We found that disease burden had the greatest impact on market areas or providers, the Secretary should be HCDs. The diseases or conditions that had the most effect June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 185

208 FIGURE FIGURE Cumulative change... X-X Mean adjusted home and community days are slightly lower in local health care 7–1 market areas with higher shares of dual-eligible beneficiaries, 2015 350 Ages 65 and older (all) Ages 65 and older with 2 or more chronic conditions 345 340 335 330 Mean adjusted market-area HCD days 325 320 315 9 1 8 10 7 3 4 5 6 2 Deciles of markets by dual-eligible share (lowest share to highest share) Note: Home and community days (HCDs) are adjusted for age, sex, and disease burden. There are over 1,200 MedPAC-defined market areas, designed to match insurance markets served by private plans. Deciles of markets were created based on the share of Medicare beneficiaries who were partially dual eligible for Medicare and Medicaid at any point in the year. Partially dual-eligible beneficiaries qualify for Medicaid payment of the Medicare premium and perhaps the cost sharing for Medicare services. Markets in Decile 1 have the lowest share of partially dual-eligible beneficiaries (9.4 percent), while markets in Decile 10 have the highest share of Medicaid-eligible beneficiaries (37.3 percent). The scale of the -axis was chosen to highlight the differences in HCDs across market areas. y Source: Analysis of fee-for-service Medicare claims data, 2015. cognizant of differences that correlate with dual eligibility. defined market areas designed to match insurance markets However, the Commission does not support the inclusion served by private plans and Dartmouth-defined HSAs, of dual-eligibility status in a risk adjustment model which are collections of zip codes that represent a local because doing so would mask disparities in clinical market area whose residents receive most of their inpatient performance. Rather, Medicare should account for social care from the hospitals in that area). We calculated risk factors by directly adjusting payment through peer HCDs in each MedPAC-defined market area and HSA grouping. The Commission continues to encourage using 3 years of FFS Medicare data (2013 to 2015) for 2 the Secretary to support research and data collection to populations: (1) all beneficiaries 65 years and older and improve Medicare’s ability to account for the effect of (2) beneficiaries 65 years and older with at least 2 chronic social risk factors on health outcomes. conditions. There were at that time about 27.3 million beneficiaries 65 years and older, and about 7.7 million of Adjusted HCDs in local health care market areas those had at least two chronic conditions. To understand HCDs for Medicare beneficiaries of As expected, we found that Medicare beneficiaries with different health status in different market areas over time, greater chronic-condition burden had fewer HCDs (Table we calculated mean, risk-adjusted HCDs for the two Note: Note and Source are in InDesign. 7-3). In 2015, the adjusted HCD rate for beneficiaries 65 different geographic area levels representing local health Source: years and older was 351 days compared with 328 days for care markets that we used in the PPA analysis (MedPAC- beneficiaries 65 years and older with 2 or more chronic Applying the Commission’s principles for measuring quality: Population-based measures and hospital quality incentives 186

209 TABLE Mean adjusted home and community days for FFS 7–3 beneficiaries were stable from 2013 to 2015 Home and community days 2013 2015 2014 351 All beneficiaries 65 years and older 351 351 Beneficiaries 65 years and older with 2 or more chronic conditions 328 332 331 Note: FFS (fee-for-service). Home and community days are adjusted for age, sex, disease burden, and market-fixed effects. There are over 1,200 MedPAC-defined market areas, which are designed to match insurance markets served by private plans. Source: Analysis of FFS Medicare claims data, 2013–2015. conditions (a difference of 23 days). From 2013 to 2015, days; SNF days slightly increased over the three years 12 the results for beneficiaries 65 years and older were stable (from 6.2 days to 6.6 days). (351 days in each year), but the average HCDs declined Distribution of adjusted HCDs in local health care slightly over the three years for beneficiaries with 2 or market areas more chronic conditions (from 331 to 328 days). Because our goal was to compare FFS quality across For both population groups, the components of the HCD health care markets and across different populations, we algorithm with the biggest impact on a market area’s looked for variation in HCD measure results across both HCDs were mortality days, SNF days, and inpatient days MedPAC-defined market areas and HSAs. We calculated (Table 7-4). For beneficiaries 65 years and older, the the distribution of HCDs for all beneficiaries 65 years components were stable over time. There was somewhat and older and for beneficiaries 65 years and older with more change from 2013 to 2015 in the HCD components 2 or more chronic conditions across MedPAC-defined for the beneficiaries 65 years and older with 2 or more market areas (Table 7-5, p. 188). The distribution among chronic conditions. In the 2013 to 2015 period, the MedPAC-defined market-area HCDs for both populations mortality days for that population increased by about 2.3 was very small. The difference between the 90th and TABLE Home and community day components were stable from 2013 to 2015 7–4 Beneficiaries 65 years and older All beneficiaries 65 years and older with 2 or more chronic conditions Component (days) 2015 2013 2014 2015 2013 2014 9.8 9.4 Mortality 23.7 9.7 21.4 20.5 6.6 2.1 2.1 2.0 6.2 6.0 Skilled nursing facility 4.8 Inpatient 1.6 1.5 1.5 4.7 4.5 Components are part of the home and community days (HCDs) calculation and represent mortality days and/or days in which beneficiaries have interactions with Note: more serious health care. Mortality, skilled nursing facility, and inpatient days have the biggest impact on a market area’s HCDs. HCDs are adjusted for age, sex, disease burden, and market-fixed effects. There are 1,200 MedPAC-defined market areas designed to match insurance markets served by private plans. Source: Analysis of fee-for-service Medicare claims data, 2013–2015. June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 187

210 TABLE Distribution of home and community days did not 7–5 vary across local health care market areas, 2015 Home and community days All beneficiaries Beneficiaries 65 years and older 65 years and older with 2 or more chronic conditions National mean 351 328 10th percentile (lowest performing) 349 323 50th (median) 351 328 332 352 90th (highest performing) 1.03 Ratio of 90th to 10th percentile 1.01 Note: Home and community days are adjusted for age, sex, disease burden, and market-fixed effects. There are over 1,200 MedPAC-defined market areas designed to match insurance markets served by private plans. Source: Analysis of fee-for-service Medicare claims data, 2015. 10th percentile MedPAC-defined market areas (of 1,200 Mean time between failure— “Mean time between • MedPAC-defined market areas) for beneficiaries 65 years failure” is a commonly used engineering measure of and older was only 3 days. The difference between the predicted elapsed time between inherent failures of a 90th and 10th percentile MedPAC-defined market areas mechanical or electronic system during normal system for beneficiaries 65 years and older with 2 or more chronic operation. Policymakers could consider how to apply conditions was only 9 days. For both populations, the this concept to measure quality of care for Medicare highest performing MedPAC-defined market area’s HCDs beneficiaries (for example, how many days between were always almost equal that of the lowest performing serious health care interactions (e.g., mean time MedPAC-defined market area’s HCDs. We found similar between hospitalizations) for Medicare beneficiaries). results when calculating HCDs for the more narrowly Successful community discharge— • The Improving defined HSAs (data not shown). Medicare Post-Acute Care Transformation Act of With so little variation across local market areas and 2014 mandated that CMS develop quality measures the challenges posed by the need to develop appropriate for PAC providers. Responding to this mandate, weights for constructing the composite measure, the CMS has developed measures for each PAC setting Commission questions the immediate utility of the HCD that assess whether PAC providers successfully measure in its current form to assess market-level FFS discharge beneficiaries to the community (e.g., rate performance. of beneficiaries discharged to the community who do not have an unplanned admission to a hospital within Future work on population-based quality a set period of time). The Commission also currently measures calculates rates of discharge to the community for some individual PAC settings. Policymakers could The Commission and policymakers may explore the consider measuring successful community discharge following claims-calculated, population-based measures to across all PAC providers for different populations. assess Medicare quality for different defined populations (e.g., FFS populations associated with local market areas Home-to-home transition time— • Home-to-home and beneficiaries served by MA plans, ACOs, hospitals, transition time is a measure that adds time spent in a post-acute care (PAC) providers, or groups of clinicians). PAC facility to time spent in the hospital to capture These measures are in line with the Commission’s the full span of a hospitalization episode (Barnett et al. quality measurement principles to use population-based outcome measures that are patient-oriented, encourage coordination, and promote delivery system change. Applying the Commission’s principles for measuring quality: Population-based measures and hospital quality incentives 188

211 2017). This measure is patient centered since patients Issues with current hospital quality and are interested in when they can return home from all value programs institutional care. Policymakers could explore the use The Commission has four main concerns about the design of this measure to assess the home-to-home transition of the current hospital quality programs. The first is that too times for different populations. many overlapping hospital quality payment and reporting programs create unneeded complexity for hospitals End-of-life care and burdensome transitions— • and the Medicare program itself (Medicare Payment Research has shown that a growing number of Advisory Commission 2016a, Medicare Payment Advisory older adults in the United States are dying at Commission 2016b). Some of the quality measures are home, but many continue to face multiple health scored in multiple programs. For fiscal years (FYs) 2020 and care transitions to different care sites and receive 2021, CMS has proposed to remove much of the duplication aggressive inpatient care in their final days (Teno et in quality measures across programs. For example, CMS al. 2013). Policymakers can consider developing a would continue to use the hospital-acquired infection quality measure that assesses potentially burdensome measures to assess performance in the Hospital-Acquired transitions in the last days or weeks of life. Conditions Reduction Program (HACRP) but would remove these measures from the Inpatient Quality Reporting Program • Low-value care— For several years, the Commission (IQRP) and Hospital Value-based Purchasing (VBP) Program has expressed concern that beneficiaries are receiving (Centers for Medicare & Medicaid Services 2018b). low-value care, or care that has little or no clinical benefit and that can potentially harm them (Medicare Second, the Commission believes that all-condition Payment Advisory Commission 2017, Medicare mortality and readmissions measures are more appropriate Payment Advisory Commission 2015b) (see also to measure hospitals’ performance, rather than the Chapter 10 in this report). The Commission has condition-specific (e.g., acute myocardial infarction) examined national FFS population rates for certain measures that are scored in the IQRP, VBP Program, and services and procedures that are considered low value. Hospital Readmissions Reduction Program (HRRP). Using Policymakers should continue to explore measures of all-condition measures would increase the number of low-value care for different populations. observations and reduce the random variation that single- condition readmission rates face under current policy (Medicare Payment Advisory Commission 2013). Applying the Commission’s principles Third, the IQRP includes process measures that are not for measuring quality to hospital quality tied to outcomes and are burdensome to report (e.g., incentives fibrinolytic therapy received within 30 minutes of hospital arrival). Also, providers may not be consistent in how they The Commission contends that Medicare payments should report some of the measures included in the IQRP, VBP not be made without consideration of the quality of care Program , and HACRP (e.g., chart-abstracted measures delivered to beneficiaries and has recently formalized a and hospital-acquired infections). For FYs 2020 and set of principles for quality measurement in the Medicare 2021, CMS has proposed removing some chart-abstracted program. For several years, the Medicare program has process measures, such as median time from emergency provided hospitals with incentive payments based on department (ED) arrival to ED departure for admitted ED the quality of care they give to FFS beneficiaries (see patients, from the IQRP because the data collection and text box on current hospital quality and value payment reporting costs outweigh the benefit of their continued use programs, pp. 190–191). The quality of hospital care has (Centers for Medicare & Medicaid Services 2018b). been improving over the years, which is partly due to these programs. However, the hospital industry has raised Fourth, the VBP Program, HRRP, and HACRP score hospitals concerns that the designs of these programs are complex, using “tournament models” (i.e., providers are scored relative overlap, and send different performance signals to to one another), not on clear, absolute, and prospectively set hospitals. In addition, aspects of the programs do not align performance targets. For example, the HACRP’s statutory with the Commission’s principles for measuring quality in design penalizes 25 percent of hospitals every year, even the Medicare program. if all hospitals significantly reduce their HAC rates. The Commission’s principles for quality measurement encourage Medicare quality programs to use fixed targets. June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 189

212 Current hospital quality and value payment programs he Medicare program adjusts hospital payment Incentives for higher quality based on four quality payment programs. Three programs adjust hospital payment based on One program adjusts payment based on T how the hospital performs on quality results: the whether a hospital reports quality measure results, Hospital Readmissions Reduction Program (HRRP), and three programs adjust payment based on quality the Hospital-Acquired Conditions Reduction Program performance. Although not tied to payment, CMS’s (HACRP), and the Hospital Value-based Purchasing public reporting of hospital quality performance on the (VBP) Program. Hospital Compare website, including their star ratings, Hospital Readmissions Reduction Program The HRRP is another avenue for comparing acute care hospitals. 13 was implemented in fiscal year 2013. As a part of this program, hospitals that have excess Medicare Reporting readmissions over a three-year period for selected The Hospital Inpatient Quality Reporting Program conditions have their inpatient prospective payment (IQRP) has been in place since fiscal year 2009 and system (IPPS) payments reduced. In fiscal year 2018, is built on the preceding voluntary Hospital Quality the readmissions policy applies to six conditions (acute Initiative. The IQRP reduces a hospital’s annual market myocardial infarction (AMI), heart failure, pneumonia, basket (the measure of inflation in costs of goods and chronic obstructive pulmonary disease, total hip and services used by hospitals in treating Medicare patients) knee arthroplasty, and coronary artery bypass graft update by 2.0 percent if it does not successfully report surgery). In 2018, the payment penalty is capped at quality measure data. In fiscal year 2018, nearly all 3 percent of a hospital’s base diagnosis related group inpatient hospitals met the IQRP requirement and (DRG) payments per year. In 2018, about 80 percent of will receive the full annual market basket update. hospitals will have payments reduced because of higher There are 61 quality measures in the fiscal year 2020 than average readmissions for at least one condition. program (based on coverage year 2018 performance). Total penalties will be about $556 million in 2018, Hospitals report about half of those measures to or 0.5 percent of Medicare’s total IPPS payments. CMS (e.g., patient experience surveys, health care– Research has shown that readmission rates for AMI, associated infections, medical record–abstracted heart failure, and pneumonia decreased more rapidly measures such as average number of minutes before after the HRRP began and that improvement was outpatients with chest pain or possible heart attack got most marked for hospitals with the lowest pre-HRRP an electrocardiogram, use of safe surgery checklist), performance (Wasfy et al. 2017). while the other half are claims-based outcome (e.g., readmissions) or cost measures that CMS calculates. Hospital-Acquired Conditions Reduction Program The (CMS has proposed to remove 39 measures from the HACRP was effective beginning in fiscal year 2015. IQRP for fiscal years 2020 and 2021 (Centers for Hospitals are ranked on their total rate of preventable Medicare & Medicaid Services 2018b).) conditions in two categories: (1) claims-calculated patient safety indicators such as pressure ulcer and sepsis (continued next page) program (HVIP) that would be patient oriented, encourage Redesigning Medicare’s hospital quality and coordination across providers and time, and promote value payment programs change in the delivery system. Since current hospital There is an opportunity to redesign Medicare’s hospital quality programs are defined in statute, the Congress quality payment programs as one hospital value incentive Applying the Commission’s principles for measuring quality: Population-based measures and hospital quality incentives 190

213 Current hospital quality and value payment programs (cont.) (AHRQ’s patient safety indicator (PSI) 90) and rates and (2) hospital-reported health care–associated data on six health care–associated infections; (3) infections such as surgical site infections and catheter- 25 percent is based on efficiency, using a 30-day associated urinary tract infections. The 25 percent of Medicare spending per beneficiary measure; and (4) hospitals with the highest rates of preventable conditions 25 percent is based on clinical care, tied to 30-day (poorest performers) receive a 1 percent reduction to mortality for three conditions—AMI, heart failure, and all inpatient payments. In 2017, the HACRP reduced 14,15 pneumonia). payment to 742 hospitals, with penalties totaling roughly The VBP Program gives a hospital $370 million (Medicare Payment Advisory Commission credit for achievement (relative to other hospitals) 2017). Before the start of the HACRP, hospitals had and improvement (relative to its own baseline been successful in reducing the number of hospital- performance). acquired conditions (HACs). An Agency for Healthcare Public reporting of quality performance Research and Quality (AHRQ) study reported that, from 2010 to 2015, HACs per discharge declined by Although not tied directly to payment, Medicare reports 21 percent, an estimated 125,000 fewer patients died in certain quality results to consumers and providers on the hospital as a result of the reduction in HACs, and CMS’s Hospital Compare website. The website shows a an estimated $28 billion in health care costs was saved hospital’s results for given measure categories alongside (Agency for Healthcare Research and Quality 2016). the state and national averages for the measure. The displayed measures are from the IQRP, HRRP, HACRP, The Hospital Value-based Purchasing Program and VBP programs as well as results from hospital Hospital VBP Program was implemented in fiscal outpatient facilities (e.g., imaging efficiency). The year 2013. As required by law, the program is budget measure categories include (1) survey of patient’s neutral; that is, the total pool of withheld payments experiences; (2) timely and effective care (i.e., cataract (currently equal to 2 percent of base inpatient DRG surgery care, heart attack care, emergency department payments) must be redistributed to hospitals based care); (3) complications and deaths (e.g., health care– on their performance on the VBP Program’s quality associated infections); (4) hospital returns; (5) use of measures. In 2018, the VBP Program increases medical imaging; and (6) payment and value of care payments to about 50 percent of IPPS hospitals and (e.g., Medicare spending per beneficiary). The Hospital decreases payments to 42 percent of them. Hospitals Compare website also presents a summary star rating earn back anywhere from 17 percent to 200 percent of (up to 5 stars) for the patient experience category and their withheld payments. For roughly a third of these another star rating that combines individual clinical, hospitals, the change in payments under the program patient experience, and efficiency measures from the was small, less than 0.25 percent of base payments. VBP Program, HRRP, and the Hospital Compare website. The Commission has commented to CMS The program uses a combination of measures from that the overall star rating system creates unneeded four quality domains to score hospitals on quality complexity in the Medicare program because it creates (the measures are also part of the IQRP): (1) 25 a new system of measures and scoring methodology for percent of the score is based on patient experience CMS to administer and for hospitals to track (Medicare of care surveys; (2) 25 percent is based on patient Payment Advisory Commission 2016a). ■ safety, using a composite patient safety measure 16 would need to create the new HVIP and eliminate the action (e.g., improving public reporting.) The HVIP current programs in legislation. We believe that CMS is intended to replace quality programs that affect FFS has the authority to make some of our suggested changes hospital payment. However, in line with the Commission’s to hospital quality payment without congressional principles, the HVIP measures and scoring methodology June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 191

214 Under an HVIP, the Medicare program would continue should align across Medicare accountable entities and to provide hospitals with quality feedback reports to help providers, including hospitals. MA plans, ACOs, and them understand their performance on the claims-based hospitals should be held accountable to a small set of population-based measures, scored against absolute measures. Reports could include benchmark and other comparative information so that hospitals could take thresholds, and have their payments adjusted through peer action to improve their results. Even though an HVIP grouping. Medicare’s use of the same set of measures and scoring framework across different populations could also would score all-condition measures, CMS could consider providing hospitals with condition-specific results (e.g., promote multipayer alignment. acute myocardial infarction mortality) calculated by Design claims data for hospitals to use for their own quality improvement. The Medicare program should not pay hospitals and other providers for reporting quality measures, but should pay Measures based on performance on these measures. Virtually all hospitals currently meet the IQRP reporting requirements Based on our quality measurement principles, we and receive their full payment update, arguing for the propose an HVIP that would include four largely CMS- need to retire the IQRP. The Congress could also consider calculated or CMS-administered quality measures: removing payment incentives tied to Medicare quality mortality, readmissions, Medicare spending per reporting programs in other sectors where pay-for- beneficiary (MSPB), and patients’ overall rating of the 17 performance programs have been implemented (e.g., hospital. These risk-adjusted measures are included skilled nursing facilities). in the existing hospital quality programs and thus are known to hospitals. (We envision that, as risk adjustment For simplicity, hospitals should have their payment models evolve, they will be incorporated into the adjusted based on performance on quality and cost HVIP measures.) Providers could choose to use other measures in a single program instead of three separate granular quality measures to manage their own quality programs. The HRRP and VBP programs should be improvement, but these would not factor into Medicare combined into one HVIP. The HACRP, which scores payment. patient safety measures such as infection rates, should also be retired as a hospital payment adjustment (see p. Hospital readmission, for any reason, is Readmissions 194 for more discussion of patient safety). disruptive to patients and caregivers and costly to the health care system, and it puts patients at additional Like the VBP Program, an HVIP would translate quality risk of hospital-acquired infections and complications. measure performance to payment and redistribute a Readmissions are also a major source of patient and budgeted amount to hospitals based on their performance. family stress and may contribute substantially to loss We would expect the new program to be budget neutral of functional ability, particularly in older patients. to the HRRP and HACRP, which, based on our analysis, Measuring and adjusting payments based on a hospital’s reduce Medicare payment by 0.5 percent. readmission rates holds the hospital accountable for ensuring that beneficiaries have the discharge information Public reporting of quality results can drive quality they need and encourages hospitals to coordinate with improvement by fostering competition among providers other providers. Since the implementation of the HRRP, and allowing providers to better identify opportunities hospitals have taken action and improved readmission for improvement. We believe that CMS could incorporate rates. The readmission measure is also important to and an HVIP into the public reporting of quality results understandable by the beneficiary and can be calculated on Hospital Compare or other websites. CMS could through claims data. report results as a consumer-friendly summary quality score (e.g., a star rating). For beneficiaries interested In our HVIP model, we scored hospitals on their in more detailed quality results, CMS could also unplanned, risk-adjusted rates of readmissions within report all available patient experience measures (e.g., 30 days of discharge for all conditions using Medicare communication, cleanliness), some condition-specific claims. Using an all-cause readmission measure (rather outcomes (e.g., pneumonia readmissions, heart failure than the six conditions used in the HRRP) increases the mortality), and HAC results. number of observations and reduces random variation. Applying the Commission’s principles for measuring quality: Population-based measures and hospital quality incentives 192

215 Our model also used three years of claims data (2014 (2) communication with doctors, (3) responsiveness of through 2016) to increase the number of observations. hospital staff, (4) communication about medicines, (5) cleanliness of hospital environment, (6) quietness of Mortality during or soon after a hospital stay Mortality hospital environment, (7) discharge information, (e.g., within 30 days) is an important outcome measure, (8) care transition, (9) overall rating, and (10) whether and it encourages hospitals to coordinate with post- the beneficiary would recommend the hospital to others. acute care providers. Like the readmission measure, (Hospitals can add their own survey items to the core this outcome measure can be determined with a high survey.) The HCAHPS measures are scored in the VBP degree of accuracy through claims. As suggested with Program; they are publicly reported on Hospital Compare the readmissions measure, an all-condition mortality and as part of the star rating system. measure would hold hospitals more accountable than condition-specific measures. Our HVIP model used an Based on the Commission’s principles, a new HVIP all-condition, risk-adjusted measure of mortality during ideally includes population-based patient experience the hospital stay and 30 days after discharge, and we used measures. High-quality hospitals and physicians appear three years of data (2014 to 2016) to increase the number to focus not only on technical excellence but also on how of observations. (The measure excludes patients who are patients perceive their care (Chatterjee et al. 2015). When in hospice care before admission.) patients have a better experience, they are more likely to adhere to treatments, return for follow-up appointments, Medicare spending per beneficiary MSPB is a claims- and engage with the health care system by seeking based value measure that we propose be included in an appropriate care (Safran et al. 1998). HVIP. This measure rewards efficient, effective hospital For simplicity, we modeled the HVIP using a single care, not volume of services, and reduces delivery system overall hospital rating measure (i.e., share of patients who fragmentation. By pairing the spending measure with gave their hospital a rating of 9 or 10 on a scale from mortality and readmissions, hospitals have an incentive 0 (lowest) to 10 (highest)) instead of a combination of to maintain episode quality while reducing episode costs. the 10 HCAHPS measures. The overall hospital rating The measure shows whether Medicare spends more, measure is strongly or moderately related to the other less, or about the same per Medicare patient treated at quality measures (e.g., communication with nurses a specific hospital compared with how much Medicare correlation ( r ) = 0.64; care transition correlation ( r ) = spends on comparable patients nationally. Our model 0.48), so by scoring a hospital’s overall rating, we likely used the MSPB values CMS currently produces for capture the other measures (Centers for Medicare & the VBP Program, which are price-standardized, risk- Medicaid Services 2017). Also, a hospital’s performance adjusted episodes that include all Medicare Part A and on some of the other HCAHPS measures, such as Part B claims paid during the period from 3 days before discharge information and care transitions, would be an inpatient hospital admission through 30 days after detected in the readmissions, mortality, and MSPB discharge. The model used the MSPB values calculated measures. Alternatively, the HVIP could use a unique with three years of data (2014 to 2016). composite measure based on a subset of the HCAHPS Patients’ overall rating of the hospital The Hospital measures that are meaningful to both beneficiaries and Consumer Assessment of Healthcare Providers and providers such as measures of communication with ® ® Systems ) is a national standardized (HCAHPS nurses, communication with doctors, responsiveness of survey instrument and data collection methodology for staff, and discharge information. measuring patients’ perspectives on their care during 18 To be scored on the overall hospital quality rating a recent hospital stay. The survey allows Medicare, measure, hospitals would need to administer the entire hospitals, beneficiaries, and others to make objective core HCAHPS survey and would receive a score of zero and meaningful comparisons of hospitals. Since 2006, for that measure if they did not. Hospitals could continue CMS and hospitals have worked with third-party to monitor the other HCAHPS measures and use them to survey vendors to collect survey results from a random manage their own quality improvement. CMS could also sample of each hospital’s adult inpatient discharges. The continue to publicly report multiple HCAHPS measures survey results are used to calculate 10 core measures on Hospital Compare. of patient experience: (1) communication with nurses, June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 193

216 20 Our HVIP model adjusts a hospital’s Patient safety The use of the and accidental punctures or lacerations. payment based on its performance on four measures that PSI 90 measure in pay-for-performance programs has been criticized for several reasons, including surveillance are part of the existing hospital quality payment programs. bias (e.g., hospitals with higher rates of postoperative We also support a Medicare-influenced system to improve patient safety outside of an HVIP. But because of concerns blood clots were often the hospitals that were most with the accuracy of some patient safety data, we do not vigilant in screening patients for them) and concerns about the accuracy of this measure in identifying meaningful propose inclusion of patient safety measures in the HVIP unintentional cases of injury (Rajaram et al. 2015). AHRQ model at this time. Under the HVIP, hospitals should has recently updated the PSI 90 measure to address some continue to have incentives to improve patient safety because doing so could potentially affect performance of these concerns, and hospitals will begin to report on the on the four HVIP measures (e.g., readmissions due to revised measure this year. At this time, we do not propose to include the measure in the new payment program, but hospital-acquired infections). we will continue to monitor the measure’s performance. As part of the IQRP, HACRP, and VBP programs, Hospital-acquired conditions are an important measure hospitals are scored on five self-reported hospital care– acquired infection rates, such as catheter-associated of patient safety, but since the only way currently to monitor a hospital’s infection rate is through self-reported urinary tract infections. Hospitals use their own claims information, we propose that the current measures of and medical records to report their infection rates through 19 the CDC’s National Health Safety Network (NHSN). infection rates not be part of a new HVIP. Rather, we suggest that hospitals be required as a Medicare condition The NHSN provides hospitals, states, and regions with of participation (COP) to report accurate infection rates comparative data needed to identify problem areas and to the NHSN and that hospitals continue to work with measure local and national progress on prevention efforts. the CDC to monitor and evaluate opportunities to lower The monitoring and evaluation of infection rates through infection rates. (CMS could exempt small and rural Medicare’s programs and other national initiatives such as hospitals that may not have sufficient patient numbers to the Partnership for Patients have improved infection rates. warrant reporting to the NHSN.) This requirement can be Over the years, there have been anecdotal reports of some built into the existing infection control COPs requiring hospitals’ intentional misreporting of infection data—for hospitals to have a designated infection control officer, example, clinicians ordering diagnostic tests in the absence a hospital-wide quality assessment and performance of clinical symptoms to potentially identify infections improvement program, and training programs to address present on admission so they are not considered hospital problems identified by the infection control officer acquired (Centers for Disease Control and Prevention (Centers for Medicare & Medicaid Services 2012). The 2016a). The CDC and CMS have reported that there is no Secretary should continue to publicly report infection rates evidence such behaviors are widespread and have released (currently found on Hospital Compare) and investigate guidance on the importance of adherence to the NHSN providers with high rates. Consistent with our principles, protocol, definitions, and criteria to ensure the reliability we also encourage CMS to support research and data and comparability of the data. However, there are concerns collection to improve patient safety measures for potential that some hospitals are better than others at reporting inclusion in the HVIP. infections and other patient safety issues (Calderwood et al. 2017). Also, even though there are specific definitions Scoring methodology and criteria to capture the infection data, hospital infection Scoring under an HVIP should provide incentives for control specialists have to make judgment calls about how hospitals to improve the quality and efficiency of their to catalog infections, which makes part of the reporting care. To maintain the independence and importance of subjective. each of the four measures, our model treats each measure as an equally weighted, separate domain, consistent with The IQRP, HACRP, and VBP programs also include a the VBP Program methodology. Each of the 4 measures claims-based composite measure of 10 underlying patient is worth 10 points for a total of 40 possible HVIP points. safety indicators (PSIs), PSI 90, which signals potential in- This model is illustrative; policymakers could give the hospital complications and adverse events and procedures, components different weights based on the priorities of the including pressure ulcers, iatrogenic pneumonia, 21 Medicare program and its beneficiaries. postoperative sepsis, postoperative pulmonary embolism, Applying the Commission’s principles for measuring quality: Population-based measures and hospital quality incentives 194

217 TABLE Illustration of point system to score performance on 7–6 measures under our potential HVIP model Relative Patients’ Risk-adjusted Risk-adjusted Medicare spending overall rating mortality rates readmissions rates per beneficiary of hospital (lower is better) (lower is better) (lower than 1 is better) (higher is better) 20% or above 15% or above 53% or below 1.16 or above 0 points 18% 13% 1.09 60% 2 points 4 points 16% 11% 1.02 67% 6 points 14% 9% 0.95 73% 8 points 12% 7% 0.88 80% 10 points 10% or below 5% or below 0.82 or below 87% or above Note: HVIP (hospital value incentive program). Each measure in the HVIP is continuously scored from 0 to 10 points, and only a subset of points is displayed here. Lower rates are better for readmissions, mortality, and Medicare spending per beneficiary (MSPB), and they receive more HVIP points. The MSPB value is based on the hospital’s spending compared with the national mean. “Patients’ overall rating of hospital” is the share of Hospital Consumer Assessment of Healthcare Providers ® survey respondents who gave the hospital an overall rating of 9 or 10 on a 10-point scale. and Systems Source: MedPAC analysis of Medicare fee-for-service hospital quality data, 2014–2016. Converting measure performance to HVIP points (score) targets (or “gates”) they needed to reach to achieve a One of the Commission’s principles is that Medicare certain point level for each measure. Table 7-6 presents a quality programs should reward providers based on subset of the scale of points associated with performance clear, absolute, and prospectively set performance targets targets in our HVIP model. rather than score providers relative to one another. Following is an example of converting measure Prospective targets allow providers to know in advance performance to points using the continuous performance- what outcomes they must achieve to avoid penalties and to-points scale highlighted in Table 7-6: Hospital A has achieve rewards; they also allow the industry as a whole to a risk-adjusted readmissions rate of 15 percent (earns 5 be rewarded if all providers improve. In addition, rewards points), risk-adjusted mortality rate of 7 percent (earns 8 should be distributed based on a continuous scale (i.e., points), Medicare spending per beneficiary value of 0.96 without payment “cliffs”), so that hospitals with similar (earns 5.9 points), and overall patient experience rating of performance will receive similar financial rewards. In 79 percent (earns 7.8 points). Hospital A receives a total of our example, hospitals earn points for their performance 26.7 of 40 possible HVIP points. on quality metrics based on a continuous scale, starting at 0 points and gradually increasing to 10 points. Each hospital’s total quality performance score, The continuous scale stretches over almost the whole which would be used to determine its HVIP payment distribution of performance, giving even top-performing adjustment, would have a maximum of 40 points. In hospitals an incentive to continue to improve. our HVIP model, each hospital has a total number of points based on its performance against our continuous In our HVIP model, each measure has a continuous performance-to-points scale (Table 7-6). The 3,021 performance-to-points scale based on the 2nd percentile hospitals included in our sample had a nearly normal of hospital performance (0 points) to the 98th percentile distribution of total quality performance scores under our of hospital performance (10 points), which is based on 22 HVIP model (Figure 7-2, p. 196). the hospitals in our data set. This scale—from the 2nd percentile to 98th percentile—is meant to represent In our HVIP model, the average total HVIP score point empirically derived scores that available evidence suggests total for all hospitals was 22.9 points (Table 7-7, p. 197). can be achieved by an optimally performing hospital On average, mortality contributed 7 of those points (Safran et al. 2007). Although scoring is continuous, because more hospitals perform better on this measure hospitals would know in advance what performance June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 195

218 FIGURE FIGURE Cumulative change... X-X Hospitals have a nearly normal distribution of 7–2 total quality performance under the potential HVIP 350 300 250 200 150 Number of hospitals 100 50 0 6.9 to 8.2 8.2 to 9.5 9.5 to 10.8 14.7 to 16.0 36.8 to 38.1 17.3 to 18.6 16.0 to 17.3 34.2 to 35.5 35.5 to 36.8 22.5 to 23.8 38.1 to 39.4 23.8 to 25.1 25.1 to 26.4 19.9 to 21.2 12.1 to 13.4 29.0 to 30.3 30.3 to 31.6 10.8 to 12.1 32.9 to 34.2 31.6 to 32.9 18.6 to 19.9 21.2 to 22.5 26.4 to 27.7 13.4 to 14.7 27.7 to 29.0 Total HVIP points (out of 40 points) Note: HVIP (hospital value incentive program). Hospitals receive 0 to 40 total HVIP points based on their performance on four equally weighted measures (readmissions, mortality, Medicare spending per beneficiary, and patients’ overall rating of hospital). There are 3,021 hospitals included in our HVIP model. Source: MedPAC analysis of Medicare fee-for-service hospital quality data, 2014–2016. compared with readmissions, MSPB, and overall patient providers’ populations, including social risk factors. experience rating, which each contributed about 5 points However, adjusting measure results for social risk to the total score. In addition, there were some differences factors can mask disparities in clinical performance, so in total HVIP scores based on hospital characteristics. Medicare should adjust performance payments through For example, in our model, major teaching hospitals peer grouping rather than through performance score had a lower average total HVIP score compared with adjustments. (In peer grouping, each provider is compared nonteaching hospitals (21.2 points compared with 23.2 with its “peers”—defined as providers with a similar 23 points, respectively). This difference is partially because patient mix.) The Commission also believes that major teaching hospitals have worse readmission rates and Medicare should target technical assistance resources to therefore fewer points in that domain of the HVIP scoring low-performing providers and should support research and model (3.8 points for teaching hospitals compared with 5.5 data collection to reduce measurement bias, including, for points for nonteaching hospitals). example, the effects of social risk factors. Based on these principles, our HVIP model distributes Converting HVIP points to payment adjustments using quality-based payments to hospitals classified in 10 peer peer grouping In measuring providers’ performance on Note: Note and Source are in InDesign. groups. Each peer group has about the same number quality measures, the Commission contends that Medicare Source: of hospitals (in our model, about 300 hospitals), and should take into account, as necessary, differences in Applying the Commission’s principles for measuring quality: Population-based measures and hospital quality incentives 196

219 TABLE Illustrative total HVIP points by hospital characteristics 7–7 Average: Patients’ Total HVIP overall rating MSPB points Number of Mortality Readmissions of hospital hospitals Hospital group score score (score) points 22.9 3,021 All hospitals 5.5 5.1 7.0 5.3 Hospital size 5.3 Large urban 1,209 22.2 7.6 4.5 4.8 Other urban 1,065 23.8 5.7 7.2 5.2 5.7 747 Rural 5.9 5.4 6.1 5.5 22.9 Teaching status 4.7 4.9 7.8 3.8 21.2 300 Major teaching 5.2 22.9 764 Other teaching 7.6 4.9 5.3 6.7 5.3 5.6 5.5 23.2 1,957 Nonteaching Ownership Nonprofit 5.5 23.9 1,826 6.8 5.3 5.7 5.1 754 21.1 4.8 7.4 4.4 For profit 441 Government 5.5 6.1 5.2 5.3 22.1 DSH No DSH 410 25.8 6.1 7.2 5.5 7.1 5.5 5.5 Moderate to low DSH 1,897 23.2 6.9 5.3 7.3 4.3 High DSH 665 4.5 20.3 4.2 Note: HVIP (hospital value incentive program), MSPB (Medicare spending per beneficiary), DSH (disproportionate share). Hospitals receive up to a total of 40 points based on their performance on four equally weighted measures (up to 10 points each): risk-adjusted, unplanned readmissions; risk-adjusted 30-day postdischarge mortality; MSPB; and patients’ overall rating of hospital. “Patients’ overall rating of hospital” is the share of Hospital Consumer Assessment of Healthcare Providers ® survey respondents that gave the hospital an overall rating of 9 or 10 on 10-point scale. “High DSH” hospitals have higher proportions of low-income and Systems patients compared to “no DSH” hospitals. There are 49 hospitals with unknown DSH status. Source: MedPAC analysis of Medicare fee-for-service hospital quality data, 2014–2016. hospitals are assigned to peer groups based on their In our HVIP model, we followed five steps to convert share of Medicare patients who are fully dual-eligible performance points to payment adjustments using beneficiaries—that is, who also fully qualify for Medicaid, currently available hospital quality and payment data. (See which can be a proxy for low income. (In fiscal 2019, the text box, pp. 202–203, describing the process used in our HRRP will use five peer groups based on the hospital’s HVIP model to convert each hospital’s HVIP points to a share of Medicare patients who are fully dual-eligible quality-based payment adjustment.) Overall, we found that beneficiaries.) Since our HVIP model is designed to be it was feasible to compute incentive payments that support budget neutral, each peer group has, in essence, a budget the Commission’s HVIP’s goals. based on a 2 percent payment withhold from each of the After scoring each hospital on the same continuous peer group’s hospitals. This budget is redistributed to the performance-to-points scale, we divided the 3,021 peer group’s hospitals based on their HVIP points. The hospitals in our HVIP sample into 10 equal-sized peer 2 percent withhold is the same as the withhold in the groups based on the hospitals’ shares of fully dual-eligible existing VBP Program, but policymakers could raise or Medicare patients (text box Steps 1 and 2, p. 202). The lower that amount. June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 197

220 TABLE Illustration of hospital payment adjustments using 7–8 peer groups under potential HVIP model Average: Peer group budget Percentage based on 2 percent adjustment to withhold of hospitals’ Total Share of fully base IPPS base IPPS payments HVIP dual-eligible payments per (in millions) Peer group points beneficiaries HVIP point % % 26.4 $211.8 0.08 (lowest share of fully dual-eligible beneficiaries) 1 6.5 24.5 2 10.8 228.2 0.08 0.08 274.9 23.9 13.1 3 4 15.2 23.7 227.8 0.08 5 23.7 208.6 0.08 17.2 6 19.3 22.6 216.4 0.09 7 22.1 22.6 165.8 0.09 8 22.3 25.3 169.5 0.09 9 21.2 179.5 0.09 30.5 0.10 (highest share of fully dual-eligible beneficiaries) 10 18.3 148.3 48.3 Note: HVIP (hospital value incentive program), IPPS (inpatient prospective payment system). There are about 300 hospitals in each of the 10 hospital peer groups. Peer groups are assigned based on the share of the hospital’s Medicare patients who are fully dual eligible for Medicare and Medicaid for a majority of the year. Fully dual-eligible beneficiaries qualify for a full range of Medicaid benefits. MedPAC analysis of Medicare fee-for-service hospital quality data, 2014–2016. Source: average share of these beneficiaries per hospital in each smaller IPPS base payments to be used in the withhold peer group ranged from less than 7 percent (Peer Group calculation. 1) to about 48 percent (Peer Group 10) (Table 7-8). The For each peer group, we also calculated the percentage average total HVIP points hospitals received in each peer adjustment to payments per point, which converts total group ranged from 26.4 (Peer Group 1) to 18.3 (Peer HVIP points to dollars and results in spending the 2 Group 10). Peer Group 10 had fewer total HVIP points percent withhold for each group (text box Step 4, p. 202). mainly because of higher average readmissions and The percentage adjustments to payments per point range lower overall patient ratings compared with Peer Group from 0.08 percent (Peer Group 1) to 0.10 percent (Peer 1 hospitals. Although, on average, Peer Group 10’s point Group 10) (Table 7-8). In other words, high-performing total was lower, some hospitals in the peer group were hospitals in Peer Group 10 have the potential to earn a high performers and received more HVIP points than the slightly higher payment adjustment per performance point average hospital. compared with the other groups because the percentage For each peer group, we calculated a budget of expected adjustment to payments per point for Peer Group 10 is HVIP payments to the group’s hospitals based on a 2 higher than the other groups. percent withhold of base inpatient prospective payment We calculated each hospital’s HVIP-based payment system (IPPS) payments from each of the group’s adjustment using its total HVIP points and its peer group’s hospitals (text box Step 3, p. 202). Under our model, a conversion factor for points-to-payment adjustment (text total of $2.03 billion is distributed to hospitals based on box Step 5, p. 202). In our HVIP model, small differences their HVIP points. The budget for each peer group ranges exist between the peer groups’ ranges of payment from about $275 million (Peer Group 3) to about $148 adjustments. In general, a hospital’s payment adjustment million (Peer Group 10) (Table 7-8). Inherent in the peer could range from –1.4 percent to 1.6 percent based on the group budgets are the number of discharges for the peer hospital’s base IPPS payment after accounting for their 2 group’s hospitals, so the budget is smaller for those peer percent withhold (Table 7-9). (By design, no hospital can groups that have hospitals with fewer discharges and thus Applying the Commission’s principles for measuring quality: Population-based measures and hospital quality incentives 198

221 TABLE Illustrative HVIP payment adjustments by hospital peer groups 7–9 Average Range of HVIP payment adjustments withhold of total base Peer group Relative to withhold After withhold IPPS payments 1 (lowest share of fully dual-eligible beneficiaries) 2% –1.1% to + 1.1% 44% to 156% 2 2 –1.1 to + 1.1 47 to 155 3 2 –1.2 to + 1.0 38 to 149 2 4 44 to 146 –1.1 to + 0.9 2 5 40 to 152 –1.2 to + 1.1 2 –1.1 to + 1.0 45 to 152 6 37 to 154 7 2 –1.3 to + 1.1 –1.4 to + 1.3 8 2 31 to 163 37 to 158 9 2 –1.3 to + 1.2 10 2 –1.3 to + 1.6 37 to 180 (highest share of fully dual-eligible beneficiaries) Note: HVIP (hospital value incentive program), IPPS (inpatient prospective payment system). There are about 300 hospitals in each of the 10 hospital peer groups. Peer groups are assigned based on the share of the hospital’s Medicare patients who are fully dual eligible for Medicare and Medicaid for a majority of the year. Fully dual-eligible beneficiaries qualify for a full range of Medicaid benefits. The average HVIP adjustments after the withhold is zero by design. MedPAC analysis of Medicare fee-for-service hospital quality data, 2014–2016. Source: lose more than its 2 percent withhold.) Thus, hospitals top performers in the HVIP model. About 1 percent (34 can recover between 31 percent and 180 percent of their 2 hospitals) were top performers in the existing programs percent withhold. but were poor performers in the HVIP model. The HACRP appeared to play a role in this trend (i.e., some hospitals Under our model, about half of the hospitals (1,510) would were poor performers in the existing programs because receive a penalty and about half (1,511 hospitals) would they received a HAC penalty but did well under the receive a reward. About 11 percent of hospitals (367) HVIP model.) This supports our concerns with potential would receive a reward more than 1.5 times the withhold. misreporting of hospital infection data in a program that About 12 percent (365) would receive a penalty of less uses a tournament model rather than fixed targets. than one-half of the withhold. Effect of peer grouping on reducing disparities Comparison of HVIP model to existing among hospitals hospital quality programs Our HVIP model uses a small set of measures, a To understand differences between hospital performance continuous performance-to-points scale, and converts in the existing programs and our HVIP model, we those points to payment adjustments relative to groups of assigned hospitals to quartiles based on their total hospitals that serve similar shares of fully dual-eligible performance in the existing programs and then quartiles populations (hospital peer groups). Since one goal of an based on their performance under the HVIP model. About HVIP is to adjust payments to account for differences in a quarter of hospitals were in the same performance social risk factors, we examined how hospitals serving 24 quartile under the existing programs and the HVIP model. large shares of low-income patients perform. Figure Three-quarters of hospitals were in the same or within one 7-3 (p. 200) compares the existing quality payment performance quartile under the existing program and the program adjustments with the HVIP model’s payment HVIP model. At the extremes, 2 percent (61 hospitals) adjustments by peer group. All the HVIP adjustments are were poor performers in the existing programs but were zero relative to the average within each peer group since June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 199

222 FIGURE Compared with existing quality payment programs, the potential HVIP FIGURE 7–3 Cumulative change... makes payment adjustments more equitable for hospitals 1-X grouped by share of fully dual-eligible beneficiaries 0.5 Average payment adjustments under existing programs 0.4 Average payment adjustments under potential HVIP model 0.3 0.2 0.1 0.0 –0.1 –0.2 relative to average –0.3 Percent payment adjustment –0.4 –0.5 Peer Group 1 Peer Group 10 Peer Group 3 Peer Group 6 (lowest share of (highest share of fully dual-eligible beneficiaries) fully dual-eligible beneficiaries) HVIP (hospital value incentive program). The existing quality programs include the Hospital Readmissions Reduction Program (HRRP), Hospital-Acquired Condition Note: Reduction Program (HACRP), and Hospital Value-based Purchasing (VBP) Program. The HRRP and HACRP are penalties, and the VBP Program is budget neutral. To make the existing programs and HVIP comparable, we included a budget-neutrality adjustment in the existing programs’ adjustment. The budget-neutrality adjustment is the overall existing program adjustment divided by overall base payments (0.93 percent). The average HVIP adjustment is the sum of each hospital’s HVIP adjustment after the withhold divided by the sum of each hospital’s base payment. The HVIP is budget neutral. Peer groups are assigned based on the share of the hospital’s Medicare patients who are fully dual eligible for Medicare and Medicaid for a majority of the year. Fully dual-eligible beneficiaries qualify for the full range of Medicaid benefits. Source: MedPAC analysis of Medicare fee-for-service hospital quality data, 2014–2016. the adjustments are budget neutral within each peer group. HVIP program, they would receive a smaller positive Under the existing programs, Peer Group 1 (lowest share adjustment—on average, 0.06 percent. Under the existing of fully dual-eligible beneficiaries) hospitals receive a 0.39 programs, the high–DSH hospitals receive, on average, a percent positive adjustment while Peer Group 10 (highest –0.22 percent adjustment; under an HVIP program, that share of fully dual-eligible beneficiaries) hospitals receive adjustment would rise to an average of –0.04 percent. a –0.41 percent adjustment. Thus, compared with the Conclusion existing quality payment programs, the HVIP approach makes payment adjustments among hospitals that serve A single quality payment program for hospitals, such as different populations more equitable. our HVIP model, would be simpler to administer and would produce more equitable results compared with the We can also see this effect in Figure 7-4, which compares existing quality payment programs. The HVIP, as a single existing and HVIP model payment adjustments program, would eliminate the complexity of overlapping for different groups of hospitals according to their program requirements, focus on outcomes, and promote disproportionate share (DSH) hospital status (which can the coordination of care. It would also align with the also be considered a proxy for low-income status). Under Commission’s principles for quality measurement, in the existing quality programs, non-DSH hospitals receive, Note: Note and Source are in InDesign. particular, by setting absolute value targets and using on average, a 0.42 percent positive adjustment; under an Source: Applying the Commission’s principles for measuring quality: Population-based measures and hospital quality incentives 200

223 FIGURE FIGURE Compared with existing quality payment programs, the potential HVIP makes 7–4 Cumulative change... 1-X payment adjustments more equitable for hospitals grouped by DSH status 0.5 Average payment adjustments under existing programs 0.4 Average payment adjustments under potential HVIP model 0.3 0.2 0.1 0.0 relative to average –0.1 Percent payment adjustment –0.2 –0.3 No DSH Moderate to low DSH High DSH HVIP (hospital value incentive program), DSH (disproportionate share hospital). The existing quality programs include the Hospital Readmissions Reduction Program Note: (HRRP), Hospital-Acquired Condition Reduction Program (HACRP), and Hospital Value-based Purchasing (VBP) Program. The HRRP and HACRP are penalties, and the VBP Program is budget neutral. To make the existing programs and HVIP comparable, we included a budget-neutrality adjustment in the existing programs’ adjustment. The budget-neutrality adjustment is the overall existing program adjustment divided by overall base payments (0.93 percent). The average HVIP adjustment is the sum of each hospital’s HVIP adjustment after the withhold divided by the sum of each hospital’s base payment. The HVIP is budget neutral. Source: MedPAC analysis of Medicare fee-for-service hospital quality data, 2014–2016. Over the next year, the Commission plans to continue to peer grouping to account for differences in provider populations. Under peer grouping in our HVIP model, refine a design for an HVIP consistent with our principles differences in payment adjustments were reduced among for quality measurement. Some topics the Commission providers serving populations of varying social risk will further explore include weighting of measures, factors. withhold values, patient experience measures, and patient safety measures. ■ Note: Note and Source are in InDesign. Source: June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 201

224 Steps to convert hospital value incentive program points to payment adjustments using peer grouping For each peer group, create a budget of Step 3: ur hospital value incentive program (HVIP) expected HVIP payments to hospitals, based on a 2 model distributes quality-based payments to percent withhold from each of the hospitals in the peer hospitals classified in 10 peer groups. Each O group (e.g., 2 percent of each hospital’s base inpatient peer group has about the same number of hospitals, prospective payment system (IPPS) payments). and hospitals are assigned to peer groups based on their share of Medicare patients who are fully dual- For each peer group, calculate the percentage Step 4: eligible beneficiaries—that is, who also fully qualify adjustment to payment per HVIP point, which converts for Medicaid, which can be a proxy for low income. total HVIP points to dollars and results in spending the Since our HVIP model is designed to be budget group’s budget defined in Step 3. neutral, each peer group has, in essence, a budget based on a 2 percent payment withhold from each of Percentage adjustment to payments per point = HVIP the peer group’s hospitals. This budget is redistributed budget for peer group / (sum (each hospital’s base IPPS to the peer group’s hospitals based on their quality payments × hospital’s total HVIP points)) performance. Compute each hospital’s adjustment for the Step 5: We followed five steps to covert each hospital’s quality coming year based on past performance and their peer measure performance to a payment adjustment that group’s percentage adjustment to payment per HVIP provides rewards or penalties. point. Convert each hospital’s performance on quality Step 1: Hospital’s HVIP-based adjustment = percentage measures to total HVIP points based on a continuous adjustment to payments per point × hospital’s total performance-to-points scale. (Every hospital is scored HVIP points on the same scale.) Multiply the hospital’s HVIP-based adjustment by Divide hospitals into 10 equal-sized peer groups Step 2: the hospital’s withhold of IPPS payments to yield the based on the hospital population’s share of fully dual- payment adjustment in dollars. eligible patients. (continued next page) Applying the Commission’s principles for measuring quality: Population-based measures and hospital quality incentives 202

225 Steps to convert hospital value incentive program points to payment adjustments using peer grouping (cont.) for the peer group is a sum of the two hospital’s Table 7-10 below describes an example of converting withholds (or $1.3 million). We then calculate the HVIP points to payment adjustments using peer percentage adjustment to payments per point for grouping. First, we convert each hospital’s quality the peer group, which converts total HVIP points to measure performance to total HVIP points based on dollars and results in spending the entire $1.3 million the continuous performance-to-points scale (Step 1). budget (Step 4). For every HVIP point that a hospital As seen at the top of the table, Hospital 1 has higher in the peer group earns, it can receive a 0.065 percent total HVIP performance with 40 points compared payment adjustment per point. Based on the hospital’s with Hospital 2’s 30 points. We assume two hospitals HVIP performance and the peer group’s percentage were assigned to a peer group because of a similar adjustment to payments per point, Hospital 1 will earn share of fully dual-eligible beneficiaries (Step 2). a payment adjustment of 2.6 percent, which is equal We withhold 2 percent of each of the hospital’s total to $130,000 (or a reward of $30,000 greater than the base IPPS payments (Step 3). Since Hospital 1 has hospital’s withhold) (Step 5). Because Hospital 2 had fewer discharges, its 2 percent withhold is less than lower HVIP points, it will have a $30,000 penalty. ■ Hospital 2’s withhold. As shown in the middle of the table, the total HVIP bonus pool to be redistributed TABLE Example of converting HVIP points to payment 7–10 adjustments for a peer group’s hospitals Hospital 2 Hospital 1 (5,000 discharges) (500 discharges) HVIP points (Step 1) 40 30 $5,000,000 Total base IPPS payments $60,000,000 2 percent withhold of IPPS payments $100,000 $1,200,000 Total HVIP budget for peer group (Step 3) $1,300,000 Percentage adjustment to payments per point (Step 4) 0.065% adjustment per point 2.60% ($130,000) Hospital HVIP-based adjustment (Step 5) 1.95% ($1,170,000) –0.05% ( –$30,000) +0.60% (+$30,000) Reward or penalty relative to 2 percent withhold Note: HVIP (hospital value incentive program), IPPS (inpatient prospective payment system). This example assumes the peer group has two hospitals (Step 2). June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 203

226 Endnotes 11 The HCD measure includes beneficiaries ages 65 years and 1 For clarity and consistency with the Commission’s past older, while the PPA measure was specified for beneficiaries work, we use the term potentially preventable admissions ages 67 years and older. The PPA measure focuses on throughout the chapter. The literature and industry also admissions tied to five chronic conditions. For the HCD refer to the measure concept as avoidable hospitalizations, calculations, chronic conditions are identified from a set of and ambulatory care–sensitive condition hospitalizations, 15 (acute myocardial infarction/ischemic heart disease, CHF, hospitalizations for potentially preventable complications . specified heart arrhythmias, dementia, hematologic disease, HEDIS is a registered trademark of NCQA. The HEDIS 2 lung disease, psychiatric disease, chronic kidney disease, potentially preventable admissions measure is called endocrine disease, vascular disease, neuromuscular disease, “hospitalizations for potentially preventable complications.” diabetes, cancer, liver disease, stroke). The conditions were chosen based on the combination of high prevalence and CMS has proposed to retain this measure as a 2019 MA Plan 3 mortality as well as associated health care spending. Finder display page measure. The agency has also signaled its intent to move the measure to the star rating program in 2022 12 One possible explanation for the increase in mortality days (Centers for Medicare & Medicaid Services 2018a). in 2015 is the very severe flu season from October 2014 to March 2015. Beneficiaries who died in the January to March 4 The expected-discharges value is the predicted number of portion of the 2014 to 2015 flu season would have fewer hospitalizations based on the age, sex, and comorbidities HCDs because they had more mortality days subtracted from (i.e., hierarchical condition categories (HCCs)) of the eligible the 365 calendar days of 2015. Beneficiaries who died in the population of beneficiaries. October to December portion of the 2014 to 2015 flu season would have more HCDs because they had fewer mortality The measure uses discharges rather than admissions 5 days subtracted from the 365 calendar days of 2014. because patients who die in the hospital are not included in potentially the measure. For consistency, we use the term The Commission recommended a readmissions reduction 13 . preventable admissions program in our 2008 report to the Congress (Medicare Payment Advisory Commission 2008). Our June 2018 report 6 Eight percent represents 740,000 potentially preventable to the Congress also includes a study mandated by the 21st admissions out of 9.5 million admissions. Century Cures Act of 2016 that examines whether changes in readmission rates under the HRRP are related to any changes 7 The Commission has previously referred to this measure as in outpatient and emergency services furnished. “healthy days at home.” The measure’s new name does not presume that beneficiaries are healthy just because they are 14 The PSI 90 measure is a composite of eight patient safety at home and is more explicitly inclusive of beneficiaries who measures: PSI 03 (pressure ulcer); PSI 06 (iatrogenic may be living in long-term care facilities. pneumothorax); PSI 07 (central venous catheter–related bloodstream infections); PSI 08 (postoperative hip fracture); 8 For example, the Endovascular Treatment for Small Core and PSI 12 (perioperative pulmonary embolism or deep Anterior Circulation Proximal Occlusion with Emphasis on vein thrombosis); PSI 13 (postoperative sepsis); PSI 14 Minimizing CT to Recanalization Times (ESCAPE) study (postoperative wound dehiscence); and PSI 15 (accidental uses a days alive and out of the hospital measure during the puncture or laceration). six months after the randomized use of pulmonary artery catheters for patients with congestive heart failure. In 2018, 2 process-of-care measures were dropped from the 15 VBP Program, and the 1 remaining process-of-care measure, Because our goal is to calculate market-specific estimates 9 PC–01 (elective delivery before 39 weeks), was moved to the of HCDs and ultimately compare payment models across patient safety domain; this measure’s weight increased from and within market areas, we used a fixed-effect model that 20 percent to 25 percent. CMS has proposed removing the includes an indicator variable for each of the markets in the PC–01 measure from the VBP Program (Centers for Medicare regression model to better estimate the age, sex, and HCC and Medicaid Services 2018b). covariates. Market areas 10 refers to the over 1,200 MedPAC-defined market areas used in the PPA analysis. Applying the Commission’s principles for measuring quality: Population-based measures and hospital quality incentives 204

227 The IQRP was mandated by the Medicare Prescription 16 22 We included only hospitals paid through the inpatient Drug, Improvement, and Modernization Act of 2003 and prospective payment system. Because we wanted to model updated by the Patient Protection and Affordable Care Act of the scoring of all four measures, we did not include hospitals 2010 (PPACA). The HRRP, VBP Program, and HACRP are with no publicly reported HCAHPS data or MSPB data (from mandated in PPACA. CMS) or risk-adjusted mortality or readmissions value of 0 or missing. A policy question is how to score missing values— CMS calculates claims-based mortality, readmissions, and 17 for example, when a hospital’s population is too small for MSPB measures. CMS oversees the Hospital Consumer HCAHPS. Another policy question is whether and how to ® patient Assessment of Healthcare Providers and Systems include critical access hospitals, which may have numbers too experience survey (including certifying survey vendors small for valid measurement. and developing standardized data collection and sampling protocols). Hospitals work with a survey vendor or follow the Based on suggestions from the Commission and the recent 23 standardized protocols themselves to collect and report the requirement legislated in the 21st Century Cures Act of 2016, core and supplemental experience data from their patients. CMS is implementing a peer-group scoring model, using five peer groups, in the HRRP. Others have tested and found ® is a registered trademark of AHRQ, a U.S. 18 CAHPS that the peer-grouping approach adequately accounts for government agency. differences among providers serving populations with social risk factors (Office of the Assistant Secretary for Planning and NHSN is operated by the Centers for Disease Control and 19 Evaluation 2016, Samson et al. 2018). Prevention. It is the nation’s most widely used health care– associated infection tracking system. Acute care hospitals, We compared the amount of quality payment adjustments in 24 long-term care hospitals, dialysis facilities, inpatient existing programs with the HVIP model payment adjustments rehabilitation facilities, and other facility types report data to by hospital characteristics (e.g., size, teaching status) (see NHSN. Table 7-A1 in online Appendix 7-A, available at http://www. medpac.gov). To make the existing programs and HVIP 20 This PSI 90 composite measure was created by AHRQ to comparable, we included a budget-neutrality adjustment for help hospitals measure adverse events and address their own the existing program adjustment calculation. quality improvement efforts. 21 In some current Medicare quality payment programs, CMS uses differential measure weighting to prioritize outcome measures over process measures. For example, the MA star rating program assigns outcome and intermediate outcome measures a weight of 3, patient experience and access measures a weight of 1.5, and process measures a weight of 1. June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 205

228 References Centers for Medicare & Medicaid Services, Department of Health Agency for Healthcare Research and Quality. 2016. National and Human Services. 2012. Medicare program; reform of hospital scorecard on rates of hospital-acquired conditions 2010 to 2015: and critical access hospital conditions of participation. Final rule. Interim data from national efforts to make health care safer . Federal Register 77, no. 95 (May 16): 29034–29076. Rockville, MD: AHRQ. http://www.ahrq.gov/professionals/ quality-patient-safety/pfp/2015-interim.html. Chatterjee, P., T. C. Tsai, and A. K. Jha. 2015. Delivering value American Journal of Managed by focusing on patient experience. Barnett, M. L., D. C. Grabowski, and A. Mehrotra. 2017. Home- 21, no. 10 (October): 735–737. Care to-home time: Measuring what matters to patients and payers. New England Journal of Medicine 377, no. 1 (July 6): 4–6. Covinsky, K. E., E. Pierluissi, and C. B. Johnston. 2011. Hospitalization-associated disability: “She was probably able to Healthy-time measures of health Burns, M., and J. Mullahy. 2016. Journal of the American Medical ambulate, but I’m not sure.” . NBER working paper no. outcomes and healthcare quality Association 306, no. 16 (October 26): 1782–1793. 22562. Cambridge, MA: National Bureau of Economic Research. Davies, S. M., K. M. McDonald, E. Schmidt, et al. 2009. Calderwood, M. S., S. S. Huang, V. Keller, et al. 2017. Variable Expanding use of the Prevention Quality Indicators: Report of case detection and many unreported cases of surgical-site . Report prepared for the Agency for Clinical Expert Review Panel infection following colon surgery and abdominal hysterectomy Healthcare Research and Quality. Rockville, MD: AHRQ. in a statewide validation. Infection Control and Hospital 38, no. 9 (September): 1091–1097. Epidemiology Freund, T., S. M. Campbell, S. Geissler, et al. 2013. Strategies for reducing potentially avoidable hospitalizations for ambulatory Centers for Disease Control and Prevention. 2016a. Adherence 11, no. 4 Annals of Family Medicine care-sensitive conditions. to the Centers for Disease Control and Prevention’s (CDC’s) (July–August): 363–370. infection definitions and criteria is needed to ensure accuracy, completeness, and comparability of infection information. https:// Gao, J., E. Moran, Y. F. Li, et al. 2014. Predicting potentially www.cdc.gov/nhsn/cms/cms-reporting.html. avoidable hospitalizations. 52, no. 2 (February): Medical Care 164–171. Centers for Disease Control and Prevention. 2016b. National and state healthcare-associated infections progress report . Atlanta, Medicare Payment Advisory Commission. 2017. Report to the GA: CDC. . Washington, DC: MedPAC. Congress: Medicare payment policy Centers for Medicare & Medicaid Services, Department of Health Medicare Payment Advisory Commission. 2016a. Letter to the and Human Services. 2018a. Advance notice of methodological Acting Administrator of the Centers for Medicare and Medicaid changes for calendar year (CY) 2019 for Medicare Advantage Services regarding the hospital star rating system. September 22. capitation rates, Part C and Part D payment policies and 2019 draft call letter. February 1. https://www.cms.gov/medicare/ Report to the Medicare Payment Advisory Commission. 2016b. health-plans/medicareadvtgspecratestats/downloads/advance2018. . Washington, DC: MedPAC. Congress: Medicare payment policy pdf. Report to Medicare Payment Advisory Commission. 2015a. Centers for Medicare & Medicaid Services, Department of Health . the Congress: Medicare and the health care delivery system and Human Services. 2018b. Fiscal year (FY) 2019 Medicare Washington, DC: MedPAC. hospital inpatient prospective payment system (IPPS) and long term acute care hospital (LTCH) prospective payment system Medicare Payment Advisory Commission. 2015b. The use of low- proposed rule, and request for information. https://www.cms.gov/ value care in Medicare is substantial. The MedPAC Blog , May 21. Newsroom/MediaReleaseDatabase/Fact-sheets/2018-Fact-sheets- http://www.medpac.gov/-blog-/medpacblog/2015/05/21/use-of- items/2018-04-24.html. low-value-care-in-medicare-is-substantial. Centers for Medicare & Medicaid Services, Department of Health Medicare Payment Advisory Commission. 2014. Report to and Human Services. 2017. HCAHPS patient-level correlations . the Congress: Medicare and the health care delivery system table. http://www.hcahpsonline.org/globalassets/hcahps/summary- Washington, DC: MedPAC. analyses/patient-level/july-2015--june-2016-discharges.pdf. Medicare Payment Advisory Commission. 2013. Report to the Congress: Medicare payment policy . Washington, DC: MedPAC. Applying the Commission’s principles for measuring quality: Population-based measures and hospital quality incentives 206

229 Medicare Payment Advisory Commission. 2011. Report to Safran, D. G., D. A. Taira, W. H. Rogers, et al. 1998. Linking . the Congress: Medicare and the health care delivery system Journal of Family primary care performance to outcomes of care. Washington, DC: MedPAC. 47, no. 3 (September): 213–220. Practice Medicare Payment Advisory Commission. 2008. Report to the Samson, L. W., L. M. Chen, A. M. Epstein, et al. 2018. Dually Congress: Reforming the delivery system . Washington, DC: enrolled beneficiaries have higher episode costs on the Medicare MedPAC. 37, no. 1 Health Affairs Spending Per Beneficiary measure. (January): 86–94. Office of Inspector General, Department of Health and Human Nationwide analysis of common characteristics in Services. 2016. Teno, J. M., P. L. Gozalo, J. P. Bynum, et al. 2013. Change in end- OIG home health fraud cases . Washington, DC: OIG. of-life care for Medicare beneficiaries: site of death, place of care, Journal of the and health care transitions in 2000, 2005, and 2009. Office of the Assistant Secretary for Planning and Evaluation, 309, no. 5 (February 6): 470–477. American Medical Association Department of Health and Human Services. 2016. Report to Congress: Social risk factors and performance under Medicare’s Wasfy, J. H., C. M. Zigler, C. Choirat, et al. 2017. Readmission . Washington, DC: ASPE. value-based purchasing programs rates after passage of the Hospital Readmissions Reduction 166, Program: A pre-post analysis. Annals of Internal Medicine Rajaram, R., C. Barnard, and K. Y. Bilimoria. 2015. Concerns no. 5 (March 7): 324–331. about using the Patient Safety Indicator–90 composite in pay- Journal of the American Medical for-performance programs. Wholey, D. R., J. B. Christianson, M. Finch, et al. 2003. 313, no. 9 (March 3): 897–898. Association American Evaluating health plan quality 1: A conceptual model. Journal of Managed Care 9 spec. no. 2 (June): SP53–64. Safran, D. G., M. Anastario, H. P. Rodriguez, et al. 2007. Evaluating the potential for an empirically-derived standard of performance excellence in ambulatory patient experience measures. Pre-publication draft. June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 207

230

231 CHAPTER 8 Medicare accountable care organization models: Recent performance and long-term issues

232

233 CHAPTER 8 Medicare accountable care organization models: Recent performance and long-term issues Chapter summary In this chapter Medicare accountable care organizations (ACOs) were created to help • Introduction moderate the growth in Medicare spending and improve quality of care for Background on ACOs • beneficiaries by giving providers greater responsibility for costs and quality. In reviewing current Medicare ACO models, we found that some models— ACO quality and financial • predominantly those at risk for both savings and losses (two-sided risk)—have performance relative to CMS-designed benchmarks produced small savings relative to their benchmarks set by CMS, and all have maintained or improved quality. Spending relative to benchmarks is important ACO quality and • because it determines which ACOs will receive “shared savings” bonuses. financial performance However, some have raised the point that benchmarks are not necessarily results according to other the best measure of what spending would have been in the absence of the researchers ACO and thus may not be a good measure of true program savings. From • New tools to allow ACOs to our review of the literature on this question, we conclude that ACOs may manage care have saved Medicare from 1 percent to 2 percent more than indicated by their performance relative to benchmarks and that two-sided ACO models appear to • Long-term issues for save more than one-sided ACO models. Medicare ACOs In light of evidence regarding two-sided ACOs and savings, we identified • Conclusion issues that need to be resolved if two-sided ACOs are going to be part of the Medicare program in the long term: Hospitals could be important Are hospitals a viable participant in ACOs? • participants in ACOs, especially given their ability to supply the capital needed to take on two-sided risk. But, while ACOs may want to constrain June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 211

234 unnecessary service use (e.g., unnecessary hospital admissions) to generate savings, hospitals may have conflicting incentives to admit patients to increase their fee-for-service (FFS) revenue. We find that hospitals may still want to participate in ACOs despite the apparent conflict in incentives around inpatient hospital care primarily because most ACO savings to date stem from reductions in the use of post-acute care and not from reductions in inpatient care. • Should asymmetric models be continued? Asymmetric models—models with greater opportunities for savings than losses—could be one strategy to help ACOs transition to two-sided risk. For example, the new Track 1+ ACO model has two asymmetries. First, the shared savings rate is 50 percent (i.e., if actual spending is less than expected spending (the benchmark), then ACOs get half of the savings and Medicare keeps the other half), while the shared loss rate is 30 percent. Second, the loss cap is lower than the savings cap. Because potential gains to ACOs are greater than potential losses, this asymmetric relationship could result in a cost for the Medicare program. Currently, CMS’s Track 1+ model is a demonstration, and savings are not required under CMS’s demonstration and waiver authority. If Track 1+ were incorporated into permanent Medicare law, the costs may need to be offset if performance is essentially random. If it is demonstrated that ACOs are modifying their behavior from what they would have done if not in ACOs and are reducing spending, then this issue will not arise. The Commission will continue to monitor the Track 1+ model to determine whether aspects of it should be extended to other ACO models to encourage uptake of two-sided risk. How should benchmarks be set initially and then rebased for subsequent • agreement periods? The basic ACO model essentially sets benchmarks as a function of historical spending for beneficiaries who would have been attributed to the ACO in the past. If ACOs reduce the level of spending or keep spending growth below the trend in FFS spending, they share in savings. If the same approach were taken in subsequent agreement periods, then ACOs would have to continuously improve over their own past performance to achieve savings, which could create diminishing returns for consistently successful ACOs and potentially discourage long-term participation. In some models, benchmarks are now being rebased using a blend of regional and historical spending. There are additional concerns related to the current benchmark methodology (e.g., the impact of beneficiaries moving in and out of the ACO), and we discuss several approaches to address these issues. Medicare accountable care organization models: Recent performance and long-term issues 212

235 Should the 5 percent bonus for clinicians in advanced alternative payment • models (A–APMs) be distributed differently to encourage A–APM Under current law, clinicians receive a 5 percent bonus on all of participation? their physician fee schedule (PFS) payments if they exceed an annual threshold level for payments or patients in A–APMs. (One-sided ACOs do not qualify as A–APMs, and thus clinicians in them do not receive the bonus.) This A–APM provision could discourage clinicians from participating in ACOs because they would be uncertain about whether they would exceed the threshold. Moving to a system in which clinicians receive a 5 percent bonus with certainty on their share of PFS payments derived from an A–APM could make the incentive more equitable and encourage participation in two-sided ACOs. • What will be the relationship between specialists and two-sided ACOs? Currently, a substantial number of specialists are on the participation lists of ACOs. ACOs may want to include specialists as a way to coordinate the care of their beneficiaries more effectively, and specialists may be incentivized to join ACOs to receive referrals and potentially share in savings. Moving forward, specialty-focused ACO models may also be an option for increasing specialist participation. • Are two-sided ACOs a long-term option in the Medicare program? Some maintain that ACOs are one way for providers to take greater accountability for a group of patients and then transition toward taking full accountability as a Medicare Advantage (MA) plan. If ACOs are regarded only as a transition step toward becoming an MA plan, then it may discourage participation in the ACO model. We have found in previous work that ACOs can be the low-cost option in some areas of the country, and their advantage of lower administrative costs could keep them as a long-term option, if benchmarks are set equitably. Given the early success and popularity of the ACO model, the above issues should be considered if Medicare’s ACOs are to continue in the long term. ■ June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 213

236

237 Introduction Background on ACOs The Commission has long maintained that Medicare Medicare ACOs began in 2012 and have grown rapidly since then to care for about one-third of Medicare FFS should encourage clinicians to improve the quality of care, overall health, and costs of care for a population beneficiaries. In Medicare, ACOs are groups of health care providers that have agreed to be held accountable of patients. In the Medicare Access and CHIP for the cost (that is, spending in Medicare Part A and Part Reauthorization Act of 2015 (MACRA), the Congress provided an incentive for clinicians to join advanced B) and quality of care for a defined group of Medicare beneficiaries. Generally, the goals of ACOs are to lower alternative payment models (A–APMs), which were costs, increase quality of care and patient experience, and predicated on putting an entity responsible for meeting quality goals for a defined patient population at financial improve provider accountability for the cost and quality of risk for Medicare spending. In response, the Medicare care provided to their patients. Theoretically, ACOs could generate savings by substituting lower cost services for program deemed certain models to be A–APMs, created higher cost services (e.g., substituting outpatient services several A–APMs, and is currently developing new ones. for inpatient services) or reducing low- or no-value The Commission has developed principles for A–APMs services. If ACOs achieve their goals, they are rewarded and commented on which A–APMs best meet those with shared savings. principles (Medicare Payment Advisory Commission 2016a). In general, accountable care organization (ACO) There are three main concepts in ACO programs: models at two-sided risk—that is, at risk for losses if spending exceeds benchmarks and sharing savings if • Attribution— Beneficiaries are primarily attributed spending is lower than benchmarks—seem to be the 1 to ACOs based on their use of services. Prospective models that best meet the Commission’s principles attribution occurs when beneficiaries are assigned to because they encourage clinicians to be responsible for an ACO at the start of the performance year (based on the quality and cost of care for a defined population of their prior year usage); retrospective attribution occurs Medicare beneficiaries. when a beneficiary is attributed at the end of the year (based on their current year usage). Unlike Medicare The Commission has determined that the balance of Advantage (MA) plans, beneficiaries attributed to incentives in MACRA between clinicians in A–APMs ACOs can use whatever providers they choose. and those not in A–APMs needs to be rethought. We recommended in our March 2018 report to the • Composition of the ACO— An ACO’s providers do Congress that the current Merit-based Incentive Payment not have to provide all services for a beneficiary, System (MIPS)—which pertains to all fee-for-service although they are responsible for total Part A and (FFS) clinicians not in A–APMs unless excluded—be Part B spending. The essential requirement is that eliminated and replaced with a voluntary value program the providers as a group have enough beneficiaries that would encourage clinicians to elect to be measured attributed to them to meet the minimum requirement for cost and quality purposes as a voluntary group for their model. ACOs can be clinician-only or can (Medicare Payment Advisory Commission 2018). include providers such as hospitals and skilled nursing This recommendation was intended, in part, to prepare facilities (SNFs). clinicians to eventually move to A–APMs. • Benchmarks— The goals of ACOs are assessed using If it is important for clinicians to move to A–APMs, and a set of quality measures (see online Appendix 8-A, if two-sided-risk ACOs are the model most in keeping available at http://www.medpac.gov, for the list of with the Commission’s principles for A–APMs, then it is measures) and spending benchmarks. The spending important to understand performance on cost and quality benchmark is an estimate of Part A and Part B and what issues need to be resolved for two-sided ACOs spending for an ACO’s beneficiaries in a given year. to be a long-term part of the Medicare program. If spending for an ACO’s beneficiaries—including health care services provided outside the ACO—is below the benchmark, then the ACO is eligible to earn June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 215

238 TABLE Characteristics of the MSSP ACO tracks 8-1 b Cap on earned: Risk Maximum shared a Losses Attribution arrangement Savings savings or loss rate c Retrospective Track 1 50% One sided 10% 0% c 60% 15% 10% Two sided Retrospective Track 2 20% 75% Two sided Prospective Track 3 15% MSSP (Medicare Shared Savings Program), ACO (accountable care organization). Note: a The actual shared savings/loss rate could change depending on the ACO’s quality score (e.g., an ACO that scores poorly on quality would receive a smaller shared savings amount than if it had earned a high quality score). b The amount an ACO can share in savings (or repay in shared losses) is capped as a percentage of the benchmark. c These tracks have preliminary prospective attribution and then retrospective attribution for final reconciliation. Source: Centers for Medicare & Medicaid Services 2017c. a “shared savings” payment. If spending is above the Medicare Shared Savings Program benchmark, then the ACO may be financially liable for The MSSP was established in the Patient Protection and shared losses. One-sided-risk arrangements are those Affordable Care Act of 2010 (PPACA) and is a permanent in which ACOs can earn shared savings but are not part of the Medicare program. It currently consists of three responsible for losses; two-sided-risk arrangements ACO tracks: Track 1, Track 2, and Track 3. Table 8-1 are those in which ACOs can earn savings and are summarizes the main differences between each ACO track. responsible for shared losses. The amount of shared MSSP ACOs are allowed to participate as a Track 1 savings an ACO is eligible to earn varies by program. ACO—which is a one-sided track—for only two three- 2 Overview of Medicare’s ACO programs year agreement periods. This stipulation provides a transition period for ACOs to prepare to take on risk The first Medicare ACOs began at the start of 2012 as part as they move to two-sided-risk models (e.g., Track 2 of the Pioneer ACO Model, which was a demonstration or Track 3). (Because they are two-sided, Track 2 and that ended in 2016. Midway through 2012, the first cohort Track 3 qualify as A–APMs and clinicians in them can of ACOs belonging to the Medicare Shared Savings be eligible for the 5 percent bonus on their fee schedule Program (MSSP)—a permanent ACO program created by revenue as established in MACRA.) Furthermore, even the Congress—began. Medicare’s ACO programs have beyond the shared savings/loss rate, there are model- grown quickly since their beginning in 2012, both through specific limits on how much an ACO can earn in savings additional demonstrations and expansion of the MSSP. or pay in losses. These savings and loss limits vary With the passage of MACRA in 2015, the Congress created for each model. For instance, Track 1 shared savings stronger incentives for providers to move into A–APMs payments are capped at 10 percent of benchmark. Track and, therefore, ACOs. The Commission has been supportive 2 shared savings are capped at 15 percent of benchmark, of ACOs since the beginning, especially two-sided risk while losses are capped at 10 percent of benchmark. ACOs that best fit our A–APM principles. For Track 3, shared savings are capped at 20 percent Medicare currently has three ACO programs that have of benchmark, while losses are capped at 15 percent of been in operation since 2016 (or earlier), including the benchmark. MSSP, the Next Generation (NextGen) ACO model, Next Generation ACO Model and the ESRD (End-Stage Renal Disease) Seamless Care Organizations (ESCOs). At the start of 2018, CMS NextGen is a demonstration that began in 2016 and introduced two new ACO models: the Track 1+ ACO was based in part on the previous Pioneer ACO Model. Model and the Vermont All-Payer ACO Model. Medicare accountable care organization models: Recent performance and long-term issues 216

239 NextGen is a two-sided-risk, prospective-attribution Hospital Track 1+ ACOs— Losses are capped at 4 • demonstration run by the CMS’s Center for Medicare & percent of the ACO’s benchmark. Medicaid Innovation (CMMI). For the 2017 and 2018 3 Clinician-only Track 1+ ACOs • Losses are capped — performance years, NextGen qualifies as an A–APM. FFS revenue. This ACO-participant at 8 percent of NextGen ACOs can choose their level of shared savings model differs from the other ACO models because and losses and can opt to share at either 80 percent or 100 it sets a limit relative to FFS revenue instead of the percent of savings and losses. Both shared savings and ACO’s benchmark, which is notable because FFS losses are capped at 15 percent of the ACO’s benchmark. revenue tends to be much lower than the total Part A Additionally, NextGen ACOs receive some regulatory and Part B benchmark. In general, this loss threshold flexibility because of their level of assumed risk. This of 8 percent is lower (and thus more attractive) to flexibility includes waivers to expand the use of telehealth ACOs than the benchmark standard. and to waive the three-day hospital stay rule before using a SNF. While ACOs with hospitals may have less incentive to join the Track 1+ demonstration because they are not eligible ESRD Seamless Care Organizations for the lower risk limit based on FFS revenue, about An ESCO is a disease-specific ACO model that applies half of the Track 1+ ACOs list hospitals as participating to ESRD beneficiaries utilizing chronic dialysis services. providers, indicating broad interest in the model. Savings ESCOs began in 2016 as a demonstration and are run for both types of Track 1+ ACO are limited to 10 percent by CMMI. Beneficiaries are assigned to ESCOs on a of benchmark. “first touch” basis, meaning that the first time an ESRD beneficiary utilizes an ESCO dialysis facility, he or she Vermont All-Payer Model will be prospectively assigned to that ESCO. ESCOs are The other new ACO model in 2018, the Vermont All-Payer split into two tracks based on their size. Large dialysis ACO Model demonstration, brings together Vermont’s organizations (LDOs) are organizations with 200 or more largest payers—Medicare, Medicaid, and commercial dialysis facilities, while non–large dialysis organizations insurers—under one ACO model focused on health (non-LDOs) are those with fewer than 200 dialysis care value and quality. There is one ACO in the model, facilities. In ESCOs, LDOs are automatically at two-sided OneCare Vermont, with model specifics (e.g., benchmark risk, while non-LDOs have the option to be at one-sided methodology) varying slightly for each payer. The overall risk or two-sided risk. For the 2017 and 2018 performance goals of the model, however, are similar across payers years, LDOs and non-LDOs at two-sided risk can qualify and are Vermont specific. In 2018, OneCare Vermont is as A–APMs. For their first performance year, the shared responsible for 122,000 individuals across payers and has savings/loss rate for LDOs is a maximum of 70 percent, 10 participating hospitals from different systems across the and it is 75 percent in their second and future performance 4 state (D’Ambrosio 2017). years; the limit on shared losses is equal to the shared loss rate for the year (e.g., 75 percent). Non-LDOs have a Similar to other ACO models, providers participating shared savings rate of 50 percent, with a limit on savings in the Vermont All-Payer Model have the potential to of 5 percent of benchmark. earn shared savings and a quality bonus payment but are also accountable for shared losses. Because the model’s Track 1+ providers assume risk for the patient population, the model qualifies as an A–APM for the 2018 performance Track 1+ is an asymmetric, two-sided-risk model year. Specific goals for the model include attributing to with prospective attribution that began in 2018. It is a the ACO, by 2022, 90 percent of the state’s Medicare demonstration through CMS’s CMMI authority and is beneficiaries (and 70 percent of all Vermont-insured jointly run with CMS’s MSSP office. ACOs that join residents) and limiting Medicare per capita expenditure Track 1+ are eligible to earn up to 50 percent in shared growth to 0.1 to 0.2 percentage points below projected savings, but because it is an asymmetric risk model, national Medicare growth. The model also includes 21 they are responsible for only 30 percent of shared losses. quality measures that focus on 3 areas prioritized by Additionally, the savings and loss limits vary based on Vermont: reducing deaths due to substance use disorders ACO composition as follows: and suicides, reducing prevalence and morbidity due to June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 217

240 TABLE The number of Medicare ACOs increased from 2017 to 2018 8–2 Number of ACOs Assigned 2018 2017 beneficiaries MSSP (total) 480 506 10.5 million 438 460 Track 1 N/A Track 2 6 8 N/A 38 Track 3 36 N/A 55 * N/A Track 1+ 0 Next Generation 45 58** 1.4 million ESCOs 37 37 16,085 Total N/A 656 562 Note: ACO (accountable care organization), MSSP (Medicare Shared Savings Program), N/A (not available), ESCO (ESRD (End-Stage Renal Disease) Seamless Care Organization). Count of assigned beneficiaries is based on the most recent data available; the total MSSP count is from 2018, the Next Generation count is from 2017, and the ESCO count is from 2016. *Track 1+ started in 2018. **At the start of 2018, there were 58 participating Next Generation ACOs. According to CMS’s website, there are currently only 51 Next Generation ACOs, meaning that 7 ACOs appear to have dropped from the program. The Vermont All-Payer ACO model is included in the Next Generation count (even though it is a separate model) because, for 2018, OneCare Vermont is considered a Next Generation ACO. “Side-by-Side Comparison: Medicare Accountable Care Organization (ACO) Models” from the Kaiser Family Foundation; MSSP 2018 Fast Facts from CMS. Source: chronic conditions, and increasing access to primary care are now responsible for almost one-third of the Medicare (Green Mountain Care Board 2018). FFS population. The only ACO in the model, OneCare Vermont, has been a ACOs are available in all 50 states (and the District of Medicare ACO since 2013, first as an MSSP Track 1 ACO Columbia, Puerto Rico, the U.S. Virgin Islands, and from 2013 to 2017. Starting in 2018, it transitioned into Guam), although not in all areas of every state. MSSP a NextGen ACO. In 2016, actual spending was above the Track 1, a one-sided model, is still the predominant model, benchmark, and OneCare Vermont generated losses of 4.6 accounting for nearly three-quarters of Medicare ACOs. percent relative to the benchmark. Vermont previously had However, MSSP Track 1 does not qualify as an A–APM; 5 other Medicare ACOs operating in the state, including the thus, most MSSP ACOs are not A–APMs. Track 1+, Track 1 ACO Community Health Accountable Care LLC, which qualifies as an A–APM, is in its first year and which had spending 16.9 percent above its benchmark in already has 55 ACOs. It is interesting to note that many 2016 and is not a Medicare ACO in 2018. ACOs include hospitals as participants, even though the financial incentives for hospitals and ACOs may appear to Although the Vermont All-Payer ACO demonstration is a be in conflict. We discuss this apparent contradiction later one-state model, it could be a starting point for all-payer in this chapter. models in other states. It could show, for example, the utility of having most of a provider’s patient population in one payment model with one set of quality indicators. We ACO quality and financial performance will monitor developments. relative to CMS-designed benchmarks Number of participating ACOs in 2018 This section summarizes the quality and financial In 2018, there are 656 Medicare ACOs (Table 8-2 shows performance of the ACO programs active in Medicare the number of ACOs by program). Together, these ACOs Medicare accountable care organization models: Recent performance and long-term issues 218

241 in 2016, the latest performance data available at this future years, the ACO’s quality score is based on how the time. Financial performance is discussed relative to the ACO performed relative to a prospective national FFS CMS benchmarks for each program. In the next section, benchmark. In the MSSP program, ACOs with higher we discuss estimates from the literature on financial quality scores receive greater shared savings bonuses. performance relative to the counterfactual—that is, In 2016, only 4 of the MSSP Track 1 ACOs (1 percent what spending would have been if the ACO did not of 438 ACOs) did not meet the quality standard because exist. Benchmarks and counterfactuals differ because they did not report a complete set of data. (One of benchmarks are designed to fulfill policy goals—for those ACOs dropped out in 2017.) All 22 of the ACOs example, to encourage clinicians to participate in ACOs or participating in Track 2 or Track 3 met the quality to increase equity across the country. Therefore, “savings” standard. MSSP quality scores are high, with average relative to benchmarks will not be the best estimate of quality scores of 93 percent for Track 1, 94 percent for program savings relative to the counterfactual. The latter Track 2, and 96 percent for Track 3. is in some ways the better estimate of whether ACOs are saving the Medicare program money. But “savings” We reviewed changes over time in some of the patient relative to the benchmarks is how the ACOs will determine experience and population-based outcome measures that whether they want to stay in the program; thus, CMS- the Commission supports. The MSSP ACOs on average computed “savings” are also important. had strong patient experience results and high-performing readmission results from 2012 to 2016, with little change MSSP ACOs in results between years. The MSSP was established by PPACA and is a permanent MSSP performance relative to benchmarks part of the Medicare program. The first MSSP ACOs (relative savings) started in April 2012, and the program has grown rapidly to 506 ACOs as of 2018. The program currently consists Summarized financial results for the MSSP ACOs from of three tracks, each with its own savings and loss 2013 to 2016 are shown in Table 8-3 (p. 220). The total specifications: Track 1, Track 2, and Track 3. benchmark amount for the MSSP ACOs is shown in the first row (e.g., $81,377 million in 2016). The second row MSSP ACOs generally perform well on quality is the total amount of actual Part A and Part B Medicare metrics spending for beneficiaries attributed to the MSSP ACOs MSSP, Pioneer, and the NextGen programs use the same (e.g., $80,725 million in 2016). “Relative savings” are set of measures to calculate an annual quality score for defined as the difference between the benchmark and the each ACO. The measure set in 2016 included 31 process actual spending. In 2016, for example, Medicare spent and outcome measures covering the following 4 quality $652 million less than the benchmark in total, although domains: patient experience measures (e.g., getting some ACOs spent more than their benchmark and some timely care), care coordination and patient safety (e.g., less. Relative savings, by this definition, were less than readmissions, screening for risk of falls), preventive health 1 percent of the benchmark in each year, although this (e.g., influenza immunization), and at-risk populations number is slowly increasing. Medicare then paid ACOs (e.g., depression remission at 12 months). (See online that saved enough to entitle them to share in savings Appendix 8-A, available at http://www.medpac.gov, for (listed as “paid to ACOs” in the table), which is shown the full list of ACO quality measures.) The measures as a negative number in the next row, for example, –$701 are reported through a combination of claims and million in 2016. Some ACOs that were in Track 2 and administrative data, a CMS-provided web interface Track 3, which are two-sided models, had actual spending designed for capturing ACO-reported clinical quality greater than their benchmark and had to share that loss measure data, and the ACO Consumer Assessment of with Medicare. They paid Medicare the amount shown ® Health Care Providers and Systems patient experience in the next row (“paid back to CMS”), for example, $9 survey. million in 2016. The net amount is the sum of relative savings, the amount paid to ACOs as shared savings, and In each ACO’s first performance year, the quality score the amount paid back to Medicare by ACOs as shared is based only on whether the ACO completely and losses. For 2016, this net amount was –$39 million. accurately reported quality data. In the ACO’s second and June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 219

242 TABLE Summary financial results of MSSP ACOs relative to benchmarks 8–3 2015 2014 2013 2016 Dollars Dollars Dollars Dollars Percent (in millions) (in millions) (in millions) Percent (in millions) Percent Percent 100.0% $42,499 $52,885 100.0% $73,298 100.0% $81,377 Benchmark 100.0% Actual Part A and Part B spending 42,265 99.5 52,594 99.0 72,868 99.4 80,725 99.2 Relative savings 0.5 291 0.6 429 0.6 652 0.8 234 Paid to ACOs –316 –0.7 –341 –0.6 –646 –0.9 –701 –0.9 Paid back to CMS 9 4 0.0 0 0.0 0 0.0 0.0 –0.1 –78 Net –0.1 –0.3 –216 –0.1 –50 –39 Note: MSSP (Medicare Shared Savings Program), ACO (accountable care organization). The number of ACOs was 220 for 2013, 333 for 2014, 392 for 2015, and 432 for 2016. There were originally 433 MSSP ACOs in 2016, but CMS reported data for only 432 ACOs. “Relative savings” is defined as the difference between the benchmark and the actual spending. ”Net” is the sum of relative savings and amounts paid to ACOs and paid back to CMS. Components may not sum to totals due to rounding. MedPAC analysis of CMS MSSP ACO public use files. Source: It may not seem logical that shared savings payments to all losses are borne by the program. For example, under ACOs can exceed total relative savings, and they cannot for this model, if one ACO had savings of $1 million and the any individual ACO. However, under the Track 1 MSSP other had losses of $1 million, Medicare would pay shared model’s one-sided risk, if actual payments exceed the savings of $500,000 to the first and collect nothing from the benchmark, the ACO does not share losses with Medicare— second; thus, relative savings would be zero and the shared TABLE Summary financial results of MSSP ACOs relative to benchmarks, by track, 2016 8–4 One-sided model Two-sided models Track 1 Track 2 Track 3 Dollars Dollars Dollars Percent (in millions) Percent (in millions) Percent (in millions) Benchmark 100.0 % $688 100.0 % $3,971 100.0 % $76,718 Actual Part A and Part B spending 76,177 99.3 93.9 3,901 98.3 647 Relative savings 0.7 42 6.1 69 1.7 541 Paid to ACOs –613 –0.8 –23 –3.4 –64 –1.6 Paid back to CMS 0.0 0 0.0 9 0.2 0 Net –72 –0.1 18 2.7 14 0.4 Note: MSSP (Medicare Shared Savings Program), ACO (accountable care organization). In 2016, the number of ACOs was 410 in Track 1, 6 in Track 2, and 16 in Track 3. There were originally 433 MSSP ACOs in 2016, but CMS reported data for only 432 ACOs. “Relative savings” is defined as the difference between the benchmark and the actual spending. ”Net” is the sum of relative savings and amounts paid to ACOs and paid back to CMS. Components may not sum to totals due to rounding. Source: MedPAC analysis of CMS MSSP ACO public use files. Medicare accountable care organization models: Recent performance and long-term issues 220

243 FIGURE Medicare margins... X-X FIGURE Distribution of MSSP ACO savings and losses, 2016 8–1 180 160 140 120 100 80 Number of ACOs 60 40 20 0 More than 10% Less than 10% 2% to 5% –2% to 2% 5% to 10% –10% to –5% –5% to –2% Relative losses or savings MSSP (Medicare Shared Savings Program), ACO (accountable care organization). Note: Source: MedPAC analysis of CMS MSSP ACO public use files. Note: Note and Source are in InDesign. Source: There is variation in reported relative savings or losses savings payments would be $500,000. On net, the program across MSSP ACOs. Much of the savings and losses would have paid out $500,000 more than the amount could be the result of random variation. As shown in predicted by the benchmarks, and we would assess that result as a net relative loss to the Medicare program. Figure 8-1, 169 of the 432 of ACOs (almost 40 percent) had savings or losses of 2 percent or less. However, some Notes about this graph: The difference between one-sided and two-sided models had significantly greater savings or losses. Among the • Data is in the datasheet. Make updates in the datasheet. is illustrated in Table 8-4, which shows the performance in 83 ACOs with reported savings of over 5 percent, most 2016 of the ACOs in Track 1, the one-sided model, and the are located in areas of high service use. For example, • WATCH FOR GLITCHY RESETS WHEN YOU UPDATE DATA!!!! ACOs in Track 2 and Track 3, the two-sided models. 20 of these ACOs with savings over 5 percent served • The column totals were added manually. beneficiaries in Florida, and 12 served beneficiaries in For Track 1 ACOs, the amount paid to ACOs in shared • I had to manually draw tick marks and axis lines because they kept resetting when I changed any data. Texas. These data are not surprising in light of our 2016 savings bonuses ($613 million) exceeded the amount report finding that a market’s historical level of service use • I can’t delete the legend, so I’ll just have to crop it out in InDesign. saved relative to the benchmarks ($541 million), is the best predictor of reported ACO savings (Medicare resulting in spending by the program exceeding • Use direct selection tool to select items for modification. Otherwise if you use the black selection tool, they will reset to graph Payment Advisory Commission 2016d). That analysis and expectations by $72 million. In contrast, because default when you change the data. its findings are discussed briefly below. Track 2 and Track 3 ACOs share in losses, these ACOs • Use paragraph styles (and object styles) to format. produced net savings for the Medicare program in 2016 Factors contributing to MSSP ACO performance • Data was from: R:\Groups\MGA\data book 2007\data book 2007 chp1 relative to the benchmark (2.7 percent and 0.4 percent, Using 2014 data, we analyzed the contribution of three respectively). All Track 2 ACOs generated savings selected factors that might contribute to ACO performance relative to the benchmark, and 69 percent of Track 3 relative to benchmarks: ACO type (hospital based, primary ACOs generated savings (11 of 16 ACOs). care based, or multispecialty practice based); size of the June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 221

244 TABLE ACOs with the highest price-adjusted benchmarks were more likely to 8–5 generate net savings to Medicare based on CMS’s benchmarks, 2016 As a share of quintile’s benchmark: Percent of ACOs: Price-adjusted Achieving mean Net savings Receiving Shared per capita savings to Medicare shared relative to savings ACO based on CMS’s Quintile benchmark payments benchmarks savings benchmark 1 (lowest price-adjusted benchmark) –1.3 % 0.2% $7,911 38.0% 11.4% –1.5 19.0 0.3 40.5 8,933 2 0.4 3 9,733 55.7 22.8 –0.1 40.5 0.4 10,511 4 60.8 1.1 2.0 77.2 13,160 59.5 2.3 5 (highest price-adjusted benchmark) Note: ACO (accountable care organization), MSSP (Medicare Shared Savings Program). Benchmarks in the second column have been price adjusted using CMS county- level standardized prices from 2015. Savings presented in the other columns are based on CMS’s benchmarks. The last column is the net of relative savings minus the amount paid to ACOs as shared savings, plus the amount paid back to CMS as shared losses. Data exclude 38 ACOs serving beneficiaries in multiple states that do not share a border (e.g., an ACO serving beneficiaries in both New York and California). MedPAC analysis of CMS MSSP ACO public use file. Source: ACO; and the historical level of service use in the ACO’s benchmarks) and separated the ACOs into quintiles based 6 markets. Because these variables are all correlated to some on the price-adjusted benchmarks. In Table 8-5, those degree, we evaluated them in a multivariate model. We ACOs with the highest price-adjusted benchmarks are used service use rather than spending because spending in the fifth quintile, while those with the lowest price- includes service use and price. Service use (relative to adjusted benchmarks are in the first quintile. When prices the national average) is something that the ACO could are standardized, we found that ACOs with the highest theoretically control; price is outside of the ACO’s control price-adjusted benchmarks—indicating higher levels and is instead a result of Medicare payment policy. The of historical service use—were more likely to achieve common practice of assuming that the ACO’s benchmark savings relative to the benchmark and earn shared savings is a good proxy for service use is a poor assumption. Our payments. Furthermore, ACOs with higher price-adjusted analysis found that: benchmarks were more likely to generate net relative savings for the program. historical service use in the area where an ACO’s • beneficiaries live is the factor that best explains These results are not surprising. ACOs with benchmarks savings relative to benchmark performance for ACOs; exhibiting high historical service use tend to have more service use to reduce; thus, they have more opportunities ACO size (10,000 or fewer beneficiaries) and southern • to generate savings. This tendency is highlighted by location also have some statistically significant results for ACOs in the highest quintile of price-adjusted explanatory value; and benchmarks (approximated service use): Over 77 percent of these high-use ACOs achieved savings relative to their the ACO’s size may have a larger effect on its odds • benchmarks, and almost 60 percent received a shared of financial success than its type—that is, whether savings payment. In contrast, only about 11 percent of the ACO is formed around a primary care practice, ACOs with the lowest level of price-adjusted benchmark multispecialty practice, or hospital. received shared savings. Similarly, shared savings Using 2016 performance data, we find there continues payments were 2.3 percent of the benchmark for ACOs to be a relationship between service use and MSSP with the highest benchmarks, and the implied net relative performance (Table 8-5). We price adjusted the 2016 savings for the program (total savings minus shared ACO benchmarks to approximate historical service use savings payments to ACOs) was 2.0 percent. The program (that is, we removed regional pricing differences in the lost 1.3 percent of the benchmark for ACOs with the Medicare accountable care organization models: Recent performance and long-term issues 222

245 TABLE Summary financial results of Pioneer ACOs relative to benchmarks 8–6 2013 2012 2014 2015 2016 Dollars Dollars Dollars Dollars Dollars (in (in (in (in (in millions) Percent millions) Percent millions) Percent millions) Percent millions) Percent Benchmark $3,381 $7,598 % 100.0 % 100.0 $5,490 % 100.0 $6,931 % 100.0 $7,142 % 100.0 Actual Part A and 7,507 Part B spending 6,811 3,320 99.3 5,453 98.2 98.7 7,046 98.8 98.0 91 Relative savings 1.8 0.7 37 1.7 120 1.4 96 61 1.2 –37 –68 –1.0 Paid to ACOs –82 –1.1 –77 –0.6 –34 –1.2 –1.0 Paid back to CMS 2.5 0.0 0 0.0 2 0.1 0.2 11 9 0.0 47 0.7 5 16 0.1 24 39 0.7 0.2 0.6 Net ACO (accountable care organization). The number of Pioneer ACOs was 32 for 2012, 23 for 2013, 20 for 2014, 12 for 2015, and 8 for 2016. “Relative Note: savings” is defined as the difference between the benchmark and the actual spending. ”Net” is the sum of relative savings and amounts paid to ACOs and paid back to CMS. Components may not sum to totals due to rounding. MedPAC analysis of CMS Pioneer ACO quality and financial results, performance years 1–5. Source: lowest benchmarks. Although, within each quintile, some Pioneer ACOs met the quality reporting requirement. ACOs achieved savings and others incurred losses relative Like the MSSP ACOs, they also had high quality scores, to their benchmark, the general trend was that ACOs’ ranging from 89 percent to 96 percent. We reviewed relative savings were positively correlated with higher changes over time in some of the patient experience and service use. population-based outcome measures that the Commission supports. The eight ACOs that participated in all five Pioneer ACOs generally performed well on years of the Pioneer program had consistently high cost and quality metrics patient experience results. On average, these Pioneer ACOs showed some meaningful improvement in two The Pioneer ACO demonstration was the first ACO design measures: health promotion and education (improvement tested in Medicare, and it was focused on organizations of almost 5 percent) and health/functional status (3 percent that had some experience in taking risk. It started with 32 improvement). All but one of the ACOs improved their ACOs in 2012 and continued through 2016. No ACOs hospital readmissions rates. were allowed to join the demonstration after it started, but participating ACOs were allowed to leave, so the Pioneer performance relative to benchmarks number of ACOs decreased as time went on; by the final (relative savings) year of the program, only eight ACOs remained. The In the final year of the demonstration, there were Pioneer demonstration was judged to be successful in 8 remaining Pioneer ACOs serving nearly 270,000 controlling cost and increasing quality by the CMS Office beneficiaries. Those remaining ACOs generated savings of the Actuary and was certified for expansion. Many of relative to their benchmarks, with a net relative savings of the lessons learned in the Pioneer demonstration (e.g., $24 million in 2016 (Table 8-6). prospective attribution and allowing ACOs to share in a larger portion of savings) were used when designing the The relative savings percentage, with and without taking Next Generation ACO program and Track 3 of the MSSP. into account shared savings, increased over the first three years, followed by lower savings in the fourth year. Two Pioneer quality factors may partially account for this trend. First, ACOs In the Pioneer program, an ACO’s quality score that stayed in the program tended to be more successful determined its savings/losses sharing rate. In 2016, all June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 223

246 TABLE Summary financial results of Next Generation ACOs relative to benchmarks, 2016 8–7 2016 Dollars (in millions) Percent Benchmark % 100.0 $5,149 Actual Part A and Part B spending 5,101 99.1 Relative savings 0.9 48 Paid to ACOs –58 –1.1 0.4 20 Paid back to CMS Net 0.2 10 Discount 53 1.0 63 1.2 Total relative savings Note: ACO (accountable care organization). There were 18 Next Generation (NextGen) ACOs in 2016. “Relative savings” is defined as the difference between the benchmark and the actual spending. Benchmarks for NextGen ACOs are constructed with a built-in discount—an ACO-specific decrease to the benchmark—to ensure savings for the program. ”Net” is the sum of relative savings and amounts paid to ACOs and paid back to CMS. Source: MedPAC analysis of CMS Next Generation ACO quality and financial results, performance year 1. than those that left. Hence, relative savings appeared to an additional two years. The NextGen demonstration increase in subsequent years as the unsuccessful ACOs qualifies as an A–APM. It has a few differences that dropped out. Second, the decrease in relative savings in distinguish it from the MSSP and Pioneer demonstrations, 2015 was likely due to the benchmarks being rebased. including higher risk sharing, new benchmark Rebasing takes into account any success achieved in the methodology, multiple payment models, and beneficiary previous years. If relative savings were achieved in the first engagement tools. The text box on the NextGen three years, the rebased benchmark will decrease before demonstration (pp. 226–227) summarizes these provisions. trending, making it more difficult to achieve relative Performance of NextGen ACOs (relative savings) savings in future years. This issue of how much rebasing should take into account ACOs’ past success in controlling There were 18 NextGen ACOs in performance year (PY) spending is discussed later in the chapter. 1 (2016); Table 8-7 shows summary financial results for 2016. Actual spending was less than the aggregate In 2016, four more ACOs left the Pioneer demonstration, benchmark, resulting in relative savings of $48 million and the remaining ACOs generated relative savings. This (0.9 percent). After taking into account payments for result is partly because these ACOs’ per capita benchmark shared savings and losses, there was net relative savings between 2015 and 2016 increased significantly. Four of $10 million (0.2 percent). However, the benchmarks for ACOs had an increase of 10 percent or more. Even so, NextGen ACOs are constructed with a built-in discount— after subtracting shared savings paid to the ACOs, the an ACO-specific decrease to the benchmark—to ensure Medicare program saw net relative savings of less than 1 savings for the program (see the text box on the NextGen percent of the benchmark. demonstration, pp. 226–227, for more information on the discount). Taking into account the discount, the Next Generation ACOs have performed well demonstration saved $63 million (1.2 percent) relative to on cost and quality metrics the benchmark. The three program performance years for the Next Generation (NextGen) demonstration are 2016 to 2018, The ACOs varied in performance. Eleven NextGen ACOs with an option for ACOs to extend their participation for had savings ranging from 0.1 percent to 4.1 percent, and Medicare accountable care organization models: Recent performance and long-term issues 224

247 TABLE Summary financial results of ESRD Seamless Care 8–8 Organizations relative to benchmarks, 2016 2016 Dollars (in millions) Percent Benchmark % $1,415 100.0 Actual Part A and Part B spending 1,340 94.7 75 Relative savings 5.3 Paid to ESCOs –51 –3.6 0 0 Paid back to CMS Net 24 1.7 Note: ESCO (ESRD (End-Stage Renal Disease) Seamless Care Organization). There were 13 ESCOs in 2016. “Relative savings” is defined as the difference between the benchmark and the actual spending. ”Net” is the sum of relative savings and amounts paid to ESCOs and paid back to CMS. MedPAC analysis of CMS ESCO quality and financial results, Performance Year 1. Source: the other seven had losses ranging from 0.1 percent to 2.6 ESCOs are a good test case for ACOs. The population is percent. Because 2016 was the first year of the NextGen well defined and has a chronic condition that dominates ACOs, any ACO that fully and accurately reported quality their care. Most beneficiaries on dialysis are treated data received a 100 percent score for quality; all NextGen at a dialysis facility three times a week and see their ACOs received 100 percent in 2016. nephrologist at least monthly. Thus, the ESCO has many opportunities to communicate with its patients and ESRD Seamless Care Organizations coordinate their care, and attribution should be clear. As part of the Comprehensive ESRD Care (CEC) Model, ESCO quality nephrologists, dialysis clinics, and other providers can join The measure set for the CEC currently includes 11 together to create ESCOs, which are ACO-like models process measures (e.g., advance care plan, influenza for the ESRD population. Similar to other ACO models, immunization), 1 outcome measure (i.e., standardized ESCOs are responsible for their attributed population’s mortality ratio), and 6 patient experience measures based quality and financial outcomes, with larger ESCOs liable on the In-Center Hemodialysis Consumer Assessment of for shared losses. ® Healthcare Providers and Systems survey. In the first ESCOs have performed well on cost metrics year of the program, all 13 ESCOs received full credit for (relative savings) and average on quality metrics the quality score because they completely and accurately reported data to calculate quality measure results. Analysis There were 13 ESCOs in PY1 (2016). All 13 produced of the 2016 results recently released by CMS shows savings relative to their benchmarks, with 12 ESCOs that the ESCOs’ patient experience results are around producing enough savings to earn shared savings the national average for dialysis facilities (e.g., rating of payments. These shared savings payments ranged from kidney doctors, rating of dialysis center). $1 million to $12 million. Quality in PY1 was essentially pay for reporting, so all ESCOs that completely and Beginning in the second year of the program (2017), accurately reported quality data received a quality score of each ESCO earns quality points on a sliding scale based 100 percent. In total, the demonstration saved 1.7 percent on its performance compared with a national benchmark relative to the benchmark (Table 8-8). or its improvement from its previous year results. The June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 225

248 Next Generation ACO demonstration: Key provisions he Next Generation (NextGen) accountable Shared Savings Program ACOs and is intended to care organization (ACO) demonstration builds promote savings, better reward ACOs that are already efficient, and provide certainty as to the benchmark on CMS’s experience with previous ACOs but T 7 at the beginning of the year. has a few differences, including higher risk sharing, Benchmarks for the new benchmark methodology, and new beneficiary first three performance years are calculated based on engagement tools. historical expenditures as in the other ACO programs, but the baseline for calculating the benchmark will come from one year of data (2014) instead of three Risk sharing years of data. The baseline 2014 expenditure data The NextGen program allows for higher risk sharing will then be trended based on regional projections for ACOs; Arrangement A allows ACOs a shared for the current year and risk adjusted. After trending savings (or loss) rate of up to 80 percent, and and risk adjustment, the benchmark is discounted; the Arrangement B ACOs can have a sharing rate of discount can be thought of as an automatic decrease up to 100 percent. Savings and losses are shared at to the benchmark, making it slightly more difficult to first dollar instead of requiring an ACO to exceed a generate savings. The size of the discount differs for minimum savings or loss rate. There is also a limit on each ACO because the discount is adjusted to take into shared savings or losses for the ACO: 15 percent of the account both a national and regional efficiency ratio. benchmark. ACOs that are more efficient than their market or the nation will receive a more favorable (smaller) discount Benchmarks to their benchmark. The intent of this approach is to The prospective benchmark calculation for the rectify previous benchmarking methods that in some NextGen ACOs differs from Pioneer and Medicare sense penalized already efficient providers. (continued next page) benchmarks, the estimated savings in the various studies total points earned for each measure is multiplied by the differ from the relative-savings computations that are used measure weight and summed to produce the ESCO total when CMS distributes shared savings. We discuss how quality score, which is used to determine the ESCO’s various savings estimates compare with the savings CMS eligibility for shared savings. Data are not yet available for has computed using administratively set ACO benchmarks. 2017. Savings relative to benchmarks and other estimates of savings can differ ACO quality and financial performance Savings relative to CMS-constructed benchmarks and results according to other researchers other estimates of ACO savings can differ because CMS constructs benchmarks to fulfill certain policy goals. For In this section, we discuss estimates from the literature example, in our early work on ACOs, we maintained of how much ACOs have saved the Medicare program. that the appropriate trend for the benchmark should be Each study’s estimate depended on the choice of the national increase in FFS spending stated in absolute counterfactual, meaning the study’s estimation of what dollar terms and that the benchmark should be stated spending would have been for the beneficiaries attributed in standardized dollars (Medicare Payment Advisory to ACOs in the absence of the ACO. The studies often Commission 2009). The rationale for that design was used a comparison group to determine the counterfactual. that an area that had historically low service use would Because the studies’ counterfactuals differ from the ACOs’ see a relatively large trend increase, and one that had Medicare accountable care organization models: Recent performance and long-term issues 226

249 Next Generation ACO demonstration: Key provisions (cont.) ACOs to pay claims for services provided by ACO In addition to the prospective benchmark calculation, participants that have written agreements with the NextGen ACOs also have the opportunity to choose ACO. CMS will continue to pay claims to other one of four ways to receive payment from CMS: providers and reconcile payments with the NextGen standard fee-for-service (FFS), FFS and infrastructure ACO’s target after the year is complete. payments, population-based payment (PBP), and (starting the second year) partial capitation. Under Beneficiary engagement the FFS and infrastructure option, ACOs receive their usual FFS payments and an additional payment to be NextGen ACOs are designed to focus on greater put toward infrastructure. At the end of the year, these beneficiary engagement by allowing beneficiaries infrastructure payments are subtracted from the savings to align themselves with the ACO and providing an ACO would receive or are added to the loss amount incentives for using ACO services. Incentives can an ACO owes. The PBP option reduces FFS claims include reward payments to beneficiaries for using by a percentage and then pays ACOs this reduction in ACO-affiliated providers and allowing a more flexible per beneficiary per month (PBPM) payments. ACOs Medicare benefit, such as covering skilled nursing then receive both PBPM payments and reduced FFS facility stays without a prior three-day hospitalization. payments. In the final option, partial capitation, CMS Beneficiaries will be able to align with an ACO by estimates expenditures for a given ACO on a PBPM filling out a form that confirms that they use a specific basis, and then participating ACOs receive PBPM provider or practice. This voluntary alignment process payments at the start of each month that cover the began in 2016, and beneficiaries who submitted an expected cost of ACO-aligned providers. Choosing alignment form were added to the prospective list of the partial capitation option places responsibility on beneficiaries starting in performance year 2 (2017). ■ historically high service use would receive a relatively spending for a control group. For instance, one study used low trend increase. Thus, our option would not reward a 20 percent sample of beneficiaries to compare changes areas of the country with already high service use. Actual in spending for beneficiaries in ACOs with changes in policy kept the national trend, but spending was not spending for a group of beneficiaries served by non-ACO stated in standardized dollars. In other words, the trended providers in ACO service areas (McWilliams et al. 2016). benchmark was not designed to necessarily best predict Under this scenario, McWilliams estimated that MSSP spending for an area’s beneficiaries but rather to meet the net savings in 2014—including bonus payments paid to goal of being equitable across the country. While CMS’s ACOs—were $287 million, or 0.7 percent of spending benchmark is designed to fulfill certain policy goals, other for ACO beneficiaries (McWilliams 2016a, McWilliams groups have used other methods to provide an alternative 2016b). assessment of whether ACOs save Medicare money. These Using the same methodology to analyze the performance alternative assessments use a counterfactual—that is, what of Pioneer ACOs, McWilliams and colleagues estimated spending on the beneficiaries in the ACO would have been that Pioneer ACOs saved $118 million (1.2 percent of in the absence of the ACO—to estimate savings. spending for ACO beneficiaries) relative to expected spending in their first year (2012), or $42 million (0.3 Savings estimates in literature (program percent of spending) when bonus payments paid to ACOs savings) are subtracted from total savings (McWilliams et al. 2015). To determine what spending would have been for beneficiaries in the absence of an ACO, most studies relied L & M Policy Research, the group CMS contracted with on comparing changes in ACO spending with changes in to formally evaluate the Pioneer ACO program, estimated June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 227

250 $280 million (3.7 percent of spending) in savings for the to improve the quality of care received while generating first year of the Pioneer ACO program (L & M Policy savings (Government Accountability Office 2015, Office of Inspector General 2017, Pham et al. 2014). While these Research 2015, Nyweide et al. 2015, Office of the Actuary 8 savings may appear modest, they are more than most care 2015). The comparison group in L & M’s analysis coordination demonstrations have achieved, including the included all “FFS Medicare beneficiaries who are not most recent Comprehensive Primary Care initiative (Dale aligned or assigned to a Medicare ACO in the Pioneer 9 et al. 2016, Nelson 2012). ACO’s local, or ‘near’ market.” This analysis did not restrict the comparison group to ACO-attributable FFS Spillover estimates beneficiaries. Thus, savings might be overstated because, to be attributable to an ACO, beneficiaries had to have a In addition to the direct savings from reduced spending primary care visit. FFS beneficiaries who did not have a on beneficiaries in ACOs, indirect savings of two kinds primary care visit in the baseline year were only in the (spillover and reduced MA benchmarks) are also possible, control group, not in the ACO. The problem is that these according to researchers. McWilliams’s (2016) research individuals tend to have low baseline spending and high on MSSP ACOs considers potential additional savings spending growth, which could have made the comparison accrued through spillover effects. Under this theory, group appear to grow faster than it would have if it ACO providers furnish better coordinated care to all included only ACO-attributable beneficiaries. their patients, thus “spilling over” to their non-ACO FFS beneficiaries. The magnitude of the spillover effect is Another analysis examined the combined performance expected to be modest and has not been tested empirically. of both MSSP and Pioneer ACOs in 2012 and 2013. Another indirect benefit could result from reduced MA It created a control group by utilizing “a random 40% benchmarks over time, as a county’s FFS spending on sample . . . of continuously enrolled fee-for-service which MA benchmarks are based is reduced. This effect beneficiaries with at least 1 evaluation and management presupposes savings from ACOs. In fact, spending in visit in a calendar year” (Colla et al. 2016). That analysis some counties with MSSP ACOs could have increased, found that, together, MSSP and Pioneer ACOs saved particularly if shared savings payments are included as approximately $592 million (about 1.1 percent of the FFS spending, and could result in an increase in MA benchmark) in 2013. benchmarks, although the magnitude would probably be small in either direction. When CMS’ Office of the Actuary (OACT) certified that expanding the Pioneer ACO Model would reduce spending Sources of savings for the program, it conducted a market-level analysis Research shows that how ACOs generate savings (Office of the Actuary 2015). OACT’s analysis compared does not necessarily align with preconceptions. Early FFS spending growth in markets with heavy MSSP and in the development of ACOs, some speculated that Pioneer penetration with markets that had few ACOs. For savings would accrue through better coordinated care markets with low rates of ACO penetration, FFS per capita and subsequent reductions in unnecessary inpatient spending decreased by 0.3 percent from 2011 to 2014, capacity, tests, imaging services, and post-acute care whereas for markets with high rates of MSSP ACOs, per (PAC) use (Fisher et al. 2007). Data from the Alternative capita spending decreased 1.2 percent, and in markets with Quality Contracts (AQCs), a commercial predecessor high rates of Pioneer ACOs, it decreased by 2.1 percent. to Medicare’s ACOs, indicated that savings could OACT’s findings that FFS spending growth decreased be generated through these avenues, specifically by more in Pioneer ACO markets, taken in conjunction with decreasing utilization of procedures, imaging, and tests Pioneer ACOs’ ability to save money relative to their and by referring patients to less expensive providers benchmarks and L & M’s positive evaluation results, led (Song et al. 2014). While AQCs were successful in OACT to certify that Pioneer ACOs were successful in these areas, Medicare ACOs—especially those in the reducing spending. MSSP—have largely created savings by decreasing Given the CMS benchmarking analyses, studies in the PAC utilization. A recent study by McWilliams and literature, and the work by OACT, it appears the ACO colleagues found that, while MSSP ACOs were scaling programs have generated savings estimated in the 0 back inpatient capacity slightly, they were generating a percent to 2 percent range. ACOs also generally appear higher proportion of their savings by decreasing PAC Medicare accountable care organization models: Recent performance and long-term issues 228

251 utilization—specifically SNF use (McWilliams et al. Telehealth 2017b). Pioneer ACOs likewise reduced PAC utilization The BBA of 2018 expanded the use of telehealth for two- to generate savings, in addition to having lower rates of sided-risk ACOs with prospective attribution. Under the inpatient stays, imaging, tests, and procedures, similar to BBA of 2018, qualifying ACOs are no longer subject to the AQCs (L & M Policy Research 2015, McWilliams et a geographic limitation on the telehealth originating site al. 2014). and are allowed to use the beneficiary’s residence as an originating site. Currently, some ACO demonstrations Additionally, while many expected ACOs to focus on allow for expanded use of telehealth (e.g., NextGen), but coordinating care for high-risk patients to save money, a ACOs are required to submit a waiver to utilize the benefit. recent study found that those savings have yet to occur In its recent telehealth discussions, the Commission has in the MSSP program. When comparing ACO savings in supported the expanded use of telehealth for risk-bearing 2014 for high-risk and low-risk patients, savings between ACOs because the ACOs are at risk for cost (unlike the two groups were relatively similar for the cohort of providers in traditional FFS) (Medicare Payment Advisory ACOs that began in 2012 (McWilliams et al. 2017a). For Commission 2018). ACOs that entered the program in 2013, more savings were accrued for low-risk patients than high-risk patients. Expanded prospective attribution Furthermore, the study found MSSP ACOs did not ACOs in retrospective attribution models (i.e., MSSP reduce hospitalizations for ambulatory care–sensitive Track 1 and Track 2) beginning or renewing their conditions. agreements on January 1, 2020, and beyond can choose to have their beneficiaries assigned prospectively. The Commission has long been in support of prospective New tools to allow ACOs to manage attribution because it gives providers more certainty at the care start of the performance year about which beneficiaries are in their ACOs and allows for better coordination of While the ACO program has grown in numbers of care throughout the year (Medicare Payment Advisory ACOs and beneficiaries, it continues to evolve. The Commission 2015a, Medicare Payment Advisory recently passed Bipartisan Budget Act of 2018 (BBA Commission 2014a). However, benchmarks for ACOs of 2018) included several changes to Medicare’s ACO changing attribution will need to be recomputed to programs, including incentives for beneficiaries to reflect the beneficiaries in the baseline who would have see ACO providers, use of telehealth, and beneficiary been attributed under prospective attribution versus assignment. Many of these changes are consistent with retrospective attribution. past Commission positions on ACOs. These changes are expected to make the program more attractive to providers Attribution based on voluntary identification by enhancing the tools they have to improve quality and by beneficiaries reduce costs. According to the BBA, the Secretary will also establish a process by which beneficiaries will be informed ACO Beneficiary Incentive Program of their option to voluntarily identify a principal Starting no later than 2020, the Secretary is to establish primary care provider. If the designated primary care an ACO Beneficiary Incentive Program, which would provider participates in an ACO, the beneficiary will be allow ACOs to pay beneficiaries up to $20 for each automatically attributed to that ACO. A similar process qualifying primary care visit with an ACO provider. is already in place for the MSSP. Currently, beneficiaries ACOs will have to apply to run such a program, which can log on to MyMedicare.gov and designate a clinician will be available only to two-sided-risk ACOs. Incentive as their “primary clinician” who is responsible for payments will not factor into an ACO’s benchmark, and coordinating their overall care (Centers for Medicare & incentive payments could be funded through previous Medicaid Services 2017a). Clinicians in ACOs have some shared savings payments. The Commission has previously latitude to encourage beneficiaries to designate them as supported giving ACOs more options for incentivizing their primary clinician. However, to date it appears that beneficiaries to use their ACO providers so that ACOs few beneficiaries are being aligned under this mechanism. have more leverage in coordinating their beneficiaries’ care (Medicare Payment Advisory Commission 2014b). June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 229

252 a hospital has an incentive to increase the volume of Long-term issues for Medicare ACOs Medicare admissions as long as the payment for an additional patient exceeds that patient’s variable cost Medicare ACOs were created to help moderate the growth and the hospital has excess capacity. (In our March 2018 in Medicare spending and improve quality of care for report, we found that the average hospital occupancy rate beneficiaries by giving providers greater responsibility was 66 percent and that variable costs were 8 percent for costs and quality. ACOs have grown rapidly (about a less than Medicare payments. Therefore, most hospitals third of Medicare FFS beneficiaries are now in ACOs), and have an incentive to increase the volume of Medicare several new initiatives have been designed to expand ACOs. admissions (Medicare Payment Advisory Commission Performance to date shows high quality being maintained, 2018).) At the same time, ACOs have an incentive to keep some savings relative to benchmarks, and slightly greater Medicare spending for their attributed beneficiaries below savings relative to what Medicare spending would have a target amount—their benchmark. If they do so, they been without ACOs. However, several issues confront can share savings with Medicare. One way to reduce or Medicare ACOs—particularly as they transition to models constrain spending is to reduce inpatient admissions. Thus, with two-sided risk—that will need to be resolved for the it would appear that the incentives for hospitals and ACOs program to be successful in reaching its goals. are in conflict. Because two-sided risk models are more likely to result in While ACOs may eventually have some effect on savings for the Medicare program, the following questions admissions, it appears to date that ACOs have not caused a arise: Can hospitals and ACOs viably coexist and, if so, large reduction in inpatient admissions, despite rhetoric to what does that mean for ACOs moving to two-sided risk? the contrary. We examined changes in inpatient admissions Should asymmetric models be continued even if they and considered why the trends should not be surprising. present the risk of excess spending for Medicare? What Assuming trends continue, opportunities for cooperation approaches to setting benchmarks should be used? What between ACOs and hospitals may exist, and concerns method should be used to distribute the 5 percent bonus about the conflicting incentive may be less germane. for clinicians participating in A–APMs? What relationship will specialists have with ACOs? Are ACOs a path to MA Reducing post-acute care (not inpatient care) is the plans or are they an end in themselves? primary source of ACO savings In interviews we conducted in 2012 and 2013, many Are hospitals a viable participant in ACOs? ACO leaders expected to generate savings by reducing In general, hospitals have greater financial resources the volume of inpatient care. In particular, physician than most clinician groups, which can make accepting leaders of ACOs saw the hospital as a key driver of downside risk easier for an ACO with a hospital spending, and reducing unnecessary hospital admissions participant than an ACO without one. In fact, about half of as a key source of savings. However, a review of the risk-bearing MSSP ACOs (Track 1+, Track 2, and Track literature finds that reducing PAC has been a much 3) list hospitals as participating providers. Thus, it may be bigger source of ACO savings than reducing inpatient important for hospital-based ACOs to thrive to make two- admissions (McWilliams et al. 2017a, McWilliams et sided ACO models more available. al. 2017b). Similarly, the AQC program, a commercial ACO program, did not generate significant reductions in There is a concern, however, that hospitals may be inpatient facility fees or inpatient professional fees (Song reluctant to reduce service volumes to meet ACO spending et al. 2014). In contrast, AQC savings were generated targets because they do not want to reduce their own FFS by reducing spending on outpatient facility fees and revenue. However, the data show that ACOs with hospitals professional fees—often by shifting services to lower can meet spending targets. We examine how they are priced providers (Song et al. 2012). Thus, decreased meeting spending targets and conclude that hospital-based hospital revenues from the actions of ACOs may be due ACOs may continue to be part of the ACO landscape into to a shift of outpatient services to lower priced settings the future. rather than a decline in the number of admissions. The finding that ACOs do not cause big reductions in Conflict between hospital and ACO incentives inpatient spending is consistent with the following three It may at first appear that the incentives for ACOs and findings. hospitals conflict. In an FFS payment environment, Medicare accountable care organization models: Recent performance and long-term issues 230

253 First, in FFS Medicare, inpatient service use varies little Third, another way to examine whether MA plans by region (Medicare Payment Advisory Commission significantly reduce inpatient use is by analyzing their bids 2017b). Our analysis of claims data from 2014 found that for self-reported spending on inpatient care. We find that across 484 market areas, inpatient use for market areas at MA plans and FFS Medicare devote similar shares of their the 90th percentile of use was 1.16 times that for market overall spending to inpatient care. This finding suggests areas in the 10th percentile of use. In contrast, PAC use for that MA plans do not reduce inpatient care to a larger market areas at the 90th percentile of use was 1.88 times degree than they reduce other services on average, which that of market areas in the 10th percentile of use. Across differs from data from 20 or 30 years ago. There is some all markets, the ratio of the maximum to minimum service evidence that HMOs historically had 35 percent to 40 use was 1.49 for inpatient and 5.66 for PAC use. This percent fewer admissions per capita than indemnity plans finding suggests ACOs would have a greater opportunity or Medicare FFS (Duggan et al. 2018, Newhouse 1993). for savings by reducing spending on PAC services in high- However, those studies used data from 2003 or earlier. use areas than by reducing spending on inpatient services. Since that time, FFS discharges per capita have fallen by about 25 percent, making reductions from the lower FFS Second, we found that admission and revenue growth vary baseline more difficult. ACOs, which have fewer tools than by hospital, but ACOs and MA plans are not the driving MA plans to control admissions, should not be expected to forces. To see whether ACOs and MA plans have had a achieve greater reduction than MA plans. material effect on hospital volumes in recent years, we examined whether county-level ACO penetration in 2015, In light of these findings, it appears that the greatest MA penetration in 2015, and growth in MA penetration opportunity for ACOs to control spending is in post-acute from 2011 to 2015 were associated with reductions in care, not inpatient care. While ACOs may eventually lead to either all-payer admissions or revenue at hospitals from small reductions in inpatient use, we have not seen evidence 2012 to 2016. We also tested to see whether hospitals that to date that they materially affect hospital revenue. were in an ACO tended to have lower volume or revenue 10 Should asymmetric models be continued? growth. We add in MA penetration because, if MA penetration does not materially affect hospital inpatient One way to encourage ACOs to take on risk is to make volume, then there is little reason to expect ACOs to the models asymmetrical—that is, to make the share of materially affect hospital inpatient volume. savings greater than the share of losses or to put higher caps on savings than on losses. A policy question is Our test consisted of a linear regression in which we whether such models should be a temporary path to controlled for, among other things, population growth increase ACO participation in these models (and give and hospitals’ size. The level of ACO penetration, MA clinicians an opportunity to participate in A–APMs) or be penetration, growth in MA penetration, and whether a permanent part of the program. the hospital participated in an ACO all failed to have a statistically significant effect on the change in a hospital’s For example, the Track 1+ model has two asymmetries. total admissions or total revenue. While hospitals in First, the model has a shared savings rate of 50 percent markets with ACOs and growing MA penetration saw and a shared loss rate of 30 percent. Second, the loss cap is small declines in inpatient use, it was not higher than lower than the savings cap for all types of Track 1+ ACOs. in the average market. This finding suggests either that There are two choices for the loss cap, both of which are MA plans and ACOs have a limited impact on Medicare less than the 10 percent of the benchmark cap on gains. inpatient admissions or that hospitals are able to replace The first choice is 4 percent of the benchmark; the second lost Medicare admissions with other patients. In contrast, is 8 percent of the Medicare FFS revenue for the ACO population and hospital size were highly significant. For participants. This choice is limited to ACOs whose only each 1 percent increase in population, hospital admissions participants are clinicians or clinicians plus a small rural increased by 0.8 percent. We also found that smaller hospital. This amount will also be much less than 10 11 hospitals tended to lose discharges faster than larger percent of the benchmark. hospitals. The net finding, that admission and revenue This design gives Track 1+ ACOs certain advantages over growth vary by hospital, but ACOs and MA plans are not ACOs in the Track 1 model, despite the downside risk in the driving forces, suggests that hospitals can coexist with Track 1+ not present in Track 1. In Track 1+, providers MA plans and ACOs. are at risk for losses, but the ACOs’ clinicians are eligible June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 231

254 for the 5 percent incentive on their physician fee schedule than in Track 1 or in unconstrained FFS and could indeed (PFS) payments because these ACOs are considered A– save money for the program while possibly increasing APMs. The 5 percent incentive considerably ameliorates quality. It seems to be a popular model thus far; in 2018, the risk of being in Track 1+ because the maximum risk 55 ACOs entered the Track 1+ model. Therefore, it will in Track 1+ for ACOs with only clinicians as participants likely increase the availability of A–APMs for clinicians is 8 percent of their FFS Medicare revenue. If they to join. Whether the increased availability of A–APMs automatically get a 5 percent bonus, risk is essentially is worth the possible increased cost to the program is an limited to 3 percent of Medicare FFS revenue. If the important policy question. The Commission will track the ACO is likely to break even—that is, has a roughly equal progress of the Track 1+ model over the next few years to probability of showing a loss or a gain—we calculate that see whether the model is saving or costing the Medicare the clinicians would see more financial advantage in Track program relative to Track 1 and FFS Medicare. 1+ than in Track 1. A recent analysis by Avalere found How should benchmarks be set initially and that, in aggregate, MSSP ACOs would have fared better rebased for subsequent agreement periods? in 2016 by $966 million if they had all been in Track 1+ rather than Track 1 (Avalere Health 2018). One of the most important policy questions when designing ACO and MA payment policy is how to set the By statute, CMS can introduce other MSSP models as part benchmarks. The goal of a benchmark for an individual of permanent Medicare law if those models are estimated ACO is to create incentives to encourage the ACO’s not to increase Medicare spending relative to the Track providers to increase quality while restraining overall 1 model (CMS has done so for the Track 2 and Track 3 Part A and Part B spending. However, a benchmark that models). However, Track 1+ is a demonstration under accomplishes that goal may not be the best estimate of the authority of CMS’s Center for Medicare & Medicaid what spending for those beneficiaries would have been Innovation (CMMI), not an additional MSSP model. in the absence of the ACO. We need to know the latter Therefore, the Track 1+ model does not have to meet that to ensure that, at the national level, the ACO program is requirement, and ACOs can join even if the model increases reducing Medicare spending over the long term while 12 spending. If Track 1+ were incorporated into permanent improving quality or at least keeping it constant. Thus, Medicare law, the costs would have to be offset. to determine whether an ACO program is “working,” we need to know whether it is creating useful incentives at the It appears that Track 1+ could put the Medicare program individual ACO level and savings at the national level. at risk of financial loss if Track 1+ ACOs’ losses relative to the benchmark are greater than ACOs’ relative savings Two approaches to setting benchmarks because of the model’s asymmetries. If Track 1+ were Generically, there are two approaches to setting incorporated into permanent Medicare law, the costs may benchmarks in Medicare: regional benchmarks, as used need to be offset if performance is essentially random. in the MA program, or historical spending, as used in the If it is demonstrated that ACOs are modifying their ACO programs. For example, in MA plans, the benchmark behavior from what they would have done if not in ACOs is set based on five years of historical FFS spending in and reducing spending, then this issue will not arise. each county, adjusted for the beneficiaries’ hierarchical Currently, ACOs can be in Track 1+ for only one three- condition category (HCC) coding scores. This approach year agreement period. Policymakers must decide whether creates incentives for MA plans to devote resources to the asymmetries in Track 1+ are appropriate and whether coding, and the result has been more coding in MA plans the model is a success; if it is a success, policymakers will than in FFS Medicare. (Under this coding incentive, MA need to decide whether aspects of the model should be beneficiaries appear to be getting sicker quicker compared extended to other ACO models (or CMS should continue with beneficiaries in FFS Medicare, whose providers— the Track 1+ model). paid differently from MA plans—lack the same incentive Whether Track 1+ will cost Medicare more relative to to code their patients at the greater intensity levels.) In what spending would otherwise have been or relative to addition, coding practices across MA plans vary widely. Track 1 will depend on the ACOs’ performance. Because We have made recommendations to address MA’s higher of the possibility of sharing in losses, clinicians in Track level of coding in aggregate and the variation by plan 1+ could be more likely to succeed at controlling spending (Medicare Payment Advisory Commission 2016c). Medicare accountable care organization models: Recent performance and long-term issues 232

255 In part to get around the dependence on risk adjustment may have lower spending and lose attribution to their using HCC scores, ACOs were built on a model that ACO because their plurality of care is no longer with the looks at historical spending for a fixed group of people ACO clinician. This scenario is also consistent with our or a fixed group practice and examines how spending for findings. A consistent relationship between service use the ACO’s beneficiary population changes from one year and attribution (or loss of attribution) could be an issue. to the next. This approach incorporated the assumption One way to limit the effect of attribution on changes that the population of beneficiaries and providers in each in spending is to use prospective attribution. Under ACO would be relatively stable. However, the “churn,” prospective attribution, the year of data used to attribute or movement of beneficiaries (and, in some cases, an individual differs from the performance year data used providers) in and out of ACOs, has been larger than to evaluate spending relative to the benchmark. Therefore, anticipated, with one study finding only 66 percent were an episode of illness that results in a beneficiary being consistently assigned over two years and about 20 percent attributed to an ACO will be in a previous year and thus in of beneficiaries left the ACO each year (McWilliams et the benchmark. al. 2014). Although changes in provider participation are This preliminary analysis suggests that, although MSSP dealt with by recalculating baseline spending, churn in ACOs are to some extent controlling the spending growth attributed beneficiaries could be an issue for benchmarking for beneficiaries who are continuously attributed, there is if those who lose ACO alignment have systematically a tendency for ACOs to have beneficiaries leaving who different characteristics from those coming into alignment. have lower growth in spending and beneficiaries joining For example, those leaving the ACO could be very high who have higher growth in spending. Attribution is related cost and those entering could be very low cost, in which to service use, which could be a source of concern when case the ACO’s benchmark would need to be refined. setting benchmarks or estimating savings. Population dynamics Rebasing benchmarks In a preliminary analysis, we compared a control In our February 2015 comment letter on the MSSP ACO population with MSSP ACO-aligned beneficiaries located proposed rule, we noted a basic conflict in the benchmark- in the same metropolitan areas. We found that beneficiaries setting mechanism and in the dynamics of rebasing attributed to MSSP ACOs for two consecutive years had (Medicare Payment Advisory Commission 2015b). spending growth about 3 percent lower than beneficiaries (Rebasing is the process of setting ACO benchmarks at the who were not in an ACO in either year. We also found start of each three-year agreement period subsequent to the that beneficiaries who were attributed in the first year and first period.) lost attribution to the ACO in the second year (and thus were in an ACO for only one year) had spending growth On the one hand, if benchmarks are rebased strictly on the that was even further below the control group. Conversely, historical experience of the ACO’s patients, the benchmark those who were attributed to an ACO in the second year will incorporate the efficiencies the ACO has realized and not in the first had much higher spending growth than in the first three years and further improvements will be the control group. That is, the people who lose alignment difficult to achieve. If an ACO were in the program for to the ACO have low spending growth, and those who repeated periods, this increased difficulty could make it join have high spending growth. (We also found that less desirable for an ACO to continue with the program. MSSP ACOs do not appear to materially affect end-of-life Such a result does not seem equitable for an ACO that has spending.) Savings estimates for MSSP ACOs should be improved its efficiency—particularly if its benchmark to evaluated taking these findings into account. begin with was below the level of ambient FFS spending in its region. There are several potential explanations for these findings. For example, a beneficiary may become sick, see an ACO On the other hand, one could set benchmarks using an clinician repeatedly, and have increased spending. Because approach similar to that for MA plans (HCC-adjusted the plurality of care will now be with an ACO clinician, local FFS spending). A regional benchmark could be this case could result in the beneficiary being aligned with calculated using FFS spending, and that amount multiplied the ACO when she otherwise would not have been, and by the HCC score for each attributed beneficiary would it would be consistent with findings in our preliminary be summed to calculate the ACO’s benchmark. However, analysis. At the same time, beneficiaries who stop seeing under such an approach, ACOs would be able to calculate clinicians because their principal condition improves June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 233

256 The blending in MSSP rebasing and the NextGen discount their benchmarks in advance, and only ACOs that are adjustment are both attempts to deal with the issue of already below their regional benchmark would participate. setting benchmarks that are equitable while still creating ACOs that had spending above the regional average would incentives for savings at the ACO level and trying to not participate because they would likely have actual ensure that Medicare program spending does not increase. spending above their benchmark. Thus, efficient ACOs Efforts should continue to monitor whether ACO programs would likely receive a shared savings bonus for doing overall are saving money while maintaining or improving what they would have done anyway, and inefficient ACOs quality. It is important to remember that benchmarks will that needed an incentive to control spending would not always incorporate policy goals, such as increasing equity participate. The result would likely cost the Medicare across the nation or encouraging participation in two- program more and not improve quality appreciably. In sided-risk ACOs, and will not—and are not intended to— addition, if HCC scores were used in benchmarking, some represent the best counterfactual to ACO participation. of the same issues that have been well documented in MA would arise—with the variability in coding intensity Should the 5 percent bonus for clinicians across practices and the incentives to spend more money in A–APMs be distributed differently to on coding being the most problematic. encourage A–APM participation? One approach to this challenge is to blend historical One step to encourage clinicians to continue to expand experience and the regional average when rebasing their participation in meaningful payment reform models benchmarks. This approach is now being taken in would be to make their eligibility for the 5 percent A– MSSP when benchmarks are rebased every three years. APM incentive more certain. Under current policy, Essentially, the average of the ACO’s risk-adjusted clinicians who participate in an A–APM can qualify for expenditures over the past three years is compared with a 5 percent A–APM incentive payment established in the FFS region’s risk-adjusted expenditure average. If the MACRA. The incentive payment is applied to all of a ACO’s per capita risk-adjusted expenditures are higher clinician’s PFS revenue from the prior year. But to qualify than the regional average, the benchmark is reduced for the incentive payment, the clinician must meet either toward the regional average; if the ACO’s expenditures are the threshold for share of revenue derived through an A– lower, the benchmark is raised toward the average. This APM or for share of patients coming through the A–APM. approach rewards ACOs whose original benchmarks (i.e., The numerical threshold for share of revenue is set in the benchmarks at the start of the three-year agreement statute and increases over time. In 2019 and 2020, to be period) were below the regional average, penalizes those eligible for the 5 percent incentive, clinicians must have with original benchmarks above the regional average, at least 25 percent of their PFS revenue in an A–APM, and compresses rebased benchmarks in a market toward 50 percent in 2021 and 2022, and 75 percent in 2023 and the regional average (Centers for Medicare & Medicaid later. The “patient count” thresholds are set by CMS. CMS Services 2017b). has set lower thresholds for the patient count option of 20 percent in 2019 and 2020, 35 percent in 2021 and 2022, The NextGen program has initially taken a different and 50 percent in 2023 and later. This lower threshold approach to accounting for efficiencies and regional appears to enable a larger share of participating clinicians variation. NextGen ACO benchmarks incorporate to qualify for the bonus. a discount to the historical spending for an ACO’s beneficiaries. That discount varies in size from 0.5 percent In addition, there is an “all-payer” option starting in to 4.5 percent. A larger discount reduces the benchmark 2021, which requires CMS to determine what share of a more than a smaller discount. The size of the discount clinician’s revenue or patients is coming through A–APM- varies based on the ACO’s efficiency relative to FFS like arrangements for other payers. CMS has started the spending in its region and relative to the national average process of collecting information for the all-payer option. of FFS spending. ACOs that are efficient in comparison In the 2019 advanced notice for MA plans, CMS proposed with their region get a smaller discount, as do ACOs in a collecting from MA plan sponsors lists of clinicians and region that is efficient compared with the national average. the contracts those clinicians hold with MA plans that Over time, however, the NextGen program will also face 13 qualify as A–APM-like contracts. pressure to blend benchmarks to avoid a downward spiral in benchmark levels. In our June 2017 report to the Congress, we described a way to simplify the incentive award process (Medicare Medicare accountable care organization models: Recent performance and long-term issues 234

257 Payment Advisory Commission 2017a). The proposal thus specialists are not required for an ACO to meet the was to eliminate the threshold calculation and instead minimum number of attributed beneficiaries. Also, some apply the 5 percent A–APM incentive payment only to could be concerned that specialists would attract high-need the clinician’s PFS revenue derived from an A–APM patients to the ACO, thereby increasing its costs. However, (instead of to all of a clinician’s PFS revenue). This if the patients are high cost to begin with and are thus in proposal would greatly simplify the system and make it the historical baseline, the ACO’s benchmark will reflect more equitable. For example, under the current system, those higher costs. In fact, one could argue that those clinicians with 24.9 percent of their revenue coming beneficiaries may be the ones who could most benefit through an A–APM get no bonus, and clinicians with 25.0 from the better care coordination that the ACO is designed percent of their revenue coming through the A–APM get a to provide. 5 percent incentive bonus on all of their PFS revenue. The Our analysis of the 2016 MSSP ACO public use file proposed system would eliminate such payment “cliffs” indicates that about 60 percent of ACO-participating or discontinuities. Instead, under our proposed refinement, 15 physicians are specialists. Being on the participant the bonus would be certain because the incentive would list does not mean that a physician will share in savings depend solely on the clinician’s revenue coming through or help manage the ACO. Each individual ACO has the the A–APM, whatever that level may be. (Additionally, latitude to decide on the relationship of the physician to such a refinement would help avoid uncertainty for the ACO as to who shares savings and how much. clinicians who would be concerned they could lose the incentive payment as the threshold rises from 25, to 50, to ACOs may have an incentive to involve specialists because 75 percent in later years.) specialists who practice in a conservative, cost-effective style and avoid unnecessary testing and procedures could A benefit of this policy is that the patient count and all- help control costs and increase the quality of care for payer options would no longer be necessary and could be beneficiaries attributed to the ACO. At the same time, eliminated because, under this revised design, the bonus participating in an ACO could be attractive to specialists. is applied only to the share of revenue coming through Participating in the ACO would give the specialist the A–APM. Under the current all-payer option, CMS access to a patient’s claims history and possibly alert the must calculate the clinicians’ total revenue from all payers specialist when the patient was admitted to a hospital or and determine what share came through A–APM-like visited an emergency room. Thus, the specialist might be contracts. That determination could represent a large able to better coordinate patient care. (In the case of two- administrative burden on all parties and intrusion of the sided-risk ACOs that are A–APMs, specialists also could government into the business relationship between MA be eligible for the 5 percent A–APM bonus on their PFS plans and clinicians. revenues.) Specialists could also receive more referrals Whether the proposed approach would result in more from the ACO’s primary care clinicians if they had a or less spending is not clear. On the one hand, more relationship with the ACO. This arrangement could prove clinicians would be eligible for some payment (e.g., in mutually beneficial to both primary care clinicians and 2019 and 2020, all those with less than 25 percent of specialists. revenue through the A–APM). On the other hand, the Furthermore, there could be a role for specialty-focused actual payments for some clinicians would be lower; for ACOs. For instance, the success of ESCOs—a specialty- example, a clinician with 30 percent of revenue through focused ACO model—indicates that specialty providers an A–APM would get a 5 percent payment adjustment could develop their own ACO-like models, which could be on 30 percent of PFS revenue, not on 100 percent of PFS done by submitting a proposal to the Physician-Focused revenue. How these changes balance out would need to be Payment Model Technical Advisory Committee (PTAC). If 14 estimated. accepted by the PTAC, the model could be recommended to the Secretary as a potential new demonstration for What relationship will specialists have with CMMI, creating even more opportunities for specialists ACOs? to participate in ACO-like models. The Commission will Another concern is that specialists are not perceived monitor the relationships between specialists and ACOs as to have a role in ACOs because attribution to ACOs is the ACO models continue to evolve, and we will examine predominantly dependent on primary care visits, and whether it is possible to ascertain the level of participation June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 235

258 in ACOs by specialists and whether the degree of If MA health care spending reductions compared with • specialists’ participation affects ACOs’ performance. ACO health care spending reductions are greater than $1,100, then MA plans would be expected to be the Are ACOs only a transition step to MA? lower cost model. The ACO program is large, continues to expand, and • If MA health care spending reductions compared with continues to evolve. However, some suggest that MA ACO health care spending reductions are less than plans are the more efficient model and that, eventually, $1,100, then ACOs would be expected to be a lower ACOs should evolve into MA plans. As a matter of policy, cost model than MA. the question is whether all ACOs should be encouraged to become MA plans or whether there are circumstances The amount of service use that MA plans will be able to in which it is better for ACOs to remain ACOs (Medicare reduce relative to FFS Medicare and ACO use will depend Payment Advisory Commission 2016b). on several factors. One may be the initial level of service use and fraud in the market. Data suggest MA plans can In the past, the Commission has discussed how no one generate substantial savings in some high-use markets model is the low-cost model in all parts of the country such as Miami. However, if there is less than $1,300 of (Medicare Payment Advisory Commission 2014c). In unnecessary spending to cut, then FFS Medicare could some markets, the tools that MA plans have to manage be a lower cost model. Second, ACO savings could be service use result in substantial savings. In other markets, affected by the ACO’s providers’ position in the market. ACOs or FFS is the lower cost model. For analytical One conceptual advantage of MA plans is their ability to purposes, that report synchronized the benchmarks at lock beneficiaries into a defined provider network. If an 100 percent of FFS spending for all three models. In fact, ACO’s participants constitute the dominant health system in 2018 we estimate MA benchmarks (including quality in a market, then the ACO model with its lower costs may bonuses) will average 107 percent of FFS spending. be more efficient because the ACO should have a similar ability to control utilization. One particularly important factor is that, although MA plans have more tools to control service use, they also have However, benchmarking could still be an issue even higher administrative costs. Data from the major insurance if an ACO is in a dominant market position. Under companies indicate that, on average, administrative costs a historically based benchmark, a regionally based in MA plans are approximately $1,300 per beneficiary. benchmark (based on regional FFS spending), or a blend, Among those costs are costs for marketing, both directly an ACO with a dominant market position would have to to beneficiaries and through brokers; enrolling members; improve on its own performance over time because its negotiating with providers; paying claims; and providing benchmark will reflect its own performance. In contrast, other insurance functions, such as prior authorization. MA MA benchmarks are based on FFS spending, not MA plans also have to qualify as state-licensed insurers, which spending. Thus, MA plans do not face the issue of their could entail considerable costs and financial resources. own historical performance dictating their benchmark. In addition, MA benchmarks are adjusted so that they Our discussions with ACOs suggest their administrative are a higher percentage of FFS spending if the county costs, in contrast to those of MA plans, are close to $200 has lower FFS spending relative to the national level. In per beneficiary per year. ACOs do not have the costs of some counties, MA benchmarks are 115 percent of the advertising, enrolling, negotiating contracts, and paying Payment Basics FFS average (see the Commission’s MA claims. Their administrative costs include the expense of document, available at http://medpac.gov/-documents-/ setting up and managing the ACO, which should include payment-basics, for a fuller discussion). data analysis and reporting quality measures. However, some companies can provide those services under Thus it is not clear a priori whether ACOs are in all contract, and some ACOs are using that approach. circumstances a stepping stone to MA or should remain as ACOs. The challenge going forward is to set MA and Therefore, which model will generate greater savings ACO benchmarks in such a way that the models can depends on whether the MA plan’s reduction in spending compete and the most efficient model can gain market on medical services offsets its higher administrative cost share in each individual market. relative to an ACO’s spending and costs. There are two basic possibilities: Medicare accountable care organization models: Recent performance and long-term issues 236

259 ACO models if they are to persist in the long term. Some Conclusion issues, such as the 5 percent incentive in MACRA, could have relatively straightforward solutions, and others, such ACOs in Medicare continue to show some success in as the role of hospitals and specialists in ACOs, are more meeting their goal of high-quality care and lower costs nuanced. Challenges such as asymmetric models and relative to their benchmarks. In addition, some analysts setting benchmarks could require policymakers to decide find that their success may be understated by their whether a preference should be given to one model (MA, performance relative to their benchmarks and that they ACO, FFS) over another and whether that preference could be saving Medicare more than the benchmarks should be temporary. ACOs in Medicare have proven to would indicate. In either case, two-sided-risk ACO models be a popular choice for providers, but whether they remain show more savings relative to one-sided models. However, that way in the long run may depend on the choices a number of issues confront Medicare two-sided-risk policymakers make going forward. ■ June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 237

260 Endnotes 10 We used American Hospital Association data to identify 1 Services that qualify for attribution are defined in regulation. hospitals that participated in an ACO. MA and ACO Use of primary care services is required in statute. penetration data were from CMS. One-sided-risk ACOs can cost money in aggregate for the 2 Eight percent of revenue for a physician-only ACO is likely to 11 Medicare program because CMS pays shared savings to be much less than 10 percent of the benchmark. We calculate successful ACOs but does not collect losses from unsuccessful that 5 percent of benchmark is the upper bound on risk under ACOs (i.e., ACOs that exceed their benchmark). the revenue risk model. These clinician-only ACOs can include hospitals and qualify 3 Unlike other CMMI ACO demonstrations in which CMMI 12 for the lower loss limit if these hospitals are small, rural has chosen a limited number of ACOs to participate after hospitals with 100 or fewer beds. a competition of sorts, ACOs can join Track 1+ simply by In 2016, OneCare Vermont was responsible for 43,685 4 applying; if they meet the requirements, they are in the Medicare beneficiaries. demonstration. In fact, the application process goes through CMS’s MSSP office, not CMMI. 5 There are other models that qualify as A–APMs, including the Bundled Payments for Care Improvement Advanced 13 See pages 43–44 of the memo to Medicare Advantage Model, Comprehensive Care for Joint Replacement Model organizations, prescription drug plan sponsors, and other (Track 1: Certified Electronic Health Record Technology), interested parties about advance notice of methodological Comprehensive Primary Care Plus Model (CPC+), and the changes for calendar year 2019 for Medicare Advantage Oncology Care Model (two-sided-risk arrangement). The capitation rates, Part C and Part D payment policies, Commission has questioned the inclusion of the CPC+ and the 2019 draft call letter from February 1, 2018 model and the Oncology Care Model as A–APMs (Medicare (available at https://www.cms.gov/Medicare/Health-Plans/ Payment Advisory Commission 2016a). MedicareAdvtgSpecRateStats/Downloads/Advance2019Part2. pdf). 6 We did not adjust for health status because we were using ACO-level, not beneficiary-level, data. Thus, this evaluation is The president’s budget included this idea of a proportional 14 only an approximation of service use. incentive for A–APM participation but did not include an estimate of savings or spending. See page 67 of “Putting 7 Certainty—that is, informing the ACOs of their benchmark America’s Health First,” available at https://www.hhs.gov/ at the beginning of the year—may require modifying the sites/default/files/fy-2019-budget-in-brief.pdf. definition of two-sided risk if ACOs can withdraw from the program after learning what their benchmarks will be. For ACOs are made up of taxpayer identification numbers (TINs), 15 example, 3 of the 21 Next Generation ACOs dropped out of and any clinician billing through that TIN is automatically on the program early on after learning what their benchmarks the participant list. Specialists make up about two-thirds of would be. This practice could affect program savings over physicians treating Medicare FFS beneficiaries. time. 8 There is no explicit mention whether these savings are net of shared savings payments paid to Pioneer ACO providers. The near market includes counties where ACO providers 9 were located in the first performance year, plus all contiguous counties. Medicare accountable care organization models: Recent performance and long-term issues 238

261 References Green Mountain Care Board. 2018. . Annual report for 2017 Avalere Health. 2018. Medicare financial incentives encourage Montpelier, VT: GMCB. providers to take on greater financial risk. Press release. January 18. http://avalere.com/expertise/managed-care/insights/medicare- Kaiser Family Foundation. 2018. Side-by-side comparison: financial-incentives-encourage-providers-to-take-on-greater- Medicare accountable care organization (ACO) models. https:// financ. www.kff.org/interactive/side-by-side-comparison-medicare- accountable-care-organization-aco-models/. Centers for Medicare & Medicaid Services, Department of Health and Human Services. 2017a. Medicare program; revisions to L & M Policy Research. 2015. Pioneer ACO evaluation findings payment policies under the physician fee schedule and other . Final report prepared for from performance years one and two revisions to Part B for CY 2018; Medicare Shared Savings the Centers for Medicare & Medicaid Services by L & M Policy Program requirements; and Medicare Diabetes Prevention Research. Baltimore, MD: CMS. 82, no. 219 (November 15): Federal Register Program. Final rule. 52976–53371. McWilliams, J. M. 2016a. Changes in Medicare Shared Savings Journal of the American Program savings from 2013 to 2014. Centers for Medicare & Medicaid Services, Department of Health Medical Association 316, no. 16 (October 25): 1711–1713. and Human Services. 2017b. Medicare Shared Savings Program: Shared savings and losses and assignment methodology— McWilliams, J. M. 2016b. Savings from ACOs—Building on Specifications . Baltimore, MD: CMS. Annals of Internal Medicine 165, no. 12 (December early success. 20): 873–875. Centers for Medicare & Medicaid Services, Department of Health and Human Services. 2017c. New accountable care organization McWilliams, J. M., M. E. Chernew, J. B. Dalton, et al. 2014. model opportunity: Medicare ACO Track 1+ Model. Fact sheet. Outpatient care patterns and organizational accountability in July. https://www.cms.gov/Medicare/Medicare-Fee-for-Service- 174, no. 6 (June): 938–945. JAMA Internal Medicine Medicare. Payment/sharedsavingsprogram/Downloads/New-Accountable- Care-Organization-Model-Opportunity-Fact-Sheet.pdf. McWilliams, J. M., M. E. Chernew, and B. E. Landon. 2017a. Medicare ACO program savings not tied to preventable Colla, C. H., V. A. Lewis, L. S. Kao, et al. 2016. Association Health hospitalizations or concentrated among high-risk patients. between Medicare accountable care organization implementation Affairs 36, no. 12 (December): 2085–2093. and spending among clinically vulnerable beneficiaries. JAMA Internal Medicine 176, no. 8 (August 1): 1167–1175. McWilliams, J. M., L. G. Gilstrap, D. G. Stevenson, et al. 2017b. Changes in postacute care in the Medicare Shared Savings D’Ambrosio, D. 2017. Care Board five-year experiment will Program. 177, no. 4 (April 1): 518–526. JAMA Internal Medicine , Burlington Free Press focus on keeping patients healthy. December 21. McWilliams, J. M., L. A. Hatfield, M. E. Chernew, et al. 2016. Supplementary appendix to: McWilliams, J. M., L. A. Hatfield, Dale, S. B., A. Ghosh, D. N. Peikes, et al. 2016. Two-year costs et al. Early performance of accountable care organizations in and quality in the Comprehensive Primary Care Initiative. New New England Journal of Medicine Medicare. 374: 2357–66. England Journal of Medicine 374, no. 24 (June 16): 2345–2356. http://www.nejm.org/doi/suppl/10.1056/NEJMsa1600142/suppl_ file/nejmsa1600142_appendix.pdf. Duggan, M., J. Gruber, and B. Vabson. 2018. The consequences of health care privatization: Evidence from Medicare Advantage McWilliams, J. M., M. E. Chernew, B. E. Landon, et al. 2015. 10, no. 1: American Economic Journal: Economic Policy exits. Performance differences in year 1 of Pioneer accountable care 133–186. New England Journal of Medicine 372, no. 20 organizations. (May 14): 1927–1936. Fisher, E. S., D. O. Staiger, J. P. Bynum, et al. 2007. Creating accountable care organizations: The extended hospital medical Medicare Payment Advisory Commission. 2018. Report to the 26, no. 1 (January–February): w44–57. Health Affairs staff. Congress: Medicare payment policy . Washington, DC: MedPAC. Medicare: Results from Government Accountability Office. 2015. Report to Medicare Payment Advisory Commission. 2017a. the first two years of the Pioneer Accountable Care Organization . the Congress: Medicare and the health care delivery system . Report to the Ranking Member, Committee on Ways and Model Washington, DC: MedPAC. Means, House of Representatives. Washington, DC: GAO. Report to the Congress: Medicare and the Health Care Delivery System June 2018 | 239

262 Medicare Payment Advisory Commission. 2017b. Report to the Report to Medicare Payment Advisory Commission. 2009. Congress: Regional variation in Medicare Part A, Part B, and the Congress: Improving incentives in the Medicare program . Part D spending and service use . Washington, DC: MedPAC. Washington, DC: MedPAC. Medicare Payment Advisory Commission. 2016a. Comment Nelson, L. 2012. Lessons from Medicare’s demonstration projects on CMS’s proposed rule on the Merit-based Incentive Payment on disease management and care coordination . Working paper System and alternative payment models, June 15. http://www. 2012–01. Washington, DC: Congressional Budget Office. medpac.gov/docs/default-source/comment-letters/medpac- Free for all? Lessons from the RAND Newhouse, J. P. 1993. comment-on-cms-s-proposed-rule-on-the-merit-based-incentive- health insurance experiment . Cambridge, MA: Harvard payment-system-and-alternative-pa.pdf?sfvrsn=0. University Press. Report to Medicare Payment Advisory Commission. 2016b. Nyweide, D. J., W. Lee, T. T. Cuerdon, et al. 2015. Association of the Congress: Medicare and the health care delivery system . Pioneer accountable care organizations vs traditional Medicare Washington, DC: MedPAC. fee for service with spending, utilization, and patient experience. Medicare Payment Advisory Commission. 2016c. Report to the Journal of the American Medical Association 313, no. 21 (June . Washington, DC: MedPAC. Congress: Medicare payment policy 2): 2152–2161. Medicare Payment Advisory Commission. 2016d. Status Office of Inspector General, Department of Health and Human report on Medicare accountable care organizations (ACOs). Medicare Shared Savings Program accountable Services. 2017. Presentation at the Commission’s October public meeting. care organizations have shown potential for reducing spending http://www.medpac.gov/docs/default-source/default-document- . Washington, DC: OIG. and improving quality library/statusreportonmedicareaccountable_oct16_pres_sec. Office of the Actuary, Centers for Medicare & Medicaid Services, pdf?sfvrsn=0. Department of Health and Human Servivces. 2015. Certification Medicare Payment Advisory Commission. 2015a. MedPAC of Pioneer Model savings from Paul Spitalnic, Chief Actuary. comment on CMS’s proposed rule on the physician fee schedule April 10. and other revisions to Part B, September 8. Pham, H. H., M. Cohen, and P. H. Conway. 2014. The Pioneer Medicare Payment Advisory Commission. 2015b. MedPAC Accountable Care Organization Model: Improving quality and comment on CMS’s Medicare Shared Savings Program: 312, no. 16 (October 22–29): 1635–1636. JAMA lowering costs. Accountable care organizations proposed rule, February 2. Song, Z., S. Rose, D. G. Safran, et al. 2014. Changes in health Medicare Payment Advisory Commission. 2014a. Comment New care spending and quality 4 years into global payment. letter to CMS on accountable care organizations, June 16. http:// England Journal of Medicine 371, no. 18 (October 30): 1704– www.medpac.gov/docs/default-source/comment-letters/comment- 1714. letter-to-cms-on-accountable-care-organizations-june-16-2014-. Song, Z., D. G. Safran, B. E. Landon, et al. 2012. The ‘Alternative pdf?sfvrsn=0. Quality Contract,’ based on a global budget, lowered medical Medicare Payment Advisory Commission. 2014b. MedPAC spending and improved quality. Health Affairs 31, no. 8 (August): comment letter to CMS on accountable care organizations, June 16. 1885–1894. Medicare Payment Advisory Commission. 2014c. Report to the Congress: Medicare and the health care delivery system . Washington, DC: MedPAC. Medicare accountable care organization models: Recent performance and long-term issues 240

263 CHAPTER 9 Managed care plans for dual-eligible beneficiaries

264

265 CHAPTER 9 Managed care plans for dual-eligible beneficiaries Chapter summary In this chapter Individuals who receive both Medicare and Medicaid (known as dual-eligible • Introduction beneficiaries) often have complex health needs but are at risk of receiving Background on dual-eligible • fragmented or low-quality care because of the challenges in obtaining beneficiaries care from two distinct programs. Many observers have argued that the two programs could be better integrated by developing managed care plans that Update on the financial • alignment demonstration provide both Medicare and Medicaid services. Supporters argue that integrated plans would improve quality and reduce federal and state spending because More states are using • they would have stronger incentives to coordinate care than either program Medicaid managed care for does when acting on its own. However, these plans have been difficult to dual eligibles develop, and only 8 percent of full-benefit dual-eligible beneficiaries are now Medicare plans that serve • enrolled in a plan with a high level of Medicare and Medicaid integration. dual eligibles differ in key respects Since 2013, CMS and 10 states have tested the use of integrated Medicare– Medicaid Plans (MMPs) as part of the financial alignment demonstration. The • Potential policies to demonstrations in nine states, with a combined enrollment of about 380,000 encourage the development dual eligibles, are still under way and will likely continue at least through of integrated plans 2019. (The other demonstration ended as planned in 2017.) There are limited • Conclusion data available on the demonstration’s effects on areas such as quality, service use, and cost because the evaluations of the demonstration are taking longer to complete than expected. However, the information available is generally positive. Although the demonstration has often been difficult to implement, enrollment now appears stable (although participation is lower than many June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 243

266 expected) and quality appears to be improving. During site visits we made to several states, we found that the participating plans have grown more confident about their ability to manage service use as the demonstration has matured, with many plans reporting declines in the use of expensive services such as inpatient care. There also continues to be widespread support for the demonstration among the diverse collection of stakeholders interviewed on our site visits. The demonstration is part of a broader effort by many states to use Medicaid managed care to provide long-term services and supports (LTSS), such as nursing home care and personal care. Between 2004 and 2018, the number of states that have managed LTSS (MLTSS) programs grew rapidly, from 8 to 24, and more states will likely develop similar programs in the future. The growing use of managed care to provide LTSS—which account for most of Medicaid’s spending on dual eligibles—means that, in many states, the development of health plans that provide both Medicare and Medicaid services is probably the most feasible approach for pursuing closer integration. Medicare now has four types of integrated plans that serve dual eligibles: the demonstration’s Medicare–Medicaid Plans, Medicare Advantage dual-eligible special needs plans (D–SNPs), fully integrated dual-eligible SNPs (FIDE SNPs), and the Program of All-Inclusive Care for the Elderly. There are significant differences among these plans in several key areas, such as their level of integration with Medicaid, ability to use passive enrollment, and payment methodology. In addition, allowing MMPs and D–SNPs to operate in the same market has been problematic in some states because competition between the plans has reduced enrollment in the more highly integrated MMPs. Policy changes to better define the respective roles of each type of plan or consolidate them in some fashion may be needed. Three potential policies that would help encourage the development of integrated plans are (1) limiting how often dual eligibles can change their coverage, (2) limiting enrollment in D–SNPs to dual eligibles who receive full Medicaid benefits, and (3) expanding the use of passive enrollment, particularly when beneficiaries first qualify for Medicare. Collectively, these policies would improve care coordination and continuity of care, require D–SNPs to focus on the dual eligibles who stand to benefit the most from integrated care, and encourage more dual eligibles to enroll in plans with higher levels of Medicare–Medicaid integration. ■ Managed care plans for dual-eligible beneficiaries 244

267 argue that integrated plans would improve quality and Introduction reduce federal and state spending because they would have stronger incentives to coordinate care than either program More than 10 million people qualify for both Medicare does when acting on its own. However, these plans have and Medicaid and are known as dual-eligible beneficiaries. been difficult to develop, and their enrollment remains low. For these individuals, the federal Medicare program Our analysis examines the use of managed care for dual covers medical services such as hospital care, post-acute eligibles, focusing on the following topics: an update on care, physician services, durable medical equipment, CMS’s financial alignment demonstration, which is testing and prescription drugs. The federal–state Medicaid two new models of care for dual eligibles and has focused program covers a variety of long-term services and on managed care plans that provide both Medicare and supports (LTSS), such as custodial nursing home care and Medicaid services; the growing use of Medicaid managed community-based care, and wraparound services, such care for dual eligibles, which is making managed care as dental benefits and transportation. The program also the most feasible approach for better Medicare–Medicaid provides assistance with Medicare premiums and, in some integration in many states; the various types of Medicare cases, cost sharing. health plans that serve dual eligibles; and three potential Dual-eligible beneficiaries are generally in poorer health policies to encourage the development of integrated plans. than other Medicare beneficiaries. For example, as a group, dual eligibles are more likely to have functional impairments, behavioral health conditions, and substance Background on dual-eligible abuse disorders. As a result, dual eligibles account for a disproportionately large share of Medicare spending: beneficiaries In 2013, the most recent year of linked Medicare and Medicaid enrollment and spending data available, they Individuals must separately qualify for both Medicare and represented about 20 percent of Medicare beneficiaries but Medicaid coverage to become dual-eligible beneficiaries. accounted for about 34 percent of total Medicare spending. Roughly half of dual eligibles first qualify for Medicare They were also costly for Medicaid, representing about based on disability (compared with 17 percent of Medicare 15 percent of enrollment and about 32 percent of total beneficiaries who are not dual eligibles) and roughly half spending in that program (Medicare Payment Advisory qualify when they turn 65. Medicaid’s eligibility rules Commission and Medicaid and CHIP Payment and Access vary somewhat across states, but most dual eligibles Commission 2018). qualify because they receive Supplemental Security Income benefits, need nursing home care or have other Policymakers have long been concerned that dual eligibles high medical expenses, or meet the eligibility criteria for are vulnerable to receiving care that is fragmented or the Medicare Savings Programs, which provide assistance poorly coordinated. Medicare and Medicaid are separate with Medicare premiums and cost sharing (Medicare programs—the first purely federal, the second largely Payment Advisory Commission and Medicaid and CHIP operated by states with federal oversight and a mix of Payment and Access Commission 2018). Some individuals federal and state funding. Each program is complex, with who are eligible for Medicaid do not participate in the its own distinct rules for eligibility, covered services, and program, particularly those who qualify for the Medicare administrative processes. Medicare and Medicaid also Savings Programs (Medicaid and CHIP Payment and have relatively little incentive to engage in activities that Access Commission 2017). In December 2016, about 10.5 might benefit the other program. For example, states have million Medicare beneficiaries (18 percent of the total) relatively little incentive to reduce the use of inpatient care were dual eligibles. by dual eligibles because Medicare would realize most of the savings. Similarly, Medicare has relatively little Dual eligibles divide into two broad groups—“full benefit” incentive to prevent dual eligibles from going into nursing and “partial benefit”—based on the Medicaid benefits homes, where Medicaid pays for most of their care. they receive. Full-benefit dual eligibles qualify for the full range of Medicaid services covered in their state, which Many observers have argued that the two programs could generally includes a broad range of primary and acute care be better integrated by developing managed care plans that services, nursing home care, and other long-term services provide both Medicare and Medicaid services. Supporters June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 245

268 TABLE Dual eligibles had much higher per capita annual 9–1 spending in 2013 than other Medicare beneficiaries Medicare Medicaid Total Dual-eligible beneficiaries All $18,112 $11,126 $29,238 Full benefit 19,256 15,222 34,478 15,200 15,895 695 Partial benefit All other Medicare beneficiaries 8,593 8,593 N/A Note: N/A (not applicable). Figures include all Medicare (Part A, Part B, and Part D) and Medicaid spending except Medicare or Medicaid spending on Part A, Part B, or Part D premiums. The Medicaid spending for partial-benefit dual eligibles is for coverage of Medicare cost sharing. Source: MedPAC analysis of linked Medicare–Medicaid enrollment and spending data. and supports. In contrast, partial-benefit dual eligibles were more likely than other Medicare beneficiaries to use receive assistance only with Medicare premiums and, in inpatient care (26 percent vs. 16 percent), and those who some cases, assistance with cost sharing. In December were hospitalized had higher inpatient costs ($19,580 vs. 2016, there were 7.5 million full-benefit dual eligibles and $16,362, respectively). The Medicaid costs for full-benefit 3.0 million partial-benefit dual eligibles. dual eligibles largely comprised spending on LTSS, such as nursing home care and home- and community-based Given the role that factors such as disability and functional waiver programs. Less than half of full-benefit dual impairment play in becoming a dual eligible, it is not eligibles (42 percent) used LTSS in 2013, but spending surprising that dual eligibles are more likely than other on those services accounted for about 80 percent of this Medicare beneficiaries to report that they are in poor population’s total Medicaid costs (Medicare Payment health (18 percent vs. 6 percent) or need help performing Advisory Commission and Medicaid and CHIP Payment three or more activities of daily living (30 percent vs. 9 and Access Commission 2018). percent) (Medicare Payment Advisory Commission and Medicaid and CHIP Payment and Access Commission 1 2018). The poorer health of this population leads in turn Update on the financial alignment to higher costs (Table 9-1). Measured on a per capita basis, the average annual Medicare cost for dual eligibles in demonstration 2013 was over $18,000, more than two times higher than for other Medicare beneficiaries. Within the dual-eligible Under the financial alignment demonstration, CMS has population, those eligible for full Medicaid benefits had been working with 13 states to test 2 new models of care higher Medicare costs and much higher Medicaid costs for full-benefit dual eligibles—a capitated model and a than those eligible for partial Medicaid benefits only. managed fee-for-service (FFS) model. Both models seek In 2013, Medicare and Medicaid together spent more to improve the coordination of Medicare and Medicaid for than $34,000 per capita, on average, on full-benefit dual dual eligibles, improve the quality of their care, and lower eligibles, with Medicare accounting for about 56 percent costs (Centers for Medicare & Medicaid Services 2011): of the combined spending and Medicaid the other 44 • Under the capitated model, managed care plans percent. provide the full range of Medicare and Medicaid The high Medicare costs for dual eligibles are driven by benefits to dual eligibles. The plans receive a blended a combination of higher utilization of all major types of Medicare–Medicaid payment rate that is reduced to services and higher per user spending for those who receive reflect expected savings from the demonstration. 2 care. For example, in 2013, full-benefit dual eligibles Managed care plans for dual-eligible beneficiaries 246

269 TABLE Overview of the financial alignment demonstrations 9–2 January 2018 enrollment MOU date Start/end dates State Model type Eligible population Capitated California Aged and disabled March 2013 April 2014 to 2019 116,721 Colorado Managed FFS Aged and disabled February 2014 September 2014 to 2017 — Illinois Capitated Aged and disabled February 2013 March 2014 to 2019 53,927 Capitated Massachusetts August 2012 19,337 October 2013 to 2018 Disabled only Michigan Capitated Aged and disabled April 2014 March 2015 to 2020 39,638 38,994 Minnesota Alternative Aged only September 2013 September 2013 to 2018 Capitated Aged and disabled August 2013 January 2015 to 2019 4,263 New York (1) 731 Capitated Aged and disabled November 2015 April 2016 to 2020 New York (2) Aged and disabled December 2012 Capitated Ohio May 2014 to 2019 75,161 14,144 July 2015 Capitated Rhode Island Aged and disabled July 2016 to 2020 October 2013 Capitated 11,598 South Carolina Aged only February 2015 to 2018 Aged and disabled Capitated March 2015 to 2020 47,527 Texas May 2014 April 2014 to 2017 Capitated Aged and disabled — Virginia May 2013 Washington Managed FFS Aged and disabled October 2012 April 2013 to 2018 19,609 Note: MOU (memorandum of understanding), FFS (fee-for-service). All states use additional eligibility criteria beyond age and disability. New York’s first demonstration targets individuals who use certain kinds of long-term services and supports, while the second targets individuals with intellectual and developmental disabilities. All demonstrations will end on December 31 of the indicated calendar year. Massachusetts, Minnesota, and Washington plan to extend their demonstrations for two years, but these extensions have not been finalized and are not reflected in the table. South Carolina can extend its demonstration for two years but has not indicated whether it will do so. The enrollment figure for Washington is for December 2017. MedPAC analysis of state MOUs, CMS demonstration guidance, and Medicare Advantage enrollment data for January 2018; personal communication with L. Source: Barnette (Centers for Medicare & Medicaid Services 2018c). • Under the managed FFS model, states provide greater that included state Medicaid officials, executives and care coordination to dual eligibles who are enrolled in care coordination staff for health plans participating in both FFS Medicare and FFS Medicaid. States receive the demonstration, several different kinds of providers, a retrospective performance payment from Medicare if and beneficiary advocacy groups. This update focuses expenditures for demonstration enrollees are below a primarily on the experience with the capitated model, target amount. which most participating states are testing, but also touches on the managed FFS model. Our update is based on a wide range of CMS guidance related to the demonstration, the evaluations of its effects Table 9-2 provides an overview of the programs that are that have been completed to date, administrative data, and part of the demonstration. There are 14 demonstrations findings from site visits to participating states. Between in 13 states (2 of those demonstrations have ended). December 2015 and February 2018, we made eight site Most participating states are testing the capitated model; visits to six states (California, Illinois, Massachusetts, only Colorado and Washington have tested the managed New York, Ohio, and Texas) and conducted phone FFS model, while Minnesota is testing an alternative 3 interviews with stakeholders in two other demonstration model. Most demonstrations are open to both disabled states (Colorado and Washington). In all, we conducted and aged dual eligibles, although one (Massachusetts) is over 80 interviews with a diverse range of stakeholders limited to disabled beneficiaries, and two (Minnesota and June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 247

270 Findings from earlier efforts to develop integrated plans of care provided in a nursing home but still live in the he financial alignment demonstration’s capitated community, similar to the Program of All-Inclusive model was influenced by an earlier set of Care for the Elderly. In 2004, an evaluation of the demonstrations under which CMS and states T program compared WPP enrollees with dual eligibles developed the first integrated plans for dual eligibles. 4 who had similar characteristics but were not enrolled. These efforts started in the 1990s and 2000s, when CMS approved demonstration projects in Wisconsin The WPP enrollees had similar or slightly lower rates (1996), Minnesota (1997), and Massachusetts (2004). of hospital use, mortality, and nursing home admission. All three states succeeded in developing integrated However, the study found that WPP did not reduce plans, and making the plans a permanent part of Medicare spending because of the methodology that Medicare was one motivation for the creation of was used to set the capitation rates for the participating Medicare Advantage dual-eligible special needs plans (Kane and Homyak 2004). plans (D–SNPs) in 2003 (Schmitz et al. 2008). The The Minnesota program, known as Minnesota Senior demonstration plans were converted into D–SNPs Health Options (MSHO), was limited to beneficiaries in 2006, and many still operate today. CMS is now who were 65 or older, and it used a traditional managed testing integrated plans on a broader scale with the care approach. The same 2004 report that evaluated financial alignment demonstration, but its evaluations the Wisconsin demonstration also assessed MSHO. are taking longer to complete than initially expected. The study found that MSHO enrollees in nursing As a result, much of the research on integrated plans homes had significantly fewer hospital admissions and and their effects on spending, service use, and quality emergency room visits than comparison groups of dual of care still draws on the experience of these earlier eligibles. However, MSHO enrollees did not perform demonstrations. significantly better in key areas such as mortality rates and change in functional status over time, and the The Wisconsin program, known as the Wisconsin quality of their nursing home care was similar (Kane Partnership Program (WPP), was designed to serve and Homyak 2004). elderly and disabled dual eligibles who need the level (continued next page) South Carolina) are limited to aged beneficiaries. CMS CMS initially planned for the demonstrations to last three approved each demonstration by signing a memorandum years, but it has extended most of them because their of understanding (MOU) with the state that summarizes evaluations have not been completed. In July 2015, CMS the key parameters of the demonstration. The first MOU announced that all states could extend their demonstrations (Massachusetts) was signed in August 2012; the last for an additional two years; in January 2017, it announced (for New York’s second demonstration) was signed in that the first three states to start their demonstrations November 2015. Most of the demonstrations started (Massachusetts, Minnesota, and Washington) could enrolling beneficiaries about a year after the signing of the extend them for another two years on top of that and said MOU. that other states could receive similar extensions if more time is needed to complete their evaluations (Centers As of January 2018, about 440,000 dual eligibles were for Medicare & Medicaid Services 2017c, Centers for enrolled in the demonstrations, making this one of Medicare & Medicaid Services 2015a). Colorado and the largest demonstrations CMS has conducted that is Virginia decided against extending their demonstrations specifically aimed at this population. The four largest 5 and concluded them at the end of 2017. The other states demonstrations—California, Ohio, Illinois, and Texas— have finalized their extensions or indicated their intent to account for about two-thirds of the national total. do so, except for South Carolina, which has not decided Managed care plans for dual-eligible beneficiaries 248

271 Findings from earlier efforts to develop integrated plans (cont.) enrollees, relative to a comparison group of dual A 2016 study of MSHO had much more positive eligibles, had lower rates of nursing facility use and findings. This study compared MSHO enrollees with lower mortality rates (JEN Associates 2015). However, dual eligibles in Minnesota who did not participate another study found that SCO enrollment did not have and were mostly enrolled in a combination of fee-for- a statistically significant effect on 30-day hospital service Medicare and Medicaid managed care. The readmission rates (Jung et al. 2015). study found that MSHO enrollees were 48 percent less likely to have an inpatient stay, 6 percent less On balance, the findings from the early experiments likely to have an outpatient emergency room visit, with integrated plans are moderately positive. 2.7 times more likely to have a visit with a primary Integrated plans have shown some ability to reduce care physician, and no more likely to have a visit with enrollees’ use of hospital services and redirect LTSS a specialist. As for long-term services and supports use from nursing home care to community-based (LTSS) use, MSHO enrollees were 13 percent more care. The available research has sometimes found likely to receive home- and community-based services that integrated plans perform no better than other and no more likely to have a nursing home admission. arrangements in some areas (such as readmission rates The authors concluded that the integrated MSHO in the Massachusetts program), but, at the same time, program was associated with desirable patterns of the research has not found that dual eligibles have service use and “may have merit for other states” fared worse in integrated plans. Our understanding of (Anderson et al. 2016). the effectiveness of integrated plans should improve significantly as more evaluations of the financial Like MSHO, the program in Massachusetts—Senior alignment demonstration become available. Care Options (SCO)—is also limited to beneficiaries ■ who are 65 and older. One study found that SCO yet (Centers for Medicare & Medicaid Services 2018b). Medicare and Medicaid services. We refer to this type Most of the demonstrations therefore appear likely to last of plan as an . The use of integrated integrated plan for five to seven years and operate until 2019 or 2020, if plans has long been suggested as a way to improve not longer. care for dual eligibles, and CMS has tested their use in other demonstrations (see text box on earlier findings). CMS is conducting the financial alignment demonstration Supporters argue that integrated plans, because of their using the authority of its Center for Medicare & Medicaid responsibility for the full range of Medicare and Medicaid Innovation (CMMI). Under this authority, the Secretary benefits, would not have the incentive that each program can test new payment models and subsequently expand operating independently has to shift costs to the other the use of a model that he determines will either (1) program and would have stronger incentives to coordinate reduce spending without affecting the quality of care care across the programs. Dual eligibles would also find or (2) improve the quality of care without increasing it easier to understand their coverage and obtain care spending. As part of this process, the CMS chief actuary because they would receive integrated materials (such as a must certify that expanding the model will not increase single membership card and provider directory instead of overall Medicare or Medicaid spending. CMS could thus separate Medicare and Medicaid versions) and have one potentially expand the use of the capitated model and point of contact for their care needs. Integrated plans, it managed FFS model in the future. has been argued, would thus improve the quality of care for dual eligibles and produce savings by reducing the use Demonstrations using the capitated model of high-cost services such as inpatient hospital care and The key feature of the capitated model, which is used nursing home care. by most states, is a managed care plan that provides all June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 249

272 FIGURE Title here... FIGURE X-X Total enrollment in Medicare–Medicaid Plans has been relatively stable since mid-2015 9–1 450,000 400,000 350,000 300,000 250,000 200,000 150,000 100,000 Medicare–Medicaid Plan enrollment 50,000 0 2015 2018 2017 2016 2013 2014 MedPAC analysis of monthly Medicare Advantage enrollment data from CMS. Source: Note: Note and Source are in InDesign. The integrated plans in the financial alignment that serve individuals with intellectual or developmental demonstration are known as Medicare–Medicaid Plans disabilities. In all, about 1.3 million beneficiaries are Source: (MMPs). They provide all Medicare-covered and all or eligible for the 10 active demonstrations. most Medicaid-covered services to their enrollees. The Under the demonstration, states can passively enroll Notes about this graph: MMPs are required to provide their enrollees with a high beneficiaries in MMPs. With passive enrollment, level of care coordination and receive a blended capitation • Data is in the datasheet. Make updates in the datasheet. beneficiaries are automatically enrolled in MMPs unless rate that combines Medicare Part A, Part B, and Part D and • I deleted the years from the x-axis and put in my own. they indicate that they do not want to join an MMP, which Medicaid payments. is known as opting out. (See the Commission’s June 2016 • I had to manually draw tick marks and axis lines because they kept resetting when I changed any data. report for a fuller discussion of how passive enrollment Beneficiary participation • The dashed line looked ok here, so I didn’t hand draw it. has been used in the demonstration and how it is used CMS has limited eligibility for the financial alignment • I can’t delete the legend, so I’ll just have to crop it out in InDesign. elsewhere in the Medicare and Medicaid programs.) demonstration to full-benefit dual eligibles—individuals Every state testing the capitated model has used passive • Use direct selection tool to select items for modification. Otherwise if you use the black selection tool, they will reset to graph who are eligible for both Medicare (Part A, Part B, and enrollment for at least some beneficiaries, although default when you change the data. Part D) and full Medicaid benefits in their state. States can 6 California, New York, and Rhode Island no longer use it. further limit eligibility based on the particular needs of • Use paragraph styles (and object styles) to format. In the other states, passive enrollment is now being used their demonstration, and every state testing the capitated largely to enroll beneficiaries who have become dually model has done so. For example, 8 of the 10 active eligible since the start of the demonstration. demonstrations operate only in certain parts of the state, usually around large metropolitan areas, and 6 exclude Total enrollment in MMPs grew gradually between 2013 beneficiaries enrolled in certain Medicaid home- and and 2015 because the individual state demonstrations community-based waiver programs, particularly those Managed care plans for dual-eligible beneficiaries 250

273 TABLE MMP participation rates, by state, as of June 2017 9–3 Eligible beneficiaries Participation rate MMP enrollment State California 118,386 424,000 28 % 51,063 35 146,000 Illinois Massachusetts 16,950 104,000 16 Michigan 39,681 105,000 38 3 New York (1) 4,708 156,000 New York (2) 575 3 20,000 75,603 111,000 68 Ohio Rhode Island 14,002 30,000 47 8,033 South Carolina 21 39,000 Texas 40,738 165,000 25 Virginia 27,958 67,000 42 1,367,000 397,697 Total 29 MMP (Medicare–Medicaid Plan). Virginia’s demonstration ended in December 2017. Note: Source: Medicare Advantage enrollment data for June 2017; personal communication with L. Barnette (Centers for Medicare & Medicaid Services 2017j). Comparing MMP enrollees and beneficiaries who opted started at different times, and many were implemented in out One question about the demonstration and its use of stages (Figure 9-1). Since mid-2015, overall enrollment passive enrollment has been whether the beneficiaries has not changed much, usually ranging between 360,000 who opted out differed from those who accepted passive and 400,000 beneficiaries per month. Enrollment dropped enrollment in an MMP. To better examine this issue, we somewhat in January 2018, due largely to the end of obtained data for the MMPs from the Medicare Advantage Virginia’s demonstration, and stood at about 383,000 Prescription Drug (MARx) system, which CMS uses to beneficiaries. process enrollment transactions for all types of Medicare The participation rates for many demonstrations have health plans. The MARx data have two advantages over been lower than expected and vary widely across states. traditional enrollment data: (1) They indicate whether Table 9-3 shows the MMP enrollment, number of eligible a beneficiary was passively enrolled in an MMP or beneficiaries, and participation rate for each demonstration enrolled voluntarily, and (2) they can be used to identify as of June 2017. Across all the demonstrations, only about beneficiaries who were scheduled for passive enrollment 29 percent of eligible beneficiaries were enrolled in an but later opted out. The MARx data that we obtained have MMP. Ohio (68 percent) and Rhode Island (47 percent) all transactions involving MMPs from October 2013 (the had the highest participation rates, while New York had start of the first capitated demonstration) through April the lowest rates (3 percent in both of its demonstrations). 2016 and thus do not have information for the second New The participation rates for MMPs have been relatively York or Rhode Island demonstrations, which started later low because many beneficiaries opted out or left the in 2016. 7 MMP after a short period of time. In the states we During this period, we found that states attempted to visited, stakeholders said many beneficiaries declined to passively enroll about 855,000 beneficiaries in MMPs and participate because they were satisfied with their existing that 41 percent of them opted out (Table 9-4, p. 252). We care, did not fully understand how the demonstration also examined whether opt-out rates varied by age, sex, would affect them, or were encouraged to opt out by race/ethnicity, and whether the beneficiary was a long- providers (Medicare Payment Advisory Commission stay nursing home resident at some point during the year. 2016). June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 251

274 TABLE Opt-out rates for MMPs varied, October 2013—April 2016 9–4 Number of beneficiaries (in thousands) Share of population Opt-out rate 41 % % All passive enrollments 100 855 Age Under 65 36 35 307 45 65 and older 549 64 Sex 522 61 44 Female Male 333 38 39 Race/ethnicity 35 299 43 White 36 Hispanic 222 26 24 African American 36 207 Asian 110 56 13 17 46 2 All other/unknown Long-term nursing home use 757 Zero months 89 41 At least 1 month 98 11 42 Note: MMP (Medicare–Medicaid Plan). Components may not sum to totals because of rounding. MedPAC analysis of MMP enrollment transaction data and Medicare enrollment data. These figures do not include records for beneficiaries who opted out by Source: contacting the state Medicaid agency or beneficiaries with end-stage renal disease. Beneficiaries ages 65 and older were more likely to opt 60 days before the actual enrollment date. (During this out than those under age 65 (45 percent vs. 35 percent), 60-day period, states send beneficiaries two notices about and women were more likely to opt out than men (44 their upcoming passive enrollment, and beneficiaries can percent vs. 38 percent). The similarity between these opt out any time before the scheduled enrollment date.) two metrics is not surprising because dual eligibles over However, beneficiaries in some states were able to opt 65 are disproportionately female. As for race/ethnicity, out by contacting the state before the start of the passive beneficiaries of Asian ancestry were the most likely to opt enrollment process. The beneficiaries who opted out in out (56 percent), while African American and Hispanic this manner do not appear in the MARx data because beneficiaries were least likely (36 percent). Finally, the states never began the process of passively enrolling them. opt-out rates for long-stay nursing home residents and CMS does not know how many beneficiaries have used other beneficiaries were similar. The figures shown in this other method to opt out. Table 9-4 are aggregated across all MMP states; the figures In addition to high opt-out rates, another challenge for for individual states will vary given the differences in their MMPs has been high disenrollment rates (enrollees demographic characteristics (such as race/ethnicity) and leaving the plan for other coverage). For example, we the eligibility criteria for each demonstration. found that 25 percent of the beneficiaries who were These opt-out rates should be viewed as somewhat passively enrolled in MMPs disenrolled within the first conservative because the MARx data do not include every three months. However, the share of beneficiaries who beneficiary who opted out. The MARx data can identify disenrolled within the first three months varied relatively beneficiaries who opted out only after CMS has begun the little across the various categories shown in Table 9-4. process of passively enrolling them, which starts at least For comparison, we also examined beneficiaries who Managed care plans for dual-eligible beneficiaries 252

275 TABLE Risk scores for beneficiaries enrolling in Medicare–Medicaid Plans 9–5 2013 2014 2015 2016* 9 5 9 1 Number of active demonstrations Total enrollment actions 59,688 New MMP enrollees 5,120 241,284 395,334 74,448 N/A Beneficiaries who opted out 255,304 24,366 Average risk score New MMP enrollees 1.14 1.39 1.59 1.59 N/A Beneficiaries who opted out 1.75 1.48 1.83 New MMP enrollees, by length of enrollment 102,510 12,276 905 1 to 3 months 69,686 4 to 6 months 25,604 6,534 343 39,277 3,872 145,994 253,547 40,878 7 months or more 59,688 Total 5,120 241,284 395,334 Average risk scores for new MMP enrollees, by length of enrollment 1 to 3 months 1.20 1.64 1.86 1.89 1.72 1.65 1.52 1.20 4 to 6 months 1.48 1.45 1.24 1.12 7 months or more Note: MMP (Medicare–Medicaid Plan), N/A (not applicable). Table does not include records for beneficiaries who opted out by contacting the state Medicaid agency or beneficiaries with end-stage renal disease. There were no opt-outs in 2013 because the only demonstration then under way (Massachusetts) did not begin passive enrollment until 2014. ”New MMP enrollees” are those who first joined an MMP in the stated year. “Length of enrollment” is based on the number of months of enrollment through December 2016. *2016 figures are for enrollment actions with January through April effective dates and do not include the second demonstration in New York or the demonstration in Rhode Island, which both started later in 2016. MedPAC analysis of MMP enrollment transaction data, Medicare enrollment data, and CMS–hierarchical conditions categories risk score data. Source: enrolled voluntarily, who represented about 15 percent of plan types, such as MMPs, to account for differences all MMP enrollees. The share of voluntary enrollees who in beneficiaries’ health status. Risk scores are based on disenrolled within the first three months was 17 percent, a combination of demographic information (such as lower than the figure for passive enrollees but still high for age, sex, and whether the beneficiary first qualified for a group that had actively chosen to enroll in an MMP. Like Medicare based on a disability) and diagnostic information the passive enrollees, the disenrollment rates for voluntary from claims; scores are scaled to show how a beneficiary’s enrollees varied little by age, sex, race/ethnicity, or nursing expected Medicare costs compare with the average home use. expected cost for all FFS beneficiaries. For example, a risk score of 1.0 indicates that the expected costs for a We also used Evidence of favorable selection for MMPs beneficiary equal the overall average, and a risk score of the MARx data and MMP enrollment data to examine 1.3 indicates that the expected costs for a beneficiary are whether beneficiaries who opted out or disenrolled were 30 percent higher than the overall average. healthier or sicker than those who enrolled in MMPs. We found that the dual eligibles who have participated We compared beneficiaries using their risk scores from in the demonstration appear to be healthier than those the CMS hierarchical condition category (CMS–HCC) who opted out (Table 9-5). For example, in 2014, the risk adjustment model. CMS uses this model to adjust beneficiaries who joined an MMP had an average risk payments to Medicare Advantage (MA) plans and other June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 253

276 score of 1.39, while the beneficiaries who opted out had network and more likely to have at least one provider an average score of 1.48. The figures for 2015 and 2016 encourage them to opt out. Similarly, beneficiaries with follow the same basic pattern, although the average risk lower risk scores may have had less interaction with the scores for new enrollees and those opting out vary from health care system in the past and therefore may be more year to year. likely to be satisfied with the plan’s provider network. Among beneficiaries who enrolled in MMPs, there were One concern about favorable selection is that plans may also differences in risk scores when the enrollees were have a financial incentive to avoid serving sicker enrollees. stratified based on the length of time they were enrolled. However, many MMPs we interviewed said they would In 2014, about 241,000 beneficiaries joined MMPs, but like to have more enrollees, and several expressed support almost 70,000 (29 percent) were enrolled for 3 months for policies that would make it harder for dual eligibles or less, and about 26,000 (11 percent) were enrolled for to disenroll from MMPs. CMS and states also mitigate 8 between 4 and 6 months. financial incentives to avoid serving sicker enrollees The beneficiaries who were by risk adjusting the Medicare and Medicaid payments enrolled for three months or less had a higher average risk to MMPs, which should reduce this incentive because score (1.64) than those who were enrolled for four to six sicker enrollees also generate more revenues for plans. months (1.52), who in turn had a higher average risk score In addition, CMS increased MMP payment rates for Part than those who were enrolled for seven months or more A and Part B services after finding that the CMS–HCC (1.24). The patterns for 2015 and 2016 were similar. As model had historically tended to underestimate costs with Table 9-4 (p. 252), the figures in Table 9-5 (p. 253) for full-benefit dual eligibles (Centers for Medicare & are aggregated across all MMP states, and the figures for Medicaid Services 2015b). individual states will vary. Taken together, these differences in risk scores indicate Health plan participation that favorable selection has occurred in the capitated A total of 68 MMPs (counted at the contract level) have demonstrations, meaning that the healthier beneficiaries participated in the demonstration. Most are sponsored 9 among those eligible have been more likely to participate. by organizations with prior experience in Medicare In this respect, the financial alignment demonstration Advantage, Medicaid managed care, or both (Weiser and is similar to other managed care programs that feature Gold 2015). However, 18 plans have left the demonstration voluntary enrollment. For example, the Commission has since it started, and only 50 are still participating. CMS found that Medicare beneficiaries who enroll in MA plans has not allowed any new MMPs to join the demonstration are healthier than FFS enrollees and that beneficiaries who so far, although new plans will be able to join in the future switch from MA plans to FFS coverage have higher risk when states reprocure their Medicaid managed care plans. scores than beneficiaries who remain in MA (Medicare Plans have left the demonstration for a variety of reasons: Payment Advisory Commission 2012b). Some older studies also found evidence of favorable selection in • Most of the departing plans (11 of 18) were part of voluntary Medicaid managed care programs (American New York’s first demonstration and left because of 10 Academy of Actuaries 1996, Scholle et al. 1997). low enrollment. The demonstration started with an Nevertheless, the presence of favorable selection means unusually large number of MMPs (21), but beneficiary that the demonstration is not fully serving relatively sicker participation has been very low (see Table 9-3, p. 251), dual eligibles, who might benefit the most from better care leaving many plans with very little enrollment. The coordination. 11 plans that left the demonstration all had fewer than 300 enrollees. The stakeholders we interviewed on our site visits indicated that many beneficiaries opted out of the Three MMPs left because of Virginia’s decision to end • demonstration to maintain access to their current providers its demonstration at the end of 2017. or because their providers encouraged them to opt out. Two plans that left in 2015—one in Massachusetts • Beneficiaries with higher risk scores would tend to have and one in Illinois—cited inadequate payment rates higher service use and see a larger number of providers. As as a primary reason for their decision. However, CMS a result, they might have been more likely to find that one increased payment rates for MMPs in 2016, and we or more of their providers was not in their MMP’s provider are not aware of any plan departures since then that have been attributed to inadequate payment rates. Managed care plans for dual-eligible beneficiaries 254

277 TABLE Enrollment in individual MMPs varies widely 9–6 Enrollees MMPs Average enrollees Number per MMP Share Number Share Enrollment range 9 Less than 1,000 18 2,432 1 % % 270 1,001 to 5,000 11 3,097 22 34,064 9 5,001 to 10,000 89,346 6,873 23 26 13 15,130 67 257,205 34 17 More than 10,000 7,661 Total 100 383,047 100 50 Note: MMP (Medicare–Medicaid Plan). MMPs are counted at the contract level. Medicare Advantage enrollment data for January 2018. Source: One plan left the Illinois demonstration at the end of • The largest MMP, sponsored by Inland Empire Health 2017 after the plan’s parent company decided to end Plan in California, had more than 25,000 enrollees (data all of its Medicaid-related business in the state. not shown). The number of plans in each demonstration varies. One question about the demonstration has been whether California and the first New York demonstration health plans need a certain level of enrollment to currently have 10 plans each, while the second New York successfully operate an MMP. Before the demonstration, demonstration and Rhode Island have only 1 plan each. many health plans believed that they would need to make The other demonstrations have between two and seven significant upfront investments to provide the level of care plans. Many MMPs serve only part of the demonstration coordination required for MMPs. CMS authorized the area. For example, Texas is conducting its demonstration use of passive enrollment in the demonstration partly to in six counties. The state has five MMPs, but only one to ensure that plans would have enough enrollment to justify three plans operate in each county. those initial investments, and many plans we interviewed indicated that passive enrollment was a key factor in their Each MMP has signed a three-way contract with CMS decision to participate in the demonstration. and the state that specifies its requirements under the demonstration. States initially selected the plans for the During our site visits and in other interviews with MMPs, demonstration and could limit the number of plans that we asked plan officials whether an MMP needed a participate. Plans also had to satisfy CMS requirements minimum level of enrollment to operate effectively. Some and pass a readiness review that examines areas such as plans did not provide a figure, but most of the plans that network adequacy, financial solvency, care management did indicated that MMPs were easier to operate with at capabilities, and plan staffing for functions like customer least 5,000 to 7,500 enrollees because they could benefit service (Medicaid and CHIP Payment and Access from economies of scale in providing care coordination, Commission 2015). such as hiring staff with clinical expertise in behavioral health, and spreading relatively fixed costs for activities The number of dual eligibles enrolled in each MMP varies such as the development of member materials. Some plans widely (Table 9-6). Nine MMPs that were operating also said that higher enrollment would make it easier for in January 2018 had fewer than 1,000 enrollees. All of them to get providers to join their networks. Except for these plans were in New York, and 5 had fewer than New York, most plans appear to have enough enrollees to 250 enrollees. A total of 30 MMPs had more than 5,000 adequately test the capitated model. enrollees, and 17 MMPs had more than 10,000 enrollees. June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 255

278 Many stakeholders we interviewed on our more recent Evaluations of the demonstration site visits said they were frustrated with the delays in CMS has contracted with RTI International to evaluate completing the evaluations. At the time of these visits, the each demonstration’s effect on areas such as access demonstrations in California, Massachusetts, New York, to care, service use, quality of care, and cost. These Ohio, and Texas had been in operation for about three evaluations will include qualitative analyses, such as years. Many stakeholders in those states believed that the findings from beneficiary focus groups and interviews demonstrations showed promise and wanted to know what with key stakeholders, as well as quantitative analyses CMS was going to do in the “post-demonstration” era. using claims, encounter, assessment, and enrollment data. RTI plans to release an annual evaluation report Given the delays with the quantitative analyses, RTI has for each demonstration and a final evaluation report that issued several reports with qualitative analyses of the synthesizes findings across all participating states (Walsh demonstration, such as findings from focus groups of et al. 2013). MMP enrollees and a review of how MMPs are providing care coordination (Ptaszek et al. 2017, Weiner et al. 2017). However, these evaluations are taking much longer to CMS has also issued other data, such as results from complete than expected. So far, only one annual evaluation surveys of MMP enrollees about their patient experience. for a capitated demonstration has been released, covering The rest of our update on the capitated model incorporates the first year of the Massachusetts demonstration (Gattine 11 findings from these other data sources and from our site et al. 2017). The delays have been due to difficulties visits. in gathering the data needed to conduct the quantitative analyses. RTI plans to measure the effects of the Care coordination demonstrations on both Medicare and Medicaid service Under the demonstration, CMS and states hope that use by comparing the dual eligibles who are eligible for greater care coordination for dual eligibles will improve the demonstrations (whether or not they participate) with the quality of their care and reduce Medicare and similar groups of dual eligibles living in other states. This Medicaid spending. MMPs are required to provide care approach requires a great deal of administrative data, such coordination using a model that has three main elements: as Medicaid FFS claims and encounter data from multiple states, MMP encounter data for both Medicare and • Each enrollee must receive an initial health Medicaid services, Medicare FFS claims, MA encounter assessment. Each demonstration has its own deadlines data, and Medicare Part D data (Walsh et al. 2013). Some for completing the assessments; most are within of those data, particularly MMP encounter data and 90 days of enrollment. The assessments must be Medicaid data from comparison states, are taking longer to comprehensive, covering physical health, behavioral obtain than anticipated. health, ability to perform activities of daily living, and cognitive status (Medicaid and CHIP Payment and CMS will release more evaluations as these data issues Access Commission 2015). The assessments must also are resolved, but the annual reports for the first one or be updated periodically, usually at least once a year. two years of each demonstration may not provide much insight into the effects of the capitated model. In the Each enrollee must have an individual care plan that • states we visited, there was broad agreement among is based in part on the results of the assessment. These stakeholders that the demonstrations had been challenging care plans must be developed by an interdisciplinary to implement. Many MMPs we interviewed said they team of providers. The membership of the team varies had needed roughly 18 to 24 months to fully establish by demonstration but usually includes the enrollee’s themselves and that their impact on enrollees’ service use care coordinator, primary care physician, LTSS during that time was limited. The first-year evaluation providers, and relevant specialists (such as behavioral of the Massachusetts demonstration took a similar view; health providers). Enrollees can also participate if they that report found “limited evidence of the demonstration’s wish. effect during the first demonstration year, partly due to initial implementation challenges but also due to the need Each enrollee is assigned to a care coordinator who • for allowing adequate time for care interventions at the often takes the lead in developing the enrollee’s beneficiary level to affect service utilization” (Gattine et care plan and provides ongoing help in finding and al. 2017). obtaining necessary care. Managed care plans for dual-eligible beneficiaries 256

279 Learning more about how MMPs provide care modified their care coordination arrangements as they coordination was a primary goal of our site visits, and gained experience and tested new approaches. Many RTI has also issued two reports on the topic as part of of the plans we interviewed had increased their use of its evaluation of the demonstration (Ptaszek et al. 2017, subcontractors to provide care coordination, particularly Weiner et al. 2017). The views that we heard during as they developed relationships with local social service our interviews with stakeholders are consistent with the agencies (such as area agencies on aging or behavioral findings in RTI’s reports. health providers) and gained a better understanding of the capabilities of those entities. Texas appears to be an Many MMPs have had trouble completing the initial health exception in this regard; the plans we interviewed there assessments on time for two reasons. First, plans have not relied entirely on internal employees to provide care been able to locate many enrollees because their contact coordination. information is out of date. RTI found that most plans had trouble finding between 20 percent and 35 percent of their The MMPs we interviewed said the level of care enrollees, and the plans we interviewed supplied similar coordination that enrollees receive depends on their care figures. Second, some plans we interviewed found it needs. High-risk enrollees, such as those who use LTSS, challenging to conduct assessments when large numbers of receive the most extensive care coordination, such as beneficiaries were passively enrolled at the same time. In regular calls from their care coordinators and in-person 2015, the share of assessments that were completed within meetings or assistance in some states. In contrast, lower 90 days was between 55 percent and 75 percent for most risk enrollees appear to have much less regular contact demonstrations (Weiner et al. 2017). Completion rates are with their care coordinators, and their interactions are higher when beneficiaries who could not be located or did more likely to be limited to periodic phone calls. not want to participate in an assessment are excluded, and RTI conducted focus groups of MMP enrollees in have been rising over time, from an average of 69 percent five states and found that most knew they had a care in 2014 to 78 percent in 2015 and 89 percent in 2016 coordinator or had interacted with that person. Most of (Centers for Medicare & Medicaid Services 2017g). the participants who had used care coordination found it Our interviews and RTI both found that plans had helpful, but some beneficiaries had not known they could difficulty with the next stage of the care coordination receive care coordination before they participated in the process—using interdisciplinary teams of providers to focus group (Ptaszek et al. 2017). Other reports have formulate care plans. One particular challenge has been found that care coordination has had a significant, positive low participation by primary care physicians, who are impact on some enrollees, leading to improvements in usually not paid for taking part (Weiner et al. 2017). their health and functioning (Carver 2016, Gattine et al. 2017, SCAN Foundation 2017). The MMPs have hired a significant number of care coordinators for the demonstration. In 2015, the plans During our later site visits, some plan representatives in the 9 demonstrations then in operation employed we interviewed indicated that the care coordination almost 4,600 care coordinators. Most coordinators have requirements for the demonstration were too prescriptive. backgrounds in social work or nursing; those who oversee Many of these comments focused on low-risk enrollees, enrollees with complex needs are more likely to have a with plans saying that their assessments did not need clinical background. About 80 percent of coordinators to be as comprehensive or be completed as quickly as worked on tasks such as providing care management those for higher risk enrollees. Another plan said that and conducting assessments; the rest worked in other interdisciplinary provider meetings were difficult to capacities such as supervision (Weiner et al. 2017). On schedule and were worthwhile only for beneficiaries with average, the MMPs have 1 care coordinator for roughly very complex needs. every 100 enrollees (if the coordinators working in other Care coordination requirements have been a major issue capacities are included, the ratio is closer to 1:80). in New York in particular, where overly prescriptive Care coordinators can work directly for the plan or one requirements appear to be the main reason that its of the plan’s subcontractors, such as a medical group or first demonstration has had such low participation. social service agency. Most of the plans we interviewed The stakeholders we interviewed said that beneficiary used a mix of these approaches, and many plans had advocacy groups had played a large role in developing the June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 257

280 demonstrations had been under way for 18 to 24 months. requirements, which were modeled after those used in the The plans we interviewed at the time had not yet seen Program of All-Inclusive Care for the Elderly (PACE), and noticeable changes in their enrollees’ service use and said that there had been relatively little input from physicians. it was unrealistic to expect savings that quickly given the Under the requirements, members of the interdisciplinary initial implementation challenges that plans had faced. team of providers (which included the beneficiary’s care coordinator and primary care provider) had to meet at the On our later visits—when the demonstrations in same time, in person, to develop the beneficiary’s care California, Massachusetts, New York, Ohio, and Texas had plan. Beneficiaries were also expected to participate in been under way for about three years—plans were much the planning meetings, and primary care providers had to more definitive. Almost every plan we interviewed said the complete training on the care planning process. use of inpatient care and emergency room visits by their enrollees had declined. (The MMPs in New York were an This approach to care coordination is feasible in PACE exception; they said they had not seen significant changes because of the central role that adult day-care centers play in service use.) Several plans said that nursing home use in that program. The providers on the interdisciplinary was also declining, although those reductions appeared to team all work at the center (and are employees of the be smaller. A few plans said they had seen lower service PACE plan) and enrollees typically visit the center several use in other areas, such as post-acute care and certain times each week to receive care. In-person meetings of the types of HCBS. However, we did not get a clear sense of care planning team, including beneficiaries if they desire, whether the use of other services like primary care had are thus relatively easy to arrange. changed. This approach did not work well in the demonstration, One particularly important area for many dual eligibles where enrollees receive care from multiple providers in is behavioral health. Many stakeholders we interviewed different locations, and providers were often expected said there was a shortage of behavioral health providers in to work with multiple plans. Stakeholders indicated that their area, but they saw this deficiency as a shortcoming of providers, especially primary care physicians, thought the the broader health care system rather than something that requirements were overly burdensome and encouraged was specific to the demonstration. Some stakeholders on their patients to opt out. One plan we interviewed said our later visits felt the demonstration had expanded access providers also opposed the demonstration because MMPs to care for individuals with moderate behavioral health could authorize only services that were explicitly listed in care needs—people who could benefit from treatment but an enrollee’s care plan (the interdisciplinary team had to did not have an illness that was severe enough to receive meet again to approve any additional services, even minor treatment from the traditional behavioral health care ones), and because providers had to attest that all of their system. facilities complied with the Americans with Disabilities Act, something they had never been required to do before. The plans we interviewed said consistently that inadequate housing had been a significant challenge in caring for CMS and New York moved relatively quickly to address some enrollees. For example, one plan said even a few these concerns, eliminating or scaling back many days in short-term housing could help homeless enrollees requirements during the first year of the demonstration. who had just been discharged from a hospital by making However, many stakeholders indicated that providers still it easier for them to get appropriate follow-up care. MMPs have a negative view of the demonstration, which has cannot spend funds on room and board for people who live made it difficult to increase enrollment. in the community (a long-standing policy in Medicaid), Service use and access to care but some plans we interviewed were trying to develop closer relationships with local housing agencies so they One key question about the capitated model has been could more easily help their enrollees find housing. whether MMPs can lower costs and improve the quality of care for dual eligibles by reducing their use of expensive Some states have included additional transportation services like inpatient care and nursing home care and benefits, such as nonmedical transportation, in their by promoting greater use of primary care and home- and demonstrations to help attract enrollment, but several community-based services (HCBS). When we made our stakeholders said the service was often unreliable. first site visits to California, Illinois, and Massachusetts However, Medicaid programs often have problems between December 2015 and February 2016, those Managed care plans for dual-eligible beneficiaries 258

281 TABLE ® survey has improved, 2015–2017 MMP performance on the CAHPS 9–7 2017 2015 2016 40 27 Number of MMPs reporting CAHPS data 45 Share of beneficiaries giving the highest rating for: Health plan 51 % 59 % % 63 Health care quality 55 59 60 58 59 Getting needed care 58 48 Getting appointments and care quickly 54 50 76 77 Doctors who communicate well 76 67 71 76 Customer service Care coordination 69 70 69 Getting needed prescription drugs 73 77 77 ® ® Note: MMP (Medicare–Medicaid Plan), CAHPS (Consumer Assessment of Healthcare Providers and Systems ). Except for the number of MMPs reporting data, the numbers in this table are the share of beneficiaries giving the highest rating (a 9 or 10 on a 10-point scale or answering “always” when asked about the ability to get appointments when needed). Rates are case-mix adjusted for response bias. Source: CAHPS survey results for MMPs released by CMS in April 2016, July 2017, and December 2017. providing transportation benefits, and it was not clear For example, CAHPS results for 2017 measure patient experience in late 2016 and early 2017. whether the problems that the MMPs had encountered in this area were any worse. CMS has released overall CAHPS results for MMPs for 2015 through 2017 (Table 9-7). The number of plans that Quality of care reported data grew as the individual state demonstrations Improving the quality of care for dual eligibles is one were implemented, increasing between 2015 and 2017 of the primary goals of the demonstration. MMPs are 12 from 27 plans to 45 plans. During that period, MMP required to submit quality data to help CMS and states performance on all measures either improved or remained oversee the demonstration and evaluate its impact. stable, with the share of enrollees giving their plan the Some requirements are modeled after the MA and Part highest possible rating rising from 51 percent to 63 D programs, while others were developed specifically percent. Enrollees also reported improvements in overall for MMPs. The MMP-specific measures are a mix of health care quality, getting appointments and care quickly, process and structure measures, such as completing health customer service, and getting needed prescription drugs. assessments on time and establishing a consumer advisory board, and utilization measures, such as emergency These results naturally raise the question of how MMPs room visits related to behavioral health and diversion of perform compared with MA plans and FFS. We do beneficiaries from nursing homes (Centers for Medicare & not have the data to make this comparison based on Medicaid Services 2017e). the method used to report CAHPS results in Table 9-7, which shows the share of beneficiaries providing the Patient experience One source of quality information is highest rating for each metric. However, we can compare the Consumer Assessment of Healthcare Providers and CAHPS results using another method that calculates the ® ® Systems (CAHPS ), a beneficiary survey that measures average score on each metric for all survey respondents patient experience. Like MA plans, MMPs are required and rescales that average so it ranges between 0 percent to administer the CAHPS survey each year. The survey and 100 percent. Using this approach, the results for is usually conducted in the spring and asks enrollees to MMPs, MA plans, and FFS are quite similar, with about assess their experience during the previous six months. June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 259

282 TABLE ® measures, Performance of MMPs and MA plans on HEDIS 9–8 based on full-benefit dual eligibles only, measurement year 2016 Enrollees Enrollees under age 65 ages 65 and older Number of HEDIS measures evaluated 40 43 Number of measures where: MMP and MA performance was similar 18 18 15 14 MA plans performed better than MMPs MMPs performed better than MA plans 10 8 ® ® Note: MMP (Medicare–Medicaid Plan), MA (Medicare Advantage), HEDIS (Healthcare Effectiveness Data and Information Set ). Better performance means that the average measure value for one type of plan was more than 5 percent greater than the average measure value for the other type of plan. MedPAC analysis of HEDIS data for 2017 (for measurement year 2016) and common Medicare environment and denominator files. Source: percent of enrollees in each sector giving their health to mental health (providing follow-up care within 7 days 85 and 30 days of an inpatient mental health admission), and plan the highest possible rating. two measures related to substance abuse (initiation of and Another source of quality Clinical quality measures engagement in substance abuse treatment). information that MMPs and MA plans both submit is ® We also compared HEDIS results for MMPs in 2015 and the Healthcare Effectiveness Data and Information Set ® 2016 and found that MMP performance had improved, (HEDIS ), a set of clinical quality measures. We used on balance. We made this comparison by finding out how HEDIS person-level data to compare MMP enrollees many MMPs (measured at the contract level) improved with full-benefit dual eligibles who were enrolled in MA 13 on a given measure during this period. There were 33 plans. We looked separately at enrollees who were measures we could examine on this basis; a plurality of under 65 and enrollees who were 65 and older because the MMPs improved on 12 measures, did worse on 8, and under-65 population tends to have poorer HEDIS results. showed no change on the other 13. Our evaluation of HEDIS data for 2016 produced mixed There are several caveats to our analysis. First, we used results (Table 9-8). We found that MMPs and MA plans full-benefit dual eligibles in MA plans as a comparison had similar results for roughly 40 percent to 45 percent of group for MMP enrollees, but there could be systematic the measures that both plans collect (18 of 40 measures differences between the two groups that affect their for enrollees under 65; 18 of 43 measures for enrollees HEDIS results. For example, MA enrollees actively 65 and older). MA plans performed better on a third of enrolled in their plans, while most MMP enrollees were the measures, while MMPs performed better on about 20 passively enrolled and were difficult to contact in some percent to 25 percent of the measures. cases. Second, older, more established plans tend to MA plans performed substantially better than MMPs perform better than new plans on quality measures, and on three measures: control of blood sugar among MMPs are still relatively new compared with MA plans. diabetics, osteoporosis management for women who have Finally, the 2 types of plans have different financial experienced a fracture, and medication reconciliation after incentives when it comes to quality measures: 11 HEDIS a hospital discharge. MMPs’ poor performance on the last measures are used in the MA star rating system while only measure is particularly concerning since they should pay 2 HEDIS measures are used in the quality incentive for close attention to transitions in care settings as part of their MMPs, which is known as the “quality withhold.” Many care coordination efforts. MMPs performed better than of the measures on which MA plans performed better are MA plans (for both age groups) on five measures: control used in the star rating system but not the quality withhold, of blood pressure among diabetics, two measures related Managed care plans for dual-eligible beneficiaries 260

283 while the reverse is true for some measures on which on historical FFS experience. For Part D drugs, MMPs MMPs performed better. are paid based on the national average bid for all Part D plans. Like Part D plans, MMPs receive a capitated direct Development of a star rating system for MMPs In 2015, subsidy payment as well as prospective payments for CMS began developing a star rating system for MMPs. estimated reinsurance costs for beneficiaries with high CMS does not expect to have a fully developed system drug costs and for beneficiary cost sharing covered by ready during the demonstration; the agency is working the Part D low-income subsidy, which all dual eligibles instead to prepare for the possibility that the Secretary receive. The two Medicare capitation payments are would expand the use of the capitated model in the future adjusted for differences in beneficiaries’ health status using CMMI authority. The MMP ratings will differ from using the same risk adjustment models that are used in the star ratings for MA plans because MMPs will be MA and Part D. assessed on their performance in providing both Medicare For Medicaid benefits, each state determines its own and Medicaid services. For example, the ratings for MMPs payment rates, subject to CMS approval. The rates include will incorporate measures related to LTSS and Medicaid- both federal and state Medicaid spending and typically covered behavioral health services (Centers for Medicare vary based on beneficiaries’ use of LTSS. Medicaid rates & Medicaid Services 2015e). The rating system will be are typically highest for beneficiaries in nursing homes tested before being used and will account for differences and lowest for those not receiving any LTSS, with rates in beneficiaries’ socioeconomic status where appropriate. for beneficiaries receiving HCBS somewhere in between. CMS will decide in the future whether the star ratings will Some states have also “carved out” certain benefits from be used to adjust MMP payments, but it has indicated that the demonstration and continue to provide them through MMPs would not be subject to payment adjustments under FFS arrangements. both the quality withhold and the star ratings at the same time (Centers for Medicare & Medicaid Services 2016c). CMS and states also reduce the Part A and Part B and Medicaid capitation rates (there is no reduction to the Taken together, Lessons from CAHPS and HEDIS results Part D capitation rate) by a certain percentage to reflect the CAHPS and HEDIS results indicate that the quality savings they assume the MMPs will be able to produce of care provided by MMPs is improving, but the plans do under the demonstration. The savings percentages vary by not perform as well as MA plans in some areas. As CMS demonstration but are generally around 1 percent in the develops a star rating system for MMPs, it may want first year, 1 percent to 2 percent in the second year, and 2 to put particular emphasis on measures where MMPs percent to 5 percent in later years. currently have poor performance. The findings from our examination of HEDIS results—with MA plans tending In 2016, CMS increased MMP payment rates for Part A to perform better than MMPs on measures that are used and Part B services after finding that the existing MA risk in the MA star rating system but not the MMP quality adjustment model underestimated costs for full-benefit withhold, and vice versa—suggest that plans pay closer dual eligibles (Centers for Medicare & Medicaid Services attention to the measures used to determine their quality 2015b, Centers for Medicare & Medicaid Services rating, particularly if that rating affects their payments. 14 2015f). This change raised the payment rates for most MMPs by about 5 percent to 10 percent and was viewed Payment adequacy favorably by the plan representatives we interviewed. Under the capitated model, MMPs receive three separate During our early visits—which took place in late 2015 capitation payments: one for Part A and Part B services, and early 2016, after the increase in payment rates had one for Part D drugs, and one for Medicaid services. The been announced but not yet implemented—stakeholder payment methodology for MMPs differs from those used views on the adequacy of the MMP rates varied greatly. in MA and Part D because MMPs do not submit bids. Many interviewees in Massachusetts said the existing Instead, for Part A and Part B services, MMPs are paid rates were too low and the initial savings assumptions had using county-specific rates that are based on historical proven to be unrealistic. Interviewees in California and FFS and MA spending for beneficiaries who meet the Illinois did not express any significant concerns about demonstration’s eligibility criteria. In most states, the the rates, although they also thought the initial savings eligible population was largely enrolled in FFS Medicare assumptions were not realistic. On our later visits, none before the demonstration, so the rates are based primarily June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 261

284 of the stakeholders we interviewed (including those we withhold, while those that pass between 20 percent and 80 met with on a follow-up visit to Massachusetts) raised percent of the measures receive part of the withhold (either any significant concerns about Medicare’s rates, which 25 percent, 50 percent, or 75 percent), and plans that pass suggests that the current rates are adequate. more than 80 percent receive the entire withhold (Centers for Medicare & Medicaid Services 2014). Quality incentives for MMPs MMP payments are also tied to the plans’ performance on certain quality measures The only data on MMP performance for the quality through a quality withhold. Under the withhold, the Part withhold that are currently available are for 2014, when five A and Part B and Medicaid components of the MMP demonstrations (California, Illinois, Ohio, Massachusetts, payment rates are reduced by a specified percentage and Virginia) were under way. The lack of data is likely (usually 1 percent in the first year of the demonstration, due to the same problems with data availability that have 2 percent in the second year, and 3 percent in later hindered work on the demonstration’s evaluations. For years) that MMPs can receive later depending on their 2014, MMPs received about 70 percent of the quality performance. withhold, on average. Every MMP received at least some of the withheld funds, and a third of plans received the MMPs are assessed on their performance on a full amount (Centers for Medicare & Medicaid Services combination of “core” measures that are used in all 2017a, Centers for Medicare & Medicaid Services 2017b, capitated demonstrations and state-specific measures. Centers for Medicare & Medicaid Services 2017d, Centers There are five core measures for the first year of the for Medicare & Medicaid Services 2017h, Centers demonstration and seven core measures for later years; the for Medicare & Medicaid Services 2017k). Since the number of state-specific measures varies, with most states quality withhold equaled 1 percent in 2014, that level of having between two and five measures. For the first year, performance means the quality withhold reduced the Part most measures are related to plan administration (such as A and Part B and Medicaid payments to MMPs by about submitting encounter data and completing assessments) 0.3 percent, on average. If MMPs perform at a similar level or patient experience (such as customer service) (Centers once the quality withhold reaches its ultimate level of 3 for Medicare & Medicaid Services 2014). For later years, percent, the withhold will reduce plan payments by roughly plans are assessed largely on clinical quality or outcome 1 percent, on average. measures such as readmission rates, medication adherence for diabetes medications, and nursing home use (Centers The quality withhold differs in several respects from the for Medicare & Medicaid Services 2016b). quality bonus program in Medicare Advantage, in which plans that have ratings of 4 stars or better and submit bids CMS and states determine whether plans “pass” that are lower than the MA benchmarks receive additional each measure by comparing their performance with a funding that they use to provide extra benefits to their benchmark. The benchmarks for the core measures are enrollees: absolute, meaning they do not change based on how other MMPs perform. In contrast, for the state-specific • The quality incentive for MA plans is structured as measures, some benchmarks are absolute while others are a bonus, while the quality incentive for MMPs is relative, meaning the benchmark depends on how other structured as a penalty. MMPs perform. For example, the benchmark for several MA plans are assessed on more measures (43) than • state-specific measures is the performance of the state’s MMPs (about a dozen measures in most states). highest scoring MMP minus 10 percentage points (Centers However, the smaller number of measures for MMPs for Medicare & Medicaid Services 2015c, Centers for is partly due to the lack of good quality measures for Medicare & Medicaid Services 2015d). Starting in the LTSS and care coordination, which are still being second year of the demonstration, plans can also pass developed. all core measures and some state-specific measures by improving their performance by a sufficient amount MA plans receive a star rating on each individual • 15 (Centers for Medicare & Medicaid Services 2016b). measure, and those ratings are combined into At the end of each year, CMS and states determine what an overall star rating. MA plans cannot improve share of the measures each MMP has passed, with each their rating on any individual measure by showing measure weighted equally. Plans that pass fewer than 20 improvement, while MMPs can “pass” most measures percent of the measures do not receive any of the quality Managed care plans for dual-eligible beneficiaries 262

285 by showing sufficient improvement. However, MA most of the actual assistance to beneficiaries (Medicare plans can receive a higher overall star rating if they Payment Advisory Commission 2016). RTI has released show improvement across multiple measures. an evaluation that covers the first 18 months of the demonstration (July 2013 to December 2014). Much like • The MA quality bonus is an all-or-nothing the initial report for the Massachusetts demonstration, the proposition; plans either receive the entire bonus or evaluation found “little evidence of the demonstration’s receive nothing. In contrast, MMPs can receive part of effect” during its initial period of operation. In the quality withhold. Washington’s case, the initial impact of the demonstration may have been limited because dual eligibles were Given these differences and the work that CMS has begun enrolled gradually, some health homes found they needed to develop a star rating system for MMPs, it is unclear to develop more capacity for providing care coordination, what kind of quality incentive MMPs might face if the and health homes found it challenging to engage enrollees Secretary expands the use of the capitated model. (Justice et al. 2017). Demonstrations using the managed fee-for- At the end of each year, states can receive a “performance service model payment” if the demonstration produces savings for Unlike the capitated model, which relies on managed the federal government. CMS calculates the savings by care plans to improve care and reduce costs, the managed comparing Part A and Part B spending for beneficiaries in FFS model aims to achieve those goals by providing the demonstration with an estimate of how much Medicare greater care coordination in an FFS environment. Two would have spent without the demonstration. Savings must states—Colorado and Washington—have been testing the be at least 2 percent for the state to receive a performance managed FFS model. Colorado ended its demonstration at payment (to guard against random variation in program the end of 2017; Washington’s demonstration is scheduled spending), and CMS deducts any additional Medicaid to end in 2018 but may be extended until 2020. costs when calculating the overall federal savings. The state’s performance payment equals 30 percent to 50 Under the managed FFS model, the state passively enrolls percent of the federal savings, depending on the state’s dual eligibles who have both FFS Medicare and FFS performance on certain quality measures. Medicaid in a Medicaid-funded entity that is responsible for providing care coordination. Beneficiaries can receive In July 2017, CMS released a report estimating that care coordination services from the entity, but their Washington’s demonstration reduced Medicare spending participation is entirely optional, and they remain enrolled by $67 million during its first two and a half years of in FFS Medicare and FFS Medicaid regardless. Colorado operation (July 2013 to December 2015), a savings of enrolled all FFS dual eligibles in its demonstration, while about 9 percent (Wilkin et al. 2017b). That figure was Washington has focused on a subset of dual eligibles who based on an estimate of what Medicare would have spent are expected to have high costs. on the dual eligibles who were assigned to a health home (about 20,000 beneficiaries) without the demonstration. Colorado’s demonstration was part of a broader effort to As noted in our June 2016 report, we are skeptical that improve care coordination in FFS Medicaid known as the savings from the demonstration could be that large the Accountable Care Collaborative (ACC). The ACC because the number of beneficiaries who actually received provides care coordination through entities that function care coordination services during this period was relatively somewhat like accountable care organizations. The state low—about 3,000 people, many of whom received care had excluded dual eligibles from the ACC when it was coordination for only part of the time. As for Colorado, an first developed and added them through the demonstration. August 2017 report estimated that its demonstration had Although the demonstration is now over, the state has actually increased Medicare spending by $10 million in its decided that dual eligibles will remain in the ACC, and first 15 months of operation (September 2014 to December there should be little day-to-day change in their care. 2015), a cost of about 4 percent (Wilkin et al. 2017a). Both reports note that their findings are preliminary and do not The Washington demonstration relies on entities known 16 account for any changes in Medicaid spending. RTI also as health homes to provide care coordination, with plans to estimate the savings from the demonstrations organizations such as area agencies on aging, mental using more rigorous, regression-based methods as part of health clinics, and community health centers providing its evaluations. June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 263

286 materials (such as provider directories) instead of Overall assessment of the financial separate Medicare and Medicaid versions, and all alignment demonstration MMPs have integrated at least some parts of the Despite the conceptual appeal of integrated plans, their use grievance and appeals processes. in Medicare has always been limited. About 30 percent of full-benefit dual eligibles are now enrolled in some type These features helped generate widespread state interest in of Medicare managed care plan, but the extent to which the demonstration, with 21 states submitting proposals to those plans integrate with Medicaid varies widely. Even test the capitated model (Medicaid and CHIP Payment and with the demonstration, only 8 percent of full-benefit dual Access Commission 2018). And while the demonstrations eligibles are enrolled in plans that have a high degree of were often challenging to implement, the experience so far 17 integration. Before the demonstration, the figure was suggests that integrated plans can be developed in many about 2 percent. states. With a few exceptions, each state’s demonstration has now been under way for at least three years. The The limited use of integrated plans has traditionally been continued delays in the evaluations are a significant attributed to several factors. First, states do not benefit concern given the widespread interest in understanding financially from any Medicare savings that integrated the demonstration’s impact on access to care, service use, plans might realize and, thus, have less incentive to costs, and quality. Nevertheless, much of the information develop such plans. Second, integrated plans have found that is currently available, while limited, is relatively it difficult to generate substantial enrollment because positive: Enrollment is stable, quality of care appears dual eligibles cannot be required to enroll in a plan to to be improving, payment rates appear adequate, plans receive their Medicare benefits. Third, CMS and states have grown more confident about their ability to manage do not have the authority to resolve the many differences service use, and stakeholders remain supportive of the between Medicare and Medicaid that make it harder to demonstration. operate an integrated plan, such as separate grievances and appeals processes and different adequacy requirements for 18 provider networks. Finally, states and health plans have had little experience using managed care to deliver LTSS, More states are using Medicaid which has made it difficult to develop integrated plans. managed care for dual eligibles The experience with the demonstration suggests that States’ interest in testing the capitated model in the policy changes addressing these barriers could lead to financial alignment demonstration has been part of a greater interest by states and health plans in developing broader shift toward the use of Medicaid managed care integrated plans: for the aged and disabled. Managed care has long been • The demonstration allows states to benefit financially the dominant delivery system in Medicaid for populations from the savings that MMPs are expected to achieve in such as children, pregnant women, and nondisabled adults. Medicare by applying the same savings assumptions For example, 25 of the 32 states (including the District to both the Medicare Part A and Part B and Medicaid of Columbia) that expanded Medicaid coverage for low- components of the MMP payment rates. Even if income adults under the Patient Protection and Affordable MMPs ultimately achieve their savings entirely by Care Act of 2010 enrolled at least 80 percent of those new 19 lowering Medicare costs, states still benefit financially. beneficiaries in managed care (Paradise 2017). CMS made it easier for MMPs to generate enrollment • However, for many years, states were much less likely to by allowing states to use passive enrollment. Many use managed care for their aged and disabled enrollees, MMPs we interviewed said passive enrollment was many of whom are dual eligibles. LTSS represents a a key factor in their decision to participate in the significant share of Medicaid spending on aged and demonstration. disabled enrollees—about 80 percent for dual eligibles and 35 percent for those who have Medicaid only. LTSS CMS has used demonstration authority to address • presents distinct challenges to health plans because its some of the administrative challenges involved in services and providers can differ greatly from traditional operating integrated plans. For example, MMPs use a medical services, and the number of health plans that single identification card and a single set of member had “both the experience and the ability to accept risk for Managed care plans for dual-eligible beneficiaries 264

287 LTSS” was limited (Saucier et al. 2012). As recently as plans often include financial incentives to serve enrollees 2004, only eight states had programs that used managed in community settings where possible (Dominiak and care plans to deliver LTSS to at least some beneficiaries Libersky 2016). 20 (Saucier et al. 2012). In addition, a state cannot require Many MLTSS programs have features that are dual eligibles to enroll in Medicaid managed care unless commonplace in Medicaid managed care but can differ it first obtains a waiver from CMS, a process that can take substantially from the Medicare Advantage program: 21 up to two years. (States do not need a waiver to require most other beneficiaries to enroll in managed care.) When • Most states require at least some beneficiaries to states require dual eligibles to enroll in Medicaid managed enroll in managed care to receive their Medicaid- care, the requirement applies only to the delivery of their covered services, while enrollment in MA plans is Medicaid services, not their Medicare services. voluntary. As a result, dual eligibles in those states may be required to enroll in an MLTSS plan for their Since 2004, the number of states with these programs— Medicaid-covered services, but the same requirement often referred to as managed LTSS (MLTSS) programs— does not apply to Medicare; for example, they can has grown rapidly, from 8 states in 2004 to 16 states in select FFS Medicare coverage or an MA plan, 2012 and 24 states today (Lewis et al. 2018, Saucier et al. which may or may not be offered by the same parent 2012). Medicaid spending on MLTSS programs has also company that sponsors their Medicaid plan. Some grown significantly; between 2009 and 2015, spending states require the sponsors of their MLTSS plans to rose from $7 billion (5 percent of all Medicaid LTSS offer a companion MA dual-eligible special needs spending) to $29 billion (18 percent of all Medicaid LTSS plan so beneficiaries can receive their Medicare and spending) (Eiken et al. 2017, Eiken et al. 2016). The use of Medicaid benefits from the same parent company if MLTSS will likely grow in the future as additional states they wish. develop MLTSS programs and states that already have programs expand them. • States use competitive procurements to select a limited number of plans to participate in the program. This We are not aware of a data source that indicates how many approach increases the likelihood that all participating dual eligibles are currently enrolled in MLTSS plans. A plans will have enough enrollment to be financially recent report found that about 1.8 million individuals were viable, helps the state obtain lower payment rates, enrolled in MLTSS programs (using a combination of and makes oversight of the plans easier. Medicaid 2016 and 2017 data), but that figure includes Medicaid- generally requires states to have at least two plans only beneficiaries, so the number of dual-eligible enrollees available before they can require beneficiaries to would be lower (Lewis et al. 2018). In rough terms, we enroll in managed care, and, in practice, states often estimate that perhaps 15 percent of full-benefit dual contract with at least three plans to ensure that eligibles were in MLTSS plans in 2017. However, the mandatory enrollment in managed care can continue 24 states that now have MLTSS programs collectively even if one plan drops out. In contrast, Medicare does account for about 80 percent of all full-benefit dual not limit the number of MA plans available in an area, eligibles. If these states expand the scope of their MLTSS although CMS requires all plans to satisfy a variety programs in the future, the share of dual eligibles enrolled of requirements such as provider network adequacy in MLTSS plans could rise significantly. standards. States have been developing MLTSS programs for three • States typically have multiyear contracts with their main reasons. First, they hope that managed care will MLTSS plans, which gives the state flexibility in lower Medicaid spending and make future spending deciding when to conduct its next procurement and growth more predictable. Second, they hope that MLTSS gives plans a greater incentive to participate, offer plans will improve the quality of care by providing competitive rates, and invest in care coordination. effective care coordination for LTSS users, who often have Many contracts have a base period and can be complex health needs. Finally, states see MLTSS programs extended for an additional period by the state at as a way to encourage the use of HCBS instead of nursing its discretion. For example, the latest contract for home care (Libersky et al. 2016). For example, some states Arizona’s MLTSS plans has a three-year base period have liberalized the eligibility criteria for HCBS as part and three optional renewals (a two-year renewal of their MLTSS programs, and payment rates for MLTSS June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 265

288 followed by two one-year renewals) for a potential of highly integrated plans. However, the experience with the financial alignment demonstration also suggests that total length of seven years (Arizona Health Care Cost operating multiple types of plans targeted at dual eligibles Containment System 2016). By comparison, the MA at the same time can be problematic. Policy changes to program uses annual contracts. better define their respective roles or consolidate them in • Many states exclude some groups of enrollees from some fashion may be needed. their MLTSS programs. For example, states have been slower to enroll individuals with developmental In addition to MMPs, Medicare has three other types of disabilities in MLTSS plans. In 2015, MLTSS health plans that serve dual eligibles and seek to integrate accounted for 24 percent of LTSS spending with Medicaid in some way: for enrollees who were elderly or had physical are MA Dual-eligible special needs plans (D–SNPs) • disabilities, but only 7 percent for enrollees with plans that limit their enrollment to dual eligibles. (In developmental disabilities (Eiken et al. 2017). Partial- contrast, most MA plans are open to all beneficiaries benefit dual eligibles are also routinely excluded from in the plan’s service area.) These plans were first MLTSS programs. States may also initially limit their offered in 2006. The authority to offer D–SNPs programs to certain parts of the state and expand them was initially set to expire at the end of 2008 but once they have gained experience. was extended numerous times before the Congress • Some states may exclude or “carve out” certain permanently authorized them earlier this year in the services from their MLTSS programs and provide Bipartisan Budget Act of 2018. Since 2013, Section them separately. For example, MLTSS plans in a 1859(f)(3)(D) of the Medicare statute has required number of states exclude at least some behavioral all D–SNPs to have contracts with states that “provide health services. However, as Medicaid managed [Medicaid] benefits, or arrange for [such] benefits to care programs mature, states tend to reduce the use be provided.” of carve-outs and make the coverage provided by • are a Fully integrated dual-eligible (FIDE) SNPs plans more comprehensive. In contrast, MA plans are subset of D–SNPs that are more highly integrated required to provide all Part A and B services, except with Medicaid than regular D–SNPs. These plans for hospice, and most plans (including all special must meet a number of additional requirements to needs plans) also provide Part D drug coverage. obtain the FIDE SNP designation, such as having a Given the growth in MLTSS programs, efforts to better Medicaid contract to provide LTSS, and can receive integrate Medicare and Medicaid in many states now take higher payments if their enrollees have sufficiently place in an environment where managed care is already high frailty levels. The FIDE SNP designation became being used to provide some services to dual eligibles. As available in 2012. Like regular D–SNPs, these plans a result, the development of health plans that provide both have now been permanently authorized. Medicare and Medicaid services is probably the most Program of All-Inclusive Care for the Elderly • feasible approach for pursuing closer integration. plans serve beneficiaries who are 55 or older (PACE) and need the level of care provided in a nursing home. This program is not specifically targeted at dual Medicare plans that serve dual eligibles eligibles like D–SNPs are, but, in practice, virtually all PACE enrollees are full-benefit dual eligibles. The differ in key respects program aims to keep people living in the community instead of nursing homes, and it uses a distinctive Although the use of managed care appears to be the most model of care based on adult day-care centers that feasible route for better integrating Medicare and Medicaid are staffed by an interdisciplinary team that provides in many states, this broad concept can be implemented therapy and medical services. PACE plans provide in numerous ways. Medicare has several types of all Medicare- and Medicaid-covered services. PACE health plans that are aimed at serving dual eligibles but is the oldest type of integrated plan; it started as a nonetheless differ in key respects. Comparing these plans demonstration in the early 1980s and was permanently highlights some of the issues that policymakers may want authorized in 1997. to consider if they decide to encourage the development Managed care plans for dual-eligible beneficiaries 266

289 TABLE Key differences between Medicare plans that serve dual eligibles 9-9 D–SNP MMP PACE FIDE SNP Regular Permanent Demonstration Permanent Authorization Permanent 9 31 States where plan 41 9 is available 45 348 124 Number of plans 50 Enrollment 41,079 383,047 159,158 1,695,074 Contracting Separate Medicare and Separate Medicare and Single 3-way contract Single 3-way contract with CMS & state structure Medicaid contracts with CMS & state Medicaid contracts Varies widely but Level of integration High High High generally low <1% <1% <1% 28% Share of enrollees who are partial- benefit dual eligibles Allowed Allowed Passive enrollment Allowed Not allowed for default for default enrollment only enrollment only Plan can provide Ye s Yes, using Yes, varies by Yes, using MA rebates state and plan MA rebates noncovered benefits Plans bid against Rates are set Plans bid against Medicare payment Rates are set MA benchmarks administratively MA benchmarks administratively methodology No* Yes, if frailty levels are No Plan eligible for Ye s similar to PACE enrollees frailty adjustment No No Ye s No States can share Medicare savings Type of quality None Quality MA quality MA quality withhold bonus program bonus program incentive D–SNP (dual-eligible special needs plan), FIDE SNP (fully integrated dual-eligible special needs plan), MMP (Medicare–Medicaid Plan), PACE (Program of All- Note: Inclusive Care for the Elderly), MA (Medicare Advantage). Figures do not include Puerto Rico. Many states have more than one type of plan. The number of D–SNPs and FIDE SNPs are based on unique combinations of contract and plan number; the number of MMPs and PACE plans are based on unique contracts. Enrollment figures are for January 2018. The figures for the share of enrollees that are partial-benefit dual eligibles are based on enrollment data for December 2016. *Starting in 2019, the MMPs in New York’s first demonstration will be eligible for a frailty adjustment if the frailty levels of their enrollees are similar to those in PACE. Dual eligibles can also enroll in other types of plans, term regular D–SNP refers to a D–SNP that is not a such as regular MA plans and special needs plans for FIDE SNP. Regular D–SNPs are the most widely used individuals who live in long-term care institutions or have type of plan, with 348 plans in 40 states and the District certain chronic conditions. of Columbia covering almost 1.7 million beneficiaries in January 2018. The use of FIDE SNPs is much more The key features for each type of plan, as well as MMPs, limited; these are available in only 9 states and cover are summarized in Table 9-9. For this comparison, the about 159,000 beneficiaries, with 3 states (Massachusetts, June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 267

290 Minnesota, and New Jersey) accounting for about 75 adding additional provisions to their D–SNP contracts (see 22 percent of the overall enrollment. Only 11 percent of all text box on D–SNPs). D–SNPs (45 of 393 plans) are FIDE SNPs. Finally, PACE CMS found in 2016 that about 75,000 full-benefit dual plans are available in 31 states, but they are typically eligibles in regular D–SNPs received all of their Medicare small, and overall enrollment has always been fairly low and Medicaid services from the same parent company and (now about 41,000). that another 75,000 received all of their Medicare services The differences among the plans start with their and a majority of their Medicaid services from the same contracting structure. All D–SNPs have a standard MA company (Centers for Medicare & Medicaid Services contract with CMS to provide Medicare services and a 2017f, Centers for Medicare & Medicaid Services 2017i). separate contract with the state that details their Medicaid Those figures indicate that only about 15 percent of the responsibilities. In contrast, MMPs and PACE plans sign full-benefit dual eligibles in regular D–SNPs are in plans three-way contracts with CMS and the state that combine that may have a significant level of Medicaid integration. all of their Medicare and Medicaid responsibilities into a The other three types of plans have higher levels of single document. For MMPs, each demonstration also has integration. FIDE SNPs are required to cover Medicaid a contract management team (CMT) composed of state LTSS services, although they are not required to cover Medicaid officials and multiple CMS representatives that behavioral health. They must also have a single enrollment oversees the day-to-day management of the three-way process, an integrated model of care that covers both contract. RTI found that both sides think the CMT has Medicare and Medicaid services, and coordinated been “a very successful vehicle for joint oversight of MMP Medicare and Medicaid assessment processes (Gibbs performance” (Chepaitis et al. 2015). and Kruse 2016). These requirements are similar to On some site visits, we asked state Medicaid officials some of the requirements for MMPs, but the level of and MMP representatives if they preferred the three-way integration in MMPs is higher because they provide all or contract over the more traditional approach of separate almost all Medicaid-covered services, and more of their Medicare and Medicaid contracts. All interviewees that administrative processes have been combined. PACE is had an opinion preferred the three-way contract. Both completely integrated because its plans are required to states and plans said that the initial development of provide all Medicare and Medicaid services. the three-way contract had been time consuming and Although all four plan types serve dual eligibles, the share challenging but that it had been easier to administer and of enrollees who are partial-benefit dual eligibles—whose oversee once in place. However, Medicaid officials in one Medicaid coverage is limited to Medicare premiums and, state said the process for amending the three-way contract in some cases, cost sharing—is much higher in regular D– could be simplified. One plan we interviewed also said SNPs (28 percent) than in the other plan types (less than the three-way contract was helpful in getting its parent 1 percent in each). D–SNPs can cover partial-benefit dual company’s Medicare and Medicaid divisions to work eligibles as long as the state agrees to it in its Medicaid together more closely. contract, while MMPs cannot cover them under the terms 23 The level of integration between regular D–SNPs and of the demonstration. Partial-benefit dual eligibles Medicaid varies widely but is generally low. Since 2013, can join PACE if they meet the program’s eligibility all D–SNPs have been required to have Medicaid contracts requirements, but, in practice, very few enroll. PACE that meet certain minimum requirements. For example, the plans must provide all Medicaid-covered services to their contract must specify which categories of dual eligibles enrollees, regardless of their actual Medicaid eligibility, can enroll, the plan’s service area, the Medicaid benefits and any enrollees who are not eligible for full Medicaid the plan will cover, and the plan’s responsibility to provide benefits have to pay a substantial premium equal to the or arrange for Medicaid benefits. However, states are not plan’s monthly Medicaid capitation payment. required to contract with D–SNPs to provide any Medicaid The plans also differ in the amount of flexibility they services, let alone services such as LTSS or behavioral have to spend their Medicare and Medicaid revenues health. Plans that do provide Medicaid services may cover on services that are not covered by either program. only a limited subset, such as Medicare cost sharing or Supporters of integrated plans argue that giving plans certain acute care services. At the same time, states that a significant degree of flexibility would result in better wish to achieve higher levels of integration can do so by Managed care plans for dual-eligible beneficiaries 268

291 Using D–SNPs to promote Medicare–Medicaid integration for dual eligibles when combined with the first set of requirements number of states are using Medicare discussed above, creates a one-to-one relationship Advantage (MA) dual-eligible special needs between a state’s Medicaid plans and its D–SNPs. plans (D–SNPs) as the vehicle for more A closely integrating Medicare and Medicaid for dual- A few states have taken additional steps to • eligible beneficiaries. States have promoted integration encourage dual eligibles to enroll in a D–SNP by adding extra requirements to their Medicaid and a Medicaid plan offered by the same parent managed care contracts and the contracts that D–SNPs company. Massachusetts, Minnesota, and New are required to sign with state Medicaid agencies. Jersey prohibit their D–SNPs from enrolling These requirements are designed to increase the beneficiaries who are not also enrolled in the number of dual eligibles who are enrolled in a D–SNP parent company’s companion Medicaid plan, while and a Medicaid managed care plan offered by the same Arizona periodically reassigns some dual eligibles parent company. The Integrated Care Resource Center, to a new Medicaid plan that “matches” their D– a technical assistance entity sponsored by CMS, SNP (i.e., both are offered by the same parent reviewed the contracts in many of these states and company). provided some examples of these extra requirements: • Some states also require their D–SNPs to provide A growing number of states (at least 10 in 2018) • a variety of additional information about their require Medicaid plans that cover aged and disabled operations, such as encounter data, bid data, and beneficiaries (many of whom are dually eligible) any MA-related correspondence between CMS and and provide long-term services and supports to the plan. This added information makes it easier offer a companion D–SNP. States may also require for states to understand the Medicare side of their the D–SNP to serve the same geographic area as the integration efforts (Verdier et al. 2016). Medicaid plan. These provisions ensure that all dual eligibles enrolled in Medicaid managed care can Although these requirements can improve the receive their Medicare benefits from the same parent integration of Medicare and Medicaid for dual company if they wish. eligibles, their reach is nonetheless limited because Medicare’s freedom-of-choice provision prohibits • A smaller number of states (at least six in 2018) do states from requiring dual eligibles to enroll in not sign D–SNP contracts with companies unless Medicaid plans and D–SNPs from the same they sponsor Medicaid managed care plans in their organization. ■ state. This requirement eliminates any D–SNPs that do not have a companion Medicaid plan and, quality care because noncovered services could reduce Act of 2018 gives MA plans greater flexibility to offer overall costs and improve outcomes in some instances. supplemental benefits that are not primarily health related D–SNPs have had the least flexibility and can provide starting in 2020. MMPs have more flexibility than D– noncovered services only as a supplemental benefit using SNPs to spend their Medicare and Medicaid revenues rebates—the additional funding that MA plans receive on noncovered services. A state can require its MMPs if they submit a bid that is lower than the benchmark. to provide certain noncovered services or give each plan CMS has traditionally required these supplemental discretion to develop its own package of noncovered benefits to be primarily health related, but D–SNPs that services. PACE plans have the most flexibility in this area, meet certain integration requirements can use rebates with broad legislative authority to spend their Medicare to cover additional services for individuals who have and Medicaid revenues on noncovered services. functional impairments. In addition, the Bipartisan Budget June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 269

292 25 The ability to passively enroll beneficiaries in each type more dual eligibles enrolled in MMPs. Other states, of plan also varies. D–SNPs can passively enroll some like California and Texas, had higher D–SNP enrollment beneficiaries using an MA provision known as “default before the demonstration and now have a significant enrollment” or “seamless conversion” that allows an number of dual eligibles enrolled in both types of plans. insurer to automatically enroll individuals who have been The low participation in New York’s first demonstration in a comprehensive Medicaid managed care plan in a (which is largely due to care coordination requirements companion D–SNP when those individuals first become that were initially too prescriptive) has meant that D– eligible for Medicare. States’ use of passive enrollment in SNPs remain the state’s predominant plan type. Finally, MMPs has been a key feature of the financial alignment Massachusetts has both plan types, but they serve different demonstration. PACE plans cannot use passive enrollment. populations and do not overlap (its MMPs serve dual eligibles under age 65, while its D–SNPs serve those ages The final areas of difference among the plans are related 65 and older). to Medicare payment issues. Rates for D–SNPs are determined using the standard MA payment system, under The availability of both plan types and differences which plans bid against a predetermined benchmark that between the MMP and D–SNP models raise the prospect CMS calculates using local FFS costs. In contrast, MMPs that insurers and other entities such as insurance brokers and PACE plans do not submit bids and are instead paid may have financial incentives to favor the use of D–SNPs using rates that are set administratively. (The payment in some instances, which could hinder efforts to encourage rates for any Medicaid services that each type of plan dual eligibles to enroll in the more highly integrated provides are set separately.) Payment rates for all four plan MMPs. In some instances, allowing MMPs and D–SNPs types are adjusted for differences in beneficiaries’ health to operate in the same areas has been problematic. To status using the MA risk adjustment model. However, some extent, the friction between MMPs and D–SNPs PACE plans receive an additional payment, known as a was unavoidable for the demonstration since the states that frailty adjustment, because the model underestimates costs were most likely to be interested in the capitated model for beneficiaries with functional impairments. FIDE SNPs were also likely to be states that had already developed can also receive a frailty adjustment if the frailty level of D–SNPs. Nevertheless, the interplay between the two 24 their enrollees is comparable to PACE enrollees. plan types is worth exploring since the Secretary could MMPs use CMMI’s authority to expand the use of MMPs in the are the only type of plan where states share some of the future. savings that the plans are expected to achieve in Medicare. D–SNP and MMP rates both include quality incentives MMP payment rates for Part A and Part B services (through the MA quality bonus program and the quality can be higher or lower than D–SNP rates withhold, respectively), while PACE rates do not have a quality incentive. Payment rates for D–SNPs are determined using the same methodology that applies to all non-employer MA Allowing D–SNPs and MMPs to operate in plans. (The only exception is the frailty adjustment that the same areas has been problematic in some FIDE SNPs receive.) Each plan submits a bid that some states indicates the amount of funding that the plan requires to The financial alignment demonstration has effectively provide the Part A and Part B benefit package in a given given states that are testing the capitated model two service area. CMS compares the bid with a benchmark ways to use managed care to better integrate Medicare amount for the area, which is determined administratively and Medicaid on a large scale: D–SNPs and MMPs. and equals a certain percentage of local FFS costs. Although PACE is another option, it has never been used Benchmarks for counties in the highest spending quartile on a widespread basis and usually covers no more than 1 equal 95 percent of FFS costs, while benchmarks for percent to 2 percent of a state’s full-benefit dual eligibles. counties in the second, third, and fourth quartiles (with the fourth quartile having the lowest spending) equal 100 Each participating state has allowed both plan types to percent, 107.5 percent, and 115 percent of FFS costs, operate in certain markets, but the extent to which a state respectively. In addition, plans that have a rating of 4 stars relies on one type of plan versus the other varies. Some or higher in the CMS star system for MA plans also have states, like Illinois and Michigan, had relatively low a bonus amount, usually 5 percent of FFS costs, added to D–SNP enrollment before the demonstration and have their benchmark. Managed care plans for dual-eligible beneficiaries 270

293 If the plan’s bid is lower than the benchmark, the plan CMS risk adjusts payments to MA plans based on • receives a payment that equals its bid plus a “rebate” that enrollees’ demographic information and diagnosis equals a percentage (between 50 percent and 70 percent, codes from their claims. These adjustments are based depending on the plan’s star rating) of the difference on experience in the FFS program, but MA plans have between the benchmark and the bid. Plans that receive an incentive to submit more diagnosis codes than FFS rebates must use them to provide additional benefits to providers because doing so increases their payments. their enrollees, such as lower cost sharing for Part A and CMS partially accounts for the effect of this additional Part B services or coverage of supplemental benefits. If the coding by applying a “coding intensity adjustment” plan’s bid is higher than the benchmark, the plan receives that reduces payments to MA plans. MMPs have a payment that equals the benchmark and must charge the same incentive to submit more diagnosis codes, beneficiaries a supplemental premium that equals the but CMS has phased in the application of the coding difference between the bid and the benchmark. (Almost intensity adjustment to their payments, usually over all MA plans bid below their benchmarks.) Finally, the a three-year period. (New MA plans are subject to payment rates are adjusted for differences in beneficiaries’ the full coding intensity adjustment from the outset.) health status using the CMS–HCC risk adjustment model. The rationale for the phase-in is that most MMP enrollees were coming from the FFS program and did In contrast, MMPs do not submit bids; instead, CMS not have any additional coding. This transition period determines their payment rates using historical FFS has meant that MMPs have received higher payments and MA spending for beneficiaries who meet the during the first two years of the demonstration than demonstration’s eligibility criteria. In most states, these they would have if they had instead entered the beneficiaries were largely enrolled in FFS Medicare before market at the same time as D–SNPs. The increase the demonstration. The rates are then reduced to reflect has varied by state but, for most MMPs, has been MMPs’ expected savings and to set aside funding for the between 5 percent and 6 percent in the first year of the demonstration’s quality withhold. demonstration and 2 percent to 4 percent in the second year. During the demonstration, MMPs have benefited from a number of adjustments that increased their overall • In 2017, CMS began using a new risk adjustment payments compared with what they would have received model that raised payments to both MA plans and as D–SNPs. These adjustments have been largely MMPs for full-benefit dual eligibles. However, CMS temporary and have affected both the base payment rates also increased MMP rates for 2016 by amounts that for MMPs and how those rates are adjusted for differences approximated the extra payments that the plans would in beneficiaries’ health status: receive under the new model, effectively allowing MMPs to benefit from the new model a year earlier • CMS has increased the MMP rates in most than D–SNPs. The increase for most MMPs in 2016 demonstrations (9 of 11) to account for the bad was between 5 percent and 10 percent. debt payments that, without the demonstration, FFS Medicare would make to providers such as For this analysis, we compared MMP payment rates for hospitals for services provided to dual eligibles. Part A and Part B services with D–SNP benchmarks, MA benchmarks also include an allowance for bad which are both determined administratively by CMS. The debt payments, but it is smaller. This adjustment has MMP rates incorporate all of the adjustments described increased the FFS component of the MMP rates in above. We did not account for the effects of each plan’s most states by about 1.75 percent. quality incentive (i.e., we did not reduce MMP rates to account for the quality withhold or increase D–SNP • For 2013 and 2014, CMS “repriced” the claims benchmarks to account for the MA quality bonus) or the that were used to measure FFS costs to reflect frailty adjustment that FIDE SNPs can receive. Table more current wage data for physicians and hospital 9-10 (p. 272) shows how the relationship between MMP employees. This adjustment increased the FFS rates and MA benchmarks has changed over time. Since component of MMP rates by about 3.8 percent in the start of the demonstration, MMP rates have declined 2013 and 1.8 percent in 2014. Starting in 2015, CMS relative to MA benchmarks as the temporary increases that began making this adjustment when calculating MA CMS made to MMP rates have expired and the reductions benchmarks, so it now applies equally to MMPs and D–SNPs. June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 271

294 TABLE MMP payment rates have declined relative to MA benchmarks 9–10 2013 2014 2015 2016 2017 11 5 9 Number of demonstrations 11 1 396,509 Total enrollment 3,988 180,730 363,491 368,281 104% 103% MMP rates as a share of MA benchmarks 103% 106 % 97% Share of MMP enrollment in counties where rates are: 0% ≤ 90% of MA benchmark 0% 4% 8% 2% 18 1 15 7 6 91% to 95% of MA benchmark 21 96% to 100% of MA benchmark 10 21 16 38 101% to 105% of MA benchmark 52 22 31 16 32 106% to 110% of MA benchmark 4 26 34 26 26 > 110% of MA benchmark 0 5 36 0 25 Note: MMP (Medicare–Medicaid Plan), MA (Medicare Advantage). Figures are based on enrollment in December of the calendar year. Figures do not include effects of the MMP quality withhold, the MA quality bonus program, or the frailty adjustment that fully integrated dual-eligible special needs plans can receive. Components may not sum to totals due to rounding. MedPAC analysis of MMP payment rate data and MA benchmarks. Source: for expected savings under the demonstration have grown equal 95 percent to 98 percent of FFS costs when the larger. For 2017, MMP rates were about 97 percent of MA full reduction for expected savings is made. The MMPs benchmarks. in these areas thus might receive lower payments if they operated as D–SNPs. Although MMP rates are lower than MA benchmarks in the aggregate, the relationship between the two varies The relationship between MMP rates and MA benchmarks from county to county and over time. Since MMP rates can also vary within a state. Illinois provides a good in most states are closely tied to FFS costs, MMP rates example. The state’s demonstration is taking place in are typically lower than MA benchmarks in counties two areas: a 6-county region that includes Chicago and with relatively low FFS spending, such as those where a 15-county region in central Illinois. When we visited benchmarks equal 107.5 percent or 115 percent of FFS Illinois in 2016, the second full year of its demonstration, costs. For example, MMP rates were substantially lower the plan representatives we interviewed said MMPs were than MA benchmarks in Virginia, where the average paid better than D–SNPs in the Chicago region, where the MMP rate in 2017 was about 100 percent of FFS costs, average MA benchmark is about 95 percent of FFS costs, while the average MA benchmark was about 110 percent and worse than D–SNPs in the central Illinois region, of FFS costs. In contrast, MMP rates can be higher than where the average benchmark is about 109 percent of FFS MA benchmarks in counties with relatively high FFS costs. The demonstration has had significant problems in spending, where benchmarks equal 95 percent of FFS the central Illinois region; one of the region’s two MMPs costs. This tendency was especially true in the early years withdrew at the end of 2015, and the remaining plan had of the demonstration, when MMP rates in these areas often to suspend operations in some counties for a few months exceeded 100 percent of FFS costs because the phasing in 2017 because of problems with its provider network. in of the coding intensity adjustment and the additional Table 9-10 shows the distribution of MMP enrollment payments for bad debt more than offset the reductions based on the relationship between MMP rates and MA for expected savings. In later years, MMP rates may still benchmarks. There have been some relatively large Managed care plans for dual-eligible beneficiaries 272

295 changes in the distribution as new demonstrations have D–SNPs in the seven participating counties, and the state started, overall enrollment has grown, and the various took several steps to encourage dual eligibles to enroll in adjustments that CMS has made to MMP rates have taken MMPs instead: effect or expired. For example, the share of enrollees living • Companies that offer both plan types had to in counties where MMP rates are greater than 110 percent transfer any D–SNP enrollees who qualified for the of MA benchmarks jumped sharply in 2016 because of the demonstration into their MMP. These companies one-time increase in MMP rates to account for the effects can continue to offer a D–SNP but can use it of the new risk adjustment model. Despite the year-to-year only for beneficiaries who do not qualify for the volatility, the share of enrollment in counties where MMP 26 demonstration. rates were lower than MA benchmarks grew noticeably between 2014 and 2017, from 29 percent to 64 percent. Companies that offer a D–SNP but not an MMP can • continue offering a D–SNP, and the beneficiaries A full comparison of how health plans are paid when in the plan were exempt from passive enrollment. operating as MMPs or D–SNPs would need to account However, these D–SNPs have not been allowed to for several other factors. For MMP rates, we would need enroll any new beneficiaries who qualify for the to account for plan performance on the quality withhold. demonstration. The only new beneficiaries who can The available data on MMP performance for the quality enroll are dual eligibles who do not qualify for the withhold (which is for 2014 only) suggest that, when fully demonstration. implemented, the withhold will reduce MMP payments by about 1 percent, on average, although the reduction for • The state is not allowing any companies to offer individual plans will vary between 0 percent and 3 percent. new D–SNPs in the counties that are part of the For D–SNP rates, we would need to account for the demonstration (California Department of Health Care competing effects of the quality bonus, which increases Services 2014). overall payments, and the bidding process, which decreases overall payments. However, our most recent During one of our visits to California, several stakeholders analysis of the MA program suggests that the two largely said that many plan sponsors and enrollment brokers offset each other: In 2018, the average benchmark for all have opposed these restrictions. (The brokers receive D–SNPs without the quality bonus was about 103 percent commissions when they help people enroll in MA plans of FFS costs, while the average payment to D–SNPs, after such as D–SNPs, but the demonstration prohibits MMPs accounting for quality bonuses and plan bids, was 102 from using brokers.) Many sponsors have circumvented percent of FFS costs. Taken together, these data points the state’s restrictions by offering what our interviewees suggest that our comparison of MMP rates and D–SNP referred to as “mirror” or “look-alike” plans. These plans benchmarks is a reasonable approximation of how overall are designed to serve dual eligibles and look like D–SNPs, payments for the two types of plans differ. but they are marketed as conventional MA plans and thus are not affected by the state’s limits on D–SNPs. This comparison of MMP rates and MA benchmarks does not account for more intensive coding of beneficiary The look-alike plans resemble D–SNPs because their diagnoses. Both plan types have an incentive to submit benefit structures have many of the same distinctive more diagnoses than many FFS providers because doing features, such as a beneficiary premium for Part D so increases the plans’ total Medicare payments. In MA, coverage, the highest allowable limit on beneficiary out-of- we have estimated that excess coding adds about 2 percent pocket costs for Part A and Part B services, and the highest to overall MA spending (Medicare Payment Advisory allowable deductible for Part D coverage. These features Commission 2018). We have not examined the extent of are not appealing to the broader Medicare population. The excess coding by MMPs. other conventional MA plans in these counties usually have no premium, a lower out-of-pocket limit, and no Part Competition between MMPs, regular D–SNPs, and D deductible—but these features matter relatively little for “look-alike” plans in California dual eligibles because Part D’s low-income subsidy (LIS), which all dual eligibles receive, covers their premium Our first example of the difficulties in having both MMPs (LIS coverage of premiums is subject to a dollar limit, and D–SNPs in the same area comes from California. but the premiums for the look-alike plans are usually very Before the demonstration, there was a large number of June 2018 | Report to the Congress: Medicare and the Health Care Delivery System 273

296 TABLE California’s demonstration has led to a proliferation of 9–11 “look-alike” MA plans that enroll dual-eligible beneficiaries 2013 2014 2015 2016 2017 MMPs Number of plans 10 10 0 8 10 116,190 0 59,757 117,413 113,673 Enrollees Share that are dual eligibles 99% 99% 98% N/A D–SNPs 25 Number of plans 21 20 18 32 72,696 155,725 186,779 104,566 80,724 Enrollees N/A 97% 98% 98% 97% Share that are dual eligibles “Look-alike” MA plans 18 Number of plans 4 7 11 19 95,047 Enrollees 61,752 5,032 82,186 11,640 91% 95% N/A 96% Share that are dual eligibles 97% Other MA plans Number of plans 119 109 121 123 120 Enrollees 1,013,621 1,103,697 1,041,715 905,196 960,069 Share that are dual eligibles 7% 8% 9% 10% N/A MA (Medicare Advantage), MMP (Medicare–Medicaid Plan), N/A (not available), D–SNP (dual-eligible special needs plan). Th