oversight

Medicare: Voluntary and Mandatory Episode-Based Payment Models and Their Participants

Published by the Government Accountability Office on 2019-01-22.

Below is a raw (and likely hideous) rendition of the original report. (PDF)

                United States Government Accountability Office
                Report to Ranking Member, Committee
                on Finance, U.S. Senate




                MEDICARE
December 2018




                Voluntary and
                Mandatory Episode-
                Based Payment
                Models and Their
                Participants




GAO-19-156
                                                December 2018

                                                MEDICARE
                                                Voluntary and Mandatory Episode-Based Payment
                                                Models and Their Participants
Highlights of GAO-19-156, a report to Ranking
Member, Committee on Finance, U.S. Senate




Why GAO Did This Study                          What GAO Found
In 2017, expenditures for health care           As of February 2018, the Centers for Medicare & Medicaid Services (CMS) had
services provided through traditional           tested, or was in the process of testing, six episode-based payment models as
Medicare totaled approximately $394             alternatives to traditional Medicare. In these models, rather than pay providers
billion and were expected to grow to            largely based on the volume and complexity of each individual service, CMS
$730 billion by 2027. This 85 percent           establishes a target payment amount to cover the costs of all the services
increase is expected to be driven by            Medicare beneficiaries may receive during a defined “episode of care” initiated
multiple factors, including increases in        by a health care event, such as a surgical hospitalization. Providers can earn
the number of services provided per             additional payments if they treat beneficiaries for less than the target amount and
beneficiary. In an effort to slow this
                                                meet certain quality metrics; in some models, providers may be penalized for
growth and improve the quality of care
                                                expenditures that exceed the target amount or if the care provided does not meet
provided to beneficiaries, CMS is
testing alternatives to traditional
                                                quality goals. Provider participation in all but one of the six models being tested
Medicare, such as episode-based                 is entirely voluntary (i.e., eligible providers may choose to participate in the
payment models. Under episode-                  model and generally have an option to leave the model before testing ends), with
based payment models, providers such            participation in the remaining model mandatory for some providers (i.e., eligible
as hospitals or physician group                 providers must participate and cannot leave the model before testing ends).
practices are held accountable for the          According to CMS data and reports that GAO reviewed, providers participating in
cost and quality of the services                the six episode-based payment models typically had more beds or larger
provided to Medicare beneficiaries
                                                practices, had higher episode volume, and were more often located in urban
during a defined episode of care.
                                                areas compared to all providers that participated in traditional Medicare.
GAO was asked to review the episode-            Stakeholders—participants, experts, and provider groups—that GAO interviewed
based payment models developed by               noted that the likelihood for financial gain under voluntary models can influence
CMS. This report (1) describes the              providers’ decisions to participate in the models.
characteristics of the providers that
participated in these models and (2)            Stakeholders also identified relative advantages of voluntary versus mandatory
compares the relative advantages of             episode-based payment models. In general, stakeholders reported that voluntary
voluntary versus mandatory episode-             models largely benefit providers. For example, these models tend to have more
based payment models, as identified             generous terms and providers can choose to participate in only those models
by stakeholders.                                where they are likely to be successful. On the other hand, mandatory models are
                                                more likely to give CMS generalizable evaluation results.
GAO analyzed CMS data on
participants and other providers and            Relative Advantages of Voluntary versus Mandatory Episode-Based Payment Models
reviewed CMS’s evaluations of the               Identified by Stakeholders
models. GAO also interviewed multiple            Voluntary                                                         Mandatory
stakeholders, including officials from
                                                 Participants often have more favorable terms of                   CMS and its evaluation contractors are able to
CMS, its contractors, selected
                                                 participation (e.g., types of model incentives and                evaluate the performance of participants that are
providers and organizations                      degrees of risk required) than in mandatory                       more representative of different types of
representing providers, as well as               models, as CMS has an incentive to make the                       providers and as such, the model’s evaluation
Medicare experts.                                model attractive so providers willingly participate.              results are more generalizable.

The Department of Health and Human               Participants have the ability to self-select models               CMS can test models with greater financial risks
Services provided technical comments             and episodes where they have identified                           and penalties because providers are required to
                                                 opportunities to successfully implement care                      participate.
on a draft of this report. GAO                   redesign and earn performance bonuses.
incorporated these comments as                                                                                     CMS can encourage transition from traditional
appropriate.                                     CMS and its evaluation contractors are able to                    Medicare to value-based care models among
                                                 test novel concepts in care redesign with early                   providers that may be reluctant to make the
                                                 adopters that are interested in performing well                   change on their own.
                                                 under the model. This allows CMS to assess the
View GAO-19-156. For more information,
                                                 feasibility of a model before additional testing.
contact Jessica Farb at (202) 512-7114 or
farbj@gao.gov.                                  Source: GAO analysis of Centers for Medicare & Medicaid Services (CMS) model evaluation reports and interviews with participants,
                                                experts, and provider groups.



                                                                                                                United States Government Accountability Office
Contents


Letter                                                                                     1
               Background                                                                  5
               Episode-Based Payment Model Participants Had More Beds or
                 Larger Practices, on Average, than Providers Nationally, and
                 Voluntary Participation Was Often Driven by Perceived
                 Financial Opportunity                                                   10
               According to Stakeholders, Voluntary Models Largely Benefit
                 Providers, While Mandatory Models Are More Likely to Give
                 CMS Generalizable Evaluation Results                                    17
               Agency Comments                                                           20

Appendix I     Data Sources for Analysis of Episode-Based Payment
               Model Participant Characteristics                                         21



Appendix II    Effect of Mandatory Episode-Based Payment Models
               on Advanced Alternative Payment Model Participation                       22



Appendix III   Characteristics of Episode-Based Payment
               Model Participants                                                        25



Appendix IV    Example of a Model Tested Under Both Mandatory
               and Voluntary Participation Requirements                                  36



Appendix V     GAO Contact and Staff Acknowledgments                                     38


Tables
               Table 1: Information on Center for Medicare and Medicaid
                       Innovation Episode-Based Payment Models Tested as of
                       February 2018                                                       6
               Table 2: Selected Episode-Based Payment Model Participant
                       Characteristics Relative to Characteristics of Medicare
                       Providers Nationally                                              11




               Page i                        GAO-19-156 Medicare Episode-Based Payment Models
Table 3: Relative Advantages of Voluntary versus Mandatory
        Episode-Based Payment Models, as Identified by
        Stakeholders                                                      17
Table 4: Data Sources for Analysis of Episode-Based Payment
        Model Participant Characteristics                                 21
Table 5: Select Characteristics of Bundled Payments for Care
        Improvement (BPCI) Model 1, Model 2, and Model 4
        Participating Hospitals and Hospitals Nationally, 2018            26
Table 6: Select Characteristics of Bundled Payments for Care
        Improvement (BPCI) Model 2 and Participating Hospitals
        and Non-Participating Hospitals Nationally, 2016                  27
Table 7: Select Characteristics of Bundled Payments for Care
        Improvement (BPCI) Model 4 Participating Hospitals and
        Non-Participating Hospitals Nationally, 2015                      28
Table 8: Select Characteristics of Bundled Payments for Care
        Improvement (BPCI) Model 3 Participating Skilled Nursing
        Facilities (SNF) and SNFs Nationally, 2016                        28
Table 9: Select Characteristics of Bundled Payments for Care
        Improvement (BPCI) Model 3 Participating Home Health
        Agencies (HHA) and HHAs Nationally, 2016                          29
Table 10: Select Characteristics of Oncology Care Model (OCM)
        Participating Physician Group Practices and Practices
        Nationally, 2017                                                  30
Table 11: Select Characteristics of Comprehensive Care for Joint
        Replacement (CJR) Participating Hospitals Before
        Participation Change, CJR Participants After Participation
        Change, and Hospitals Nationally, 2018                            32
Table 12: Select Characteristics of Comprehensive Care for Joint
        Replacement (CJR) Hospitals After Participation Change,
        by Participation Status, 2018                                     34




Page ii                       GAO-19-156 Medicare Episode-Based Payment Models
Abbreviations

APM                        Alternative Payment Model
BPCI                       Bundled Payments for Care Improvement
CMS                        Centers for Medicare & Medicaid Services
CJR                        Comprehensive Care for Joint Replacement
HHA                        Home Health Agency
HHS                        Department of Health and
                           Human Services
Innovation Center          Center for Medicare and Medicaid Innovation
MSA                        metropolitan statistical area
OCM                        Oncology Care Model
QPP                        Quality Payment Program
SNF                        Skilled Nursing Facility



This is a work of the U.S. government and is not subject to copyright protection in the
United States. The published product may be reproduced and distributed in its entirety
without further permission from GAO. However, because this work may contain
copyrighted images or other material, permission from the copyright holder may be
necessary if you wish to reproduce this material separately.




Page iii                              GAO-19-156 Medicare Episode-Based Payment Models
                       Letter




441 G St. N.W.
Washington, DC 20548




                       December 21, 2018

                       The Honorable Ron Wyden
                       Ranking Member
                       Committee on Finance
                       United States Senate

                       Dear Senator Wyden:

                       In 2017, expenditures for health care services provided through traditional
                       Medicare totaled approximately $394 billion and were expected to grow to
                       $730 billion by 2027, an increase of approximately 85 percent. 1 The
                       projected increase in Medicare expenditures is expected to be driven by
                       multiple factors, including an increase in the volume—that is, the number
                       of—and complexity of services provided per beneficiary. We have
                       reported that traditional Medicare’s payment structure generally pays
                       providers based on the volume and complexity of the services they
                       provide, rather than on the value of those services (i.e., the quality of the
                       services relative to their cost). 2

                       In an effort to slow the growth of Medicare spending and improve the
                       quality of care provided to beneficiaries, the Centers for Medicare &
                       Medicaid Services (CMS) tests new ways of delivering and paying for
                       health care services. 3 Specifically, CMS’s Center for Medicare and
                       Medicaid Innovation (Innovation Center), established in 2010, tests and
                       evaluates payment and service delivery models—referred to throughout
                       this report as value-based payment models—to reduce spending and




                       1
                         This was the most recent year for which Medicare expenditure data were available at the
                       time of our review. See The 2018 Annual Report of the Boards of Trustees of the Federal
                       Hospital Insurance and Federal Supplementary Medical Insurance Trust Fund
                       (Washington, D.C.: June 5, 2018).
                       2
                        See for example GAO, High Risk Series: An Update, GAO-17-317 (Washington, D.C.:
                       Feb. 15, 2017).
                       3
                        CMS is the agency within the Department of Health and Human Services (HHS) that is
                       responsible for overseeing the Medicare program.




                       Page 1                               GAO-19-156 Medicare Episode-Based Payment Models
improve quality of care for Medicare beneficiaries. 4 If a model improves
quality without increasing Medicare spending, reduces Medicare
spending without reducing quality, or both improves quality and reduces
Medicare spending, CMS may consider expanding the model beyond the
initial duration and set of participants in the model. 5

Of the more than 35 value-based payment models the Innovation Center
had tested or was in the process of testing as of February 2018, 6 were
episode-based payment models—a type of model designed to test a
payment methodology that holds hospitals, physician group practices,
and other types of providers accountable for the cost and quality of care
provided to beneficiaries during an “episode of care” for a specific health
condition. 6 While episodes of care are defined differently across each of
the six models, they generally comprise a health care event, such as a
surgical hospitalization, and the services beneficiaries receive during a
limited time period thereafter. CMS’s payment methodology also varies
across the six models, but generally consists of (1) a set amount for
spending on all the services beneficiaries receive during the defined
episode of care—known as the target price—as well as (2) bonus
payments if spending is below the target price and penalties if spending is
above the target price. The bonuses and penalties are used to create an
incentive for providers to treat patients for less than the target price while
also meeting certain quality metrics.

CMS has designated provider participation in episode-based payment
models as either voluntary (i.e., interested and eligible participants may
choose to participate in the model and generally have an option to leave
the model before testing is completed) or mandatory (i.e., eligible
providers are required to participate in the model and do not have an
4
 The Innovation Center was established by section 1115A of the Social Security Act, as
added by section 3021 of Patient Protection and Affordable Care Act. See Pub. L. No.
111-148, §§ 3021, 10306, 124 Stat. 119, 389, 939 (2010) (codified at 42 U.S.C. § 1315a).
The Innovation Center also tests and evaluates payment and service delivery models to
reduce spending and improve quality of care for Medicaid and Children’s Health Insurance
Program beneficiaries.
5
 See 42 U.S.C. § 1315a(c). Federal law requires the Secretary of HHS and the Chief
Actuary of CMS to make certain determinations before expanding models regarding the
expected effects of such expansion, such as quality of care, net program expenditures,
and coverage for applicable individuals.
6
 For more information on the Innovation Center and other types of models it has tested,
see GAO, CMS Innovation Center: Model Implementation and Center Performance,
GAO-18-302 (Washington, D.C.: March 26, 2018).




Page 2                               GAO-19-156 Medicare Episode-Based Payment Models
option to leave the model before testing is completed).The majority of
models have been voluntary. In 2016, CMS implemented its only
mandatory episode-based payment model to date. In the final rule
establishing the model, CMS indicated that the mandatory participation
was intended to test the payment model across a wide range of hospitals,
including some that would not otherwise participate. However, in 2017,
CMS issued guiding principles for the Innovation Center, which indicated
the agency’s intent to focus on voluntary models. That year CMS issued a
final rule making participation in the one mandatory model voluntary for
certain providers in approximately half of the geographic areas where the
model was being tested. 7 According to CMS, the change to voluntary
participation provides hospitals and CMS with greater flexibility to
voluntarily test alternative approaches for paying for and delivering health
care services. In 2018, the Secretary of Health and Human Services
(HHS) said the Department was exploring new mandatory and voluntary
episode-based payment models.

You asked us to provide information on voluntary and mandatory
episode-based payment models in Medicare, including the characteristics
of participating providers. This report

1. describes the characteristics of the providers that participated in
   Medicare episode-based payment models as of February 2018,
   including the factors that influenced their decision to participate; and
2. compares the relative advantages of voluntary versus mandatory
   episode-based payment models, as identified by stakeholders.

To describe the characteristics of the providers that participated in
Medicare episode-based payment models, we analyzed CMS data and
reviewed model evaluation reports published by the Innovation Center.
Our review focused on the six episode-based payment models
implemented by the Innovation Center as of February 2018—Bundled
Payments for Care Improvement (BPCI) models 1, 2, 3, and 4; the
Comprehensive Care for Joint Replacement (CJR) model; and the
Oncology Care Model (OCM). 8 To identify participant characteristics and

7
 In this rule, CMS also announced it was cancelling three additional mandatory episode-
based payment models that it had established earlier in 2017. Medicare Program, 82 Fed.
Reg. 57,066 (Dec. 1, 2017).
8
 In October 2018, the Innovation Center began testing a new episode-based payment
model—BPCI Advanced.




Page 3                              GAO-19-156 Medicare Episode-Based Payment Models
compare these characteristics to those of all providers nationally that
participated in traditional Medicare and were eligible for the model, we
analyzed the most current CMS data available at the time of our review,
including the following: model participation files, which list those providers
that participated in the BPCI and CJR models; Provider of Service and
Inpatient Prospective Payment Impact files, which contain information on
certain characteristics of Medicare hospitals; and available model
evaluation reports, which contain information on summary characteristics
of participants in certain models. To assess the reliability of CMS data on
BPCI and CJR participants, we reviewed relevant documentation,
interviewed knowledgeable agency officials, compared results to
published reports, and performed electronic data tests. To assess the
reliability of hospital-level CMS data, we reviewed relevant documentation
and performed electronic data tests. On the basis of these steps, we
determined that all the data we used were sufficiently reliable for the
purposes of our report. See appendix I table 4 for a summary of the
specific data sources we analyzed for each model.

To identify factors that influenced participants’ decisions to participate in
voluntary episode-based payment models, we reviewed information from
available model evaluation reports and interviewed stakeholders,
including CMS officials from the Office of the Actuary and the Innovation
Center, officials from five provider organizations representing provider
types that participate in Medicare episode-based payment models, nine
Medicare payment experts, and representatives from five hospitals that
participated in the CJR model. We selected the Medicare experts based
on their published research on episode-based payment models and
Medicare payment policy, as well as through recommendations from
other stakeholders. We selected the representatives from the five
hospitals to obtain additional viewpoints on mandatory and voluntary
models; these hospitals varied in terms of size and whether their
participation in the model was voluntary or mandatory. 9 Because we used
a nonprobability sample in selecting participating hospitals, our interviews
are not generalizable.

To compare the relative advantages of voluntary and mandatory episode-
based payment models, we interviewed the same individuals as
described above as well as the three third-party contractors conducting

9
 CJR participation was mandatory for hospitals in 67 randomly selected metropolitan
statistical areas for the first two performance years and became voluntary for some
hospitals in the third performance year.




Page 4                              GAO-19-156 Medicare Episode-Based Payment Models
                        episode-based payment model evaluations for CMS. We also reviewed
                        publicly available evaluation reports from the Innovation Center to obtain
                        information on the methods used to assess participant performance in
                        each model.

                        We were also asked to examine the effect, if any, of mandatory episode-
                        based payment models on physicians’ ability to participate in the
                        Advanced Alternative Payment Models (APM) track of the Quality
                        Payment Program (QPP)—Medicare’s program for tying certain physician
                        payments to quality of care and the value of that care. See appendix II for
                        information on this topic and our methodology for collecting this
                        information.

                        We conducted this performance audit from November 2017 to December
                        2018 in accordance with generally accepted government auditing
                        standards. Those standards require that we plan and perform the audit to
                        obtain sufficient, appropriate evidence to provide a reasonable basis for
                        our findings and conclusions based on our audit objectives. We believe
                        that the evidence obtained provides a reasonable basis for our findings
                        based on our audit objectives.



Background
Episode-Based Payment   The Innovation Center’s six episode-based payment models test different
Models                  approaches for paying providers for their services during an episode of
                        care. These approaches are designed to create incentives for providers to
                        improve the quality and efficiency of the care they deliver to Medicare
                        beneficiaries. The models differ in various ways—such as the types of
                        services, clinical conditions, and providers targeted by the model as well
                        as the specific payment methodologies. See table 1 for information on the
                        six episode-based payment models initiated by the Innovation Center.




                        Page 5                         GAO-19-156 Medicare Episode-Based Payment Models
Table 1: Information on Center for Medicare and Medicaid Innovation Episode-Based Payment Models Tested as of February
2018

                                                                               Participating       Participation      Time
Model and description                                                          providers           type               period
Bundled Payments for Care Improvement (BPCI) Model 1                           Hospitals           Voluntary          April 2013 –
Retrospective Acute Care Hospital Stay Only—Hospitals received                                                        December 2016
discounted payments for Medicare services provided during an inpatient
hospital stay and were held financially responsible for any Medicare
spending on services provided 30 days after discharge that exceeded
historical trends.
BPCI Model 2, Retrospective Acute & Post- Acute Care Episode—                  Hospitals,          Voluntary          October 2013 –
Hospitals and physician group practices received additional payments or        physician group                        September 2018
made recoupment payments if the total spending for Medicare services           practices
provided during an episode of care—inpatient stay and up to 90 days after
discharge—was over or under a target price. Participants could select
from up to 48 clinical conditions included in the model.
BPCI Model 3, Retrospective Post-Acute Care Only—Post-acute care               Skilled nursing     Voluntary          October 2013 –
providers received payments or made recoupment payments if total               facilities, home                       September 2018
spending for Medicare services were over or under a target price.              health agencies,
Episodes of care began with post-acute care services and included all          inpatient rehab
services up to 90 days after initiation of post-acute care services with a     facilities, long-
participating post-acute care provider. Participants could select from up to   term care
48 clinical conditions included in the model.                                  hospitals,
                                                                               physician group
                                                                               practices
BPCI Model 4, Prospective Acute Care Hospital Stay Only—Tested a               Hospitals           Voluntary          October 2013 –
payment arrangement with a single advance payment for all Medicare                                                    September 2018
services furnished by providers during an inpatient stay in an acute care
hospital and related readmissions for 30 days after hospital discharge.
Physicians and other practitioners submitted “no-pay” claims to Medicare
and were paid by the hospital out of the advance, bundled payment.
Participants could select from up to 48 clinical conditions included in the
model.
Comprehensive Care for Joint Replacement (CJR) —Hospitals                      Hospitals           Mandatory          April 2016 –
retrospectively receive additional payments or make recoupment                                     (years 1 and 2)    December 2020
payments if the total spending for Medicare services provided during an                            Mandatory and
episode of care—inpatient stay and 90 days after discharge for hip or                              Voluntary (years
knee replacement surgery—is over or under a target price and if care                               3, 4, 5)a
meets certain quality performance thresholds.
Oncology Care Model (OCM)—Physician group practices receive a                  Physician group     Voluntary          July 2016 –
monthly payment for each Medicare beneficiary during a 6-month episode         practices,                             June 2021
of care following the administration of chemotherapy. Providers can earn       payers
additional performance-based payments if the total costs for Medicare
services provided during the episode are under a target price. Starting in
2017, practices could receive higher performance-based payments by
taking on financial risk for spending that exceeds the target price.
Source: Centers for Medicare & Medicaid Services. | GAO-19-156




                                                                 Page 6            GAO-19-156 Medicare Episode-Based Payment Models
Notes:
a
 CJR participation was mandatory for hospitals in 67 randomly selected metropolitan statistical areas
for the first two performance years, 2016 and 2017. On December 1, 2017, a final rule was issued
making provider participation voluntary for 33 of the 67 geographic areas and for hospitals designated
as rural or low-volume, effective January 1, 2018. Participation remained mandatory for hospitals that
were not designated as rural or low-volume in the remaining 34 geographic areas.



These models may include both upside risk and downside risk for
participating providers. In models with only upside risk, providers are
rewarded financially for keeping spending below the models’ target price
and providing quality care. If providers fail to lower costs or provide
quality care, they may not earn these rewards. In contrast, in models that
include downside risk, participants may be penalized with reduced
payments or by other means for any expenditures that exceed the
model’s target price or if the care provided does not meet quality goals.
The type and level of risk can also vary over the course of the model.
Some models begin with no risk or only upside risk and add or increase
the extent of downside risk participants face over the course of model
testing.

The example below shows the difference between the traditional
Medicare payment approach and the CJR model, one of the six episode-
based payment models. Under traditional Medicare, CMS pays providers
separately for the various services associated with a patient’s hip or knee
replacement surgery. The providers whose services may be involved
could include:

     •    the hospital that provides services, such as bed and board and
          nursing care, during the patient’s inpatient stay;
     •    the physician who performs the surgery;
     •    the post-acute care provider who treats the patient after discharge
          from the hospital; and
     •    any other providers who treat the patient for complications related
          to the surgery.

In contrast, under the CJR model, CMS establishes a single amount—
known as the target price—to account for any services the beneficiary
receives during the surgical episode, which generally includes the




Page 7                                    GAO-19-156 Medicare Episode-Based Payment Models
                              inpatient stay and 90 days after discharge. 10 The CJR model holds the
                              hospital where the surgery occurred accountable for the cost and quality
                              of services provided during the entire episode. Operationally, providers
                              submit claims to CMS for each service provided to beneficiaries during
                              this episode, and Medicare initially pays for each as a separate claim
                              under traditional Medicare. At the end of the year, CMS compares what
                              Medicare spent paying all the traditional Medicare claims for the CJR
                              episode to the target price—a process known as retrospective
                              reconciliation.

                              The CJR model includes both upside and downside financial risks for the
                              participating hospitals. Therefore, if total episode spending was below the
                              target price and the participating hospital met certain quality
                              requirements, the hospital may be rewarded with an additional Medicare
                              payment. However, if the total episode spending was higher than the
                              target price, the participating hospital may have to repay Medicare for
                              some or all of the amount of episode spending above the target price.
                              CJR’s reconciliation payment methodology encourages participating
                              hospitals to coordinate with other providers involved in delivering care
                              during the episode and reduce the amount Medicare spends on the
                              health care services associated with the episode. 11



Model Evaluation and          The Innovation Center evaluates each model to assess its impact on
Certification for Expansion   quality of care and Medicare spending. 12 Each evaluation is performed by
                              a third-party contractor that compares spending and outcome data for the
                              model’s participating providers and their patients to a matched
                              comparison group of providers and patients with similar characteristics. If

                              10
                                CMS sets prospective CJR episode target prices for each hospital each year based on a
                              mix of the hospital’s historical episode payments and regional average historic episode
                              payments. CMS then applies a payment discount of up to 3 percent of the hospital’s target
                              price based on the hospital’s performance on certain quality measures—such as
                              measures of the rate of knee or hip replacement complications. The better the hospital’s
                              quality score, the lower the discount, thus incentivizing good performance on the
                              measures. Target prices are set differently for other episode-based payment models.
                              11
                                To encourage accountability by other providers involved in the episode, participating
                              hospitals are permitted to share their Medicare payments with partnering providers.
                              12
                                As of October 2018, the Innovation Center had published a final evaluation report for
                              BPCI model 1; five annual evaluation reports for BPCI models 2-4; one evaluation report
                              for CJR, including results from the first fully mandatory year of participation; and a
                              baseline report for OCM with data for participants prior to the start of the program.




                              Page 8                                GAO-19-156 Medicare Episode-Based Payment Models
the Innovation Center determines that there is enough evidence from the
evaluation results to demonstrate that the model reduces Medicare
spending while maintaining or improving quality, the Center may formally
request that CMS’s Office of the Actuary analyze the financial impact of a
potential expansion. 13 As of March 2018, the Innovation Center had not
formally requested that the CMS’s Office of the Actuary review any
episode-based payment models for expansion, though it had made
requests for two other types of value-based payment models.




13
  The Secretary of HHS may, through rulemaking, expand (including implementation on a
nationwide basis) the duration and scope of a model if (1) the CMS Chief Actuary certifies
that expansion would reduce or not result in any increase in net program spending, (2) the
Secretary determines that expansion is expected to reduce spending without reducing the
quality of care or improve the quality of patient care without increasing spending, and (3)
the Secretary determines that expansion would not deny or limit the coverage or provision
of benefits. See 42 U.S.C. § 1315a(c).




Page 9                                GAO-19-156 Medicare Episode-Based Payment Models
Episode-Based
Payment Model
Participants Had
More Beds or Larger
Practices, on
Average, than
Providers Nationally,
and Voluntary
Participation Was
Often Driven by
Perceived Financial
Opportunity
In General, Participants in   CMS evaluation reports and our analysis of CMS data show that for the
Voluntary and Mandatory       five voluntary and one partially mandatory episode-based payment
                              models implemented as of February 2018, participants were typically
Episode-Based Payment
                              larger in size—measured in terms of the mean number of beds in a facility
Models Had More Beds or       or the mean number of clinicians in the provider’s practice—had higher
Larger Practices and          episode volume, and had other characteristics that differed when
Higher Episode Volume         compared with other providers of the same type that participated in
than Providers Nationally     traditional Medicare and were eligible for the model nationally (see table
                              2). 14




                              14
                                CJR is the one model with mandatory participation. Our analysis examined
                              characteristics of CMS participants at the end of the second performance year—
                              December 2017—when participation was mandatory for hospitals in 67 selected
                              metropolitan statistical areas (MSA).




                              Page 10                             GAO-19-156 Medicare Episode-Based Payment Models
Table 2: Selected Episode-Based Payment Model Participant Characteristics Relative to Characteristics of Medicare Providers
Nationally

                                                            Cumulative
                                                             number of                                      Higher            Higher
                                        Provider           participating                  Larger           episode           episode         Urban       Teaching
                                        typea                                                  b
 Model                                                        providers                    size            volumec          spendingd      locatione      status
 Bundled Payments                       hospital                          24
 for Care Improvement                                                                        ●              no data          no data          ●             ●
 (BPCI) Model 1
 BPCI Model 2                           hospital                         423                 ●                  ●               ●             ●             ●
 BPCI Model 3                           skilled                          873
                                        nursing                                              ●                  ●               ●             ●             N/A
                                        facility
                                        home                             116
                                        health                                               ●                  ●               ●             ○             N/A
                                        agency
 BPCI Model 4                           hospital                          23                 ●                  ●               ○             ●             ●
 Oncology Care Model                    physician                        190
 (OCM)                                  group                                                ●                  ●               ●           no data         ●
                                        practice
 Comprehensive Care                     hospital                         800
 for Joint
                   f
                                                                                             ●              no data          no data          ●             ●
 Replacement (CJR)
Legend
● = model participants more often had this characteristic relative to providers nationally
○ = model participants less often had or were not different for this characteristic relative to providers nationally
no data = comparative data were not available for this characteristic
N/A = characteristic is not applicable for model participant provider type
Source: GAO analysis of Centers for Medicare & Medicaid Services (CMS) model evaluation reports and data. | GAO-19-156

                                                                Notes: Characteristics compared for most models reflect participants and data from the most recent
                                                                evaluation report for each model as of October 2018 or from our analysis of CMS data as of
                                                                September 2017.
                                                                a
                                                                 Physician group practices can also participate in BPCI models 2 and 3 and inpatient rehabilitation
                                                                facilities and long-term care hospitals can participate in BPCI model 3. We excluded these provider
                                                                types from our analysis because CMS has not published information on characteristics of these
                                                                providers relative to non-participants.
                                                                b
                                                                 For hospitals and skilled nursing facilities, size is measured as the mean number of beds in the
                                                                facility; for home health agencies, size is measured as the mean number of employed nurses; and for
                                                                physician group practices, size is measured as the mean number of physicians per practice. The
                                                                table compares mean values for the size of model participants to mean values for size across
                                                                Medicare providers eligible to participate in each model nationally.
                                                                c
                                                                 Episode volume refers to the mean number of episodes (for relevant clinical conditions) for which the
                                                                provider billed Medicare during a set period of time before the model was implemented. Table
                                                                compares mean values for episode volume of model participants to mean episode volume across
                                                                Medicare providers eligible to participate in each model nationally.
                                                                d
                                                                 Episode spending refers to the mean amount of Medicare payments the provider received for
                                                                services related to relevant episodes during a set period of time before the model was implemented.
                                                                For BPCI, the table compares mean values for episode spending of participants to mean spending
                                                                across non-participants eligible to participate in each model. For OCM, the table compares mean




                                                                Page 11                                              GAO-19-156 Medicare Episode-Based Payment Models
values for episode spending of model participants to mean spending across Medicare providers
eligible to participate in OCM nationally.
e
 According to the United States Census Bureau, an urban area is defined as having a population of at
least 2,500, representing densely developed territory, and encompassing residential, commercial, and
other nonresidential urban land uses. Table compares mean values for urban location of model
participants to mean values for urban location across Medicare providers eligible to participate in
each model nationally.
f
Our analysis examines characteristics of CJR participants when participation was mandatory for
hospitals in 67 metropolitan statistical areas.



     •    Larger Size: Providers participating in all six models were, on
          average, larger than providers of the same type nationally. For
          example, CMS’s model evaluation reports and our analysis of
          CMS data found that as of 2018, hospital participants in BPCI
          models 1, 2, and 4 and CJR had a mean number of beds that
          were at least 25 percent greater than hospitals nationally.
          Likewise, physician group practices in OCM as of 2017 were
          considerably larger in terms of mean number of physicians per
          practice than practices nationally (42 and 14 physicians,
          respectively), according to CMS’s first OCM evaluation report.
     •    Higher episode volume: Participants in BPCI models 2, 3, 4, and
          OCM—the four models for which we had evaluation report data for
          participants and non-participants—had a higher volume of
          relevant episodes, on average, than did providers of the same
          type nationally during the baseline period. 15 For example, the first
          OCM evaluation report indicated that physician group practices in
          OCM had, on average, nearly four times as many attributed
          cancer episodes per practice from 2014 through June 2015 as did
          practices nationally.
     •    Higher episode spending: For providers in three of the four
          models for which data were available from evaluation reports,
          baseline spending for relevant episodes was, on average, also
          higher relative to providers of the same type that did not

15
  The baseline period refers to a specified period of time before a model is implemented,
during which CMS collects data on various measures of Medicare utilization and spending
among participating providers and non-participants in order to detect volume and
spending changes before and after model implementation. Baseline episode volume
refers to the mean number of episodes (for relevant clinical conditions) for which the
provider billed Medicare during a set period of time before the model was implemented.
Summary data on baseline episode volume and episode spending were available from
CMS evaluation reports for BPCI models 2-4 and OCM. As of October 2018, baseline
episode volume and spending data were available for CJR participants, but not for all
hospitals nationally.




Page 12                                  GAO-19-156 Medicare Episode-Based Payment Models
          participate in the model. 16 For example, the most recent BPCI
          model 2 evaluation report noted that during the 2011 baseline
          period and across all clinical episodes, participating hospitals had
          an average of $1,159—or 6 percent—more Medicare Part A
          spending per episode than did non-participating hospitals.
     •    Urban location: Five of the six models implemented as of
          February 2018 had participants disproportionately located in urban
          areas, compared with providers of the same type nationally. 17 For
          example, our analysis of CMS data showed that more than 90
          percent of hospitals in BPCI models 1, 2, and 4 were located in
          urban areas, compared to 75 percent of hospitals nationally.
     •    Teaching status: Model participants were more often teaching
          hospitals or affiliated with teaching hospitals than providers
          nationally. Specifically, 57 percent of BPCI model 1 hospitals, 37
          percent of BPCI model 2 hospitals, and 36 percent of BPCI model
          4 hospitals had accredited teaching programs compared with 23
          percent of hospitals nationally, according to our analysis of CMS
          data. In addition, CMS’s first OCM evaluation report indicated that
          16 percent of physician group practices participating in OCM were
          affiliated with academic medical centers, compared with 6 percent
          of practices nationally. (See appendix III for additional information
          on characteristics of model participants and appendix IV for
          additional information on characteristics of CJR participants
          specifically.)




16
  Baseline episode spending refers to the mean amount of Medicare payments the
provider received for services related to relevant episodes during a set period of time
before the model was implemented. Episode spending for BPCI models 2 and 4 is defined
as mean standardized Medicare Part A allowed payments in 2011 for an inpatient stay
plus relevant services received during a 90-day post-discharge period. Episode spending
for BCPI model 3 is defined as the mean standardized Medicare Part A payments in 2011
for the 90 days after admission to a post-acute care facility. For OCM, episode spending is
defined as the mean total standardized payments from 2014 through June 2015 for Part
A, B, and D costs for patients who received a chemotherapy service.
17
  According to the United States Census Bureau, an urban area is defined as having a
population of at least 2,500, representing densely developed territory, and encompassing
residential, commercial, and other nonresidential urban land uses.




Page 13                              GAO-19-156 Medicare Episode-Based Payment Models
Providers Participated in        According to stakeholders we interviewed—providers, provider groups,
Voluntary Models Primarily       and Medicare experts—and providers that CMS’s evaluation contractor
                                 interviewed for BPCI evaluation reports, a primary factor providers
Based on Perceived
                                 considered when deciding to participate in a voluntary episode-based
Financial Gain, According        payment model was their potential opportunity for financial gain under the
to Stakeholders                  model. 18 In addition, stakeholders noted that gaining access to CMS data
                                 and the opportunity to gain experience with episode-based payment
                                 models were other reasons some providers chose to participate in the
                                 models.

Opportunity for financial gain   Stakeholders explained that providers decided to participate in voluntary
                                 episode-based payment models primarily after examining the financial
                                 opportunities the models presented relative to the extent of financial risk
                                 the models required them to assume. For example, according to CMS’s
                                 BPCI evaluation reports, the most common reason providers said they
                                 chose to participate in BPCI model 2 was because the model included
                                 services delivered after a hospitalization, such as post-acute care
                                 services, and the providers thought they would most likely be able to
                                 deliver these services for less than the target price and thereby receive
                                 additional payment from CMS. 19 In addition, several providers and other
                                 stakeholders said that in general providers decide to participate in models
                                 when they determine they have or can acquire sufficient technical and
                                 staffing resources, infrastructure, and relationships with other providers in
                                 the care continuum that they need to cut spending below target prices
                                 and meet quality metrics.

                                 Some providers discontinued their participation in a model after
                                 reassessing their potential for financial gain. In many cases, this occurs
                                 when participating providers enter the phase of the model in which they
                                 face downside risk—financial penalties if they do not meet certain quality




                                 18
                                   For models with mandatory participation, providers did not have to decide whether or not
                                 they would participate, and the types of providers who did participate were a function of
                                 CMS’s criteria for selecting participants.
                                 19
                                   See The Lewin Group, CMS Bundled Payments for Care Improvement Initiative Models
                                 2-4: Year 3 Evaluation and Monitoring Annual Report, a report prepared for the Centers
                                 for Medicare & Medicaid Services (Falls Church, Va., October 2017).




                                 Page 14                              GAO-19-156 Medicare Episode-Based Payment Models
                     metrics or spending targets. 20 For example, a group representing
                     oncology providers told us that many physician practices were willing to
                     participate in OCM when they only faced the possibility of losing bonus
                     payments (i.e. “upside risk”), but they said providers might not continue to
                     participate when the model required them to bear financial responsibility
                     for episodes and face the prospect of financial penalties (i.e., “downside
                     risk”). According to several providers we interviewed and those
                     interviewed for BPCI evaluations, providers also reassessed their
                     participation based on their current performance in the model (i.e.,
                     whether they were able to keep the amount they bill Medicare for their
                     services below the model’s target price and avoid penalties) and on how
                     much it cost them to implement the model (e.g., investments in staffing or
                     technology). These perspectives are consistent with our analysis of the
                     BPCI participation data, which shows that approximately 39 percent of
                     hospitals that participated in BPCI model 2 during the preparation phase
                     (when no downside risk was required), went on to continue participating
                     after the model transitioned to the phase in which providers faced
                     downside risk. 21 Furthermore, providers continued to drop out during the
                     downside risk phase, with more than 100 hospitals having withdrawn from
                     the model as of March 2018, our analysis shows. 22

Access to CMS data   Officials from one provider group told us that many physician group
                     practices that signed up for OCM did so to gain access to robust claims
                     data from CMS. During OCM implementation, CMS is providing

                     20
                        Several of the episode-based payment models phased-in financial penalties in later
                     stages of the model, with limited downside financial risk required for participation in the
                     early years. For example, OCM participants could enter the model in 2016 under an
                     upside only risk arrangement and select, starting in 2017, to transition into the downside
                     risk arrangement. OCM participants are required to transition to downside risk or end their
                     participation in the model if they have not qualified for a performance-based incentive
                     payment by the time of the initial reconciliation of the fourth six-month performance period
                     (estimated summer 2019).
                     21
                       Models 1, 3, and 4 also had significant drop-out of participants during the downside risk
                     phase.
                     22
                       Other research has also examined the characteristics of hospitals that dropped out of
                     BPCI model 2. See The Lewin Group, CMS Bundled Payments for Care Improvement
                     Initiative Models 2-4: Year 5 Evaluation and Monitoring Annual Report, a report prepared
                     for the Centers for Medicare & Medicaid Services (Falls Church, Va., October 2018); K.E.
                     Joynt Maddox, E. J. Orav, J. Zheng, and A. M. Epstein, “Participation and Dropout in the
                     Bundled Payments for Care Improvement Initiative.” Journal of the American Medical
                     Association, vol. 319, no. 2 (2018); and D. Kahvecioglu, C. Ogbue, and R. Talati, “Bundled
                     Payments Initiative Participation and Retention.” Journal of the American Medical
                     Association, vol. 319, no. 20 (2018).




                     Page 15                               GAO-19-156 Medicare Episode-Based Payment Models
                            participating providers with claims information on their beneficiaries, as
                            well as feedback on their performance. The officials explained that these
                            data were often the most comprehensive and real-time data oncology
                            practices had ever received, because the data included claims for all
                            services received by their patients in all care settings (e.g., end of life
                            care and emergency room visits), not just those for cancer-related visits.
                            According to CMS, access to comprehensive claims data allows
                            participating providers to understand what services their patients utilize
                            outside of their oncology practices, which helps the providers better
                            manage care for their patients.

                            Providers and other stakeholders indicated that having access to CMS
                            claims data also informed participants’ understanding of how successful
                            they could be in a model. For example, BPCI providers interviewed by
                            evaluation contractors noted that they used the data provided by CMS
                            when determining which BPCI episodes to join. 23 Specifically, BPCI
                            providers used the CMS data to compare their historical data to national
                            or regional benchmarks to identify the services for which they had lower
                            spending or were more efficient compared to other providers in their
                            markets and thus had the greatest financial opportunity.

Experience in value-based   Several stakeholders also reported that providers chose to participate in
payment models              episode-based payment models to gain experience with value-based
                            payments. They said providers anticipate that these types of payments
                            are likely to become the standard form of Medicare payment in future
                            years. For example, approximately half of the BPCI model 2 hospitals that
                            contractors interviewed for the model’s second annual evaluation said
                            that they participated in a BPCI model because it provided an opportunity
                            to learn about bundled payments and to experiment with new payment
                            models.




                            23
                               BPCI models 2, 3, and 4 permitted potential participants to join episodes for up to 48
                            clinical conditions, such as stroke or hip and knee replacement. For providers that
                            participated in the preparatory phase of these models, CMS provided historical Medicare
                            claims data on services provided by various providers in the participant’s geographic
                            region.




                            Page 16                              GAO-19-156 Medicare Episode-Based Payment Models
                                                                Model evaluation reports published by CMS and our interviews with
According to                                                    stakeholders—CMS officials, evaluation contractors, experts, and
Stakeholders,                                                   provider organizations—identified different benefits when testing
                                                                voluntary and mandatory episode-based payment models. In general,
Voluntary Models                                                stakeholders reported that voluntary models largely benefit providers by
Largely Benefit                                                 offering more favorable terms to encourage participation, while mandatory
                                                                models are more likely to give CMS and its evaluation contractors
Providers, While                                                generalizable evaluation results (see table 3).
Mandatory Models
Are More Likely to
Give CMS
Generalizable
Evaluation Results

Table 3: Relative Advantages of Voluntary versus Mandatory Episode-Based Payment Models, as Identified by Stakeholders


  Voluntary Participation                                                                        Mandatory Participation
  Participants often have more favorable terms of participation                                  CMS and its evaluation contractors are able to evaluate the
  (e.g., types of model incentives and degrees of risk required),                                performance of participants that are more representative of different
  as CMS has an incentive to make the model attractive enough                                    types of providers, and as such, the model’s evaluation results are
  for providers to willingly participate.                                                        more generalizable.
  Participants have the ability to self-select models and episodes                               CMS can test models with greater financial risks and penalties
  where they have identified opportunities to successfully                                       because providers are required to participate.
  implement care redesign and earn performance bonuses.                                          CMS can encourage transition from traditional Medicare to value-
  CMS and its evaluation contractors are able to test novel                                      based care models among providers that may be reluctant to make
  concepts in care redesign with early adopters that are                                         the change on their own.
  interested in performing well under the model. This allows
  CMS to assess the feasibility of a model before more extensive
  testing and evaluation.
Source: GAO analysis of Centers for Medicare & Medicaid Services (CMS) model evaluation reports and interviews with stakeholders. | GAO-19-156

                                                                Notes: GAO interviewed officials from the Innovation Center, CMS’s Office of the Actuary, the three
                                                                third-party contractors conducting episode-based payment model evaluations for CMS, nine Medicare
                                                                payment experts, representatives from five provider organizations, and five participants in the
                                                                Comprehensive Care for Joint Replacement model.



                                                                Stakeholders we interviewed explained that the advantages associated
                                                                with voluntary episode-based payment models largely benefit providers.
                                                                For example, several model participants and experts told us that when
                                                                participation is voluntary, providers can choose to join only those models,
                                                                and select specific episodes of care within those models, for which they



                                                                Page 17                                               GAO-19-156 Medicare Episode-Based Payment Models
have a high likelihood of keeping the amount they bill Medicare below the
model’s episode target price. Stakeholders also said participants in
voluntary models can choose to end their participation in the model or in
specific episodes of care if they are not performing well, thereby avoiding
any financial penalties levied under the model. In comparison,
participants in mandatory models are limited in their ability to end
participation if they cannot meet model requirements, which, some
stakeholders suggested, can adversely affect patient care and the
financial viability of some providers. See appendix IV for more information
on the advantages of testing the CJR model with both voluntary and
mandatory participation.

Stakeholders also identified the relative advantages associated with
mandatory episode-based payment models as advantages largely for
CMS and its evaluation contractors. Specifically, CMS officials and other
stakeholders told us that a principal advantage of mandatory models is
that they generally have larger and more diverse participant populations
and likely include providers that may not have otherwise participated if the
model were voluntary. 24 According to CMS officials, an evaluation
contractor, and several experts we interviewed, larger, more diverse
patient populations give CMS and third-party contractors access to more
generalizable data to evaluate the effects of the models. This in turn
allows CMS to determine whether a model is likely to reduce costs and
improve care quality among all types of providers, not just participants
that elected to join voluntary models. In addition, stakeholders told us that
mandatory participation helps ensure that model participants remain in a
model, making it easier to evaluate the effects of the model on cost and
quality.

In contrast, voluntary models suffer from participant drop outs, as shown
in our analyses and the evaluations of BPCI models—for example 13 of
24 participants dropped out of BPCI model 1 by the second performance
year. Evaluation contractors told us that in order to mitigate the influence
of these dropouts, the evaluators may have to adjust the methodology
they use to compare participants to non-participants, and at times, they
must aggregate episodes to increase sample sizes, both of which affect
the ability to generalize the evaluation’s findings.

24
 Research has also examined data obtained from episode-based payments model with
mandatory and voluntary participation. See, A. S. Navathe, et.al. “Comparison of Hospitals
Participating In Medicare’s Voluntary and Mandatory Orthopedic Bundle Programs.”
Health Affairs, vol.37, no.6 (2018).




Page 18                              GAO-19-156 Medicare Episode-Based Payment Models
Because voluntary and mandatory episode-based payment models each
have their own distinct advantages, their utility may depend on CMS’s
needs. In general, voluntary models attract smaller groups of motivated
providers—which may be ideal for CMS when testing a more novel
concept in care redesign before attempting more extensive testing. In the
case of mandatory models, these may be particularly useful for testing
models of care delivery and payment that have already shown some
potential for reducing costs and improving quality—that is, models that
CMS is considering for broader implementation.

Stakeholders also identified other factors CMS should consider when
designing and evaluating both voluntary and mandatory models.
Specifically, several experts and provider organizations emphasized the
importance of effectively risk-adjusting the target prices providers receive
under the models to account for the varying costs of providing health care
services to patient populations with relatively sicker patients. 25 Otherwise,
stakeholders explained, model participants may have an incentive to treat
only relatively healthy patients in order to meet the target price and
receive any bonus payments for doing so.

Stakeholders also noted that evaluations of voluntary and mandatory
episode-based payment models should account for changes in the
characteristics of patients treated by providers and in the volume of
services delivered by providers. These changes, according to
stakeholders, could indicate that participants are reducing episode
spending through means other than efficiency improvements—such as
treating only healthier patients that they know will require fewer services
during the episode—or increasing the number of procedures performed.
According to stakeholders, these responses have implications for the
ability of the model to reduce Medicare spending and improve care if the
model were more widely implemented. While model evaluations to date
include some steps to examine changes in patient mix, evaluation




25
  Risk adjustment is the process by which Medicare adjusts its payments to account for
differences in patient characteristics that can lead to differences in spending amounts. The
method for adjusting for patient differences may depend on the model and its participants.




Page 19                               GAO-19-156 Medicare Episode-Based Payment Models
                  contractors told us that they intend to take additional steps to examine
                  changes to patient mix and volume in future evaluations. 26


                  We provided a draft of this report to HHS for comment. The Department
Agency Comments   provided technical comments, which we incorporated as appropriate.


                  As agreed with your office, unless you publicly announce the contents of
                  this report earlier, we plan no further distribution until 30 days from the
                  report date. At that time, we will send copies to the appropriate
                  congressional committees, relevant agencies, and other interested
                  parties.

                  In addition, the report will be available at no charge on the GAO website
                  at http://www.gao.gov.

                  If you or your staff have any questions about this report, please contact
                  Jessica Farb at (202) 512-7114 or farbj@gao.gov. Contact points for our
                  Office of Congressional Relations and Office of Public Affairs can be
                  found on the last page of this report. Other major contributors to this
                  report are listed in appendix V.

                  Sincerely yours,




                  Jessica Farb
                  Director, Health Care



                  26
                     Research has also examined changes in patient volume before and after BPCI testing
                  as well as changes in hospital-level case mix based on demographic, socioeconomic,
                  clinical, and utilization factors. See A.S. Navathe, et.al., “Association of Hospital
                  Participation in a Medicare Bundled Payment Program With Volume and Case Mix of
                  Lower Extremity Joint Replacement Episodes.” JAMA, vol. 320, no. 9 (2018).




                  Page 20                             GAO-19-156 Medicare Episode-Based Payment Models
Appendix I: Data Sources for Analysis of
                                        Appendix I: Data Sources for Analysis of
                                        Episode-Based Payment Model Participant
                                        Characteristics


Episode-Based Payment Model Participant
Characteristics
                                        Table 4 summarizes the data sources we analyzed to determine
                                        characteristics of episode-based payment model participants and
                                        providers nationally that participated in traditional Medicare and were
                                        eligible for the model.

Table 4: Data Sources for Analysis of Episode-Based Payment Model Participant Characteristics

                                                                   Sources (time periods reflected)
Model                              Number of model participants               Participant and non-participant characteristics
Bundled Payments for Care          BPCI participant files (April 2013 –       Hospital-level Provider of Service and Inpatient
Improvement (BPCI) Model 1         December 2016); BPCI Model 1               Prospective Payment Impact files (as of September
                                   Year 2 Evaluation Report                   2017)
BPCI Model 2                       BPCI participant files (October 2013 –     Hospital-level Provider of Service and Inpatient
                                   March 2018); BPCI Models 2-4               Prospective Payment Impact files (as of September
                                   Year 5 Evaluation Report                   2017); BPCI Models 2-4 Year 5 Evaluation Report
                                                                              (2011 and 2013 for providers as of December 2016)
BPCI Model 3                       BPCI participant files (October 2013 –     BPCI Models 2-4 Year 5 Evaluation Report (2011
                                   March 2018); BPCI Models 2-4               and 2013 for providers as of December 2016)
                                   Year 5 Evaluation Report
BPCI Model 4                       BPCI participant files (October 2013 –     Hospital-level Provider of Service and Inpatient
                                   March 2018); BPCI Models 2-4               Prospective Payment Impact files (as of September
                                   Year 5 Evaluation Report                   2017); BPCI Models 2-4 Year 3 Evaluation Report
                                                                              (2011 and 2013 for providers as of September
                                                                              2015)
Comprehensive Care for Joint       CJR participant files (February 2018)      Hospital-level Provider of Service and Inpatient
Replacement (CJR)                                                             Prospective Payment Impact files (as of September
                                                                              2017)
Oncology Care Model (OCM)          OCM Evaluation Report (January 2017)       OCM Evaluation Report (2014-June 2015)
Source: GAO | GAO-19-156




                                        Page 21                               GAO-19-156 Medicare Episode-Based Payment Models
Appendix II: Effect of Mandatory Episode-
              Appendix II: Effect of Mandatory Episode-
              Based Payment Models on Advanced
              Alternative Payment Model Participation


Based Payment Models on Advanced
Alternative Payment Model Participation
              We were asked to examine the effect, if any, that participating in
              mandatory episode-based payment models, such as the Comprehensive
              Care for Joint Replacement (CJR) model, has on physicians’ ability to
              participate in the Advanced Alternative Payment Models (APM) track of
              the Quality Payment Program (QPP). QPP is a Medicare quality payment
              incentive program created to implement the Medicare Access and CHIP
              Reauthorization Act of 2015. 1 QPP is mandatory for certain physicians
              participating in Medicare and includes two tracks or options for
              participation, one of which is participating through the Advanced APM
              track. 2 Physicians that meet certain thresholds of participation, measured
              by payment or patients in an Advanced APM, are eligible for an APM
              incentive payment. 3 Advanced APMs are payment models that require
              the participating provider to take on financial risk and meet other
              requirements established by law. As of October 2018, the Center for
              Medicare and Medicaid Innovation (Innovation Center) was testing three
              episode-based payment models that qualify as Advanced APMs: the CJR
              model; the Oncology Care Model (OCM); and the Bundled Payments for
              Care Improvement (BPCI) Advanced model. 4 Of these, CJR model
              participation is mandatory for some hospitals and, in turn, the physicians
              performing total knee replacement or total hip replacement episodes at
              those hospitals. Model participation in both OCM and BPCI Advanced are
              voluntary.


              1
               CMS implemented QPP in response to the Medicare Access and CHIP Reauthorization
              Act of 2015, which changed the way Medicare Part B incorporates quality measurement
              into payments for certain physicians and clinicians, whom we refer to collectively as
              physicians for purposes of this report. Pub. L. No. 114-10, 129 Stat. 87 (2015).
              2
               The other QPP track is the Merit-based Incentive Payment System. Under the Merit-
              based Incentive Payment System, physicians are assessed and scored on their
              performance in four categories: quality, cost, improvement activities, and promoting
              interoperability. Depending on their score, physicians may receive a positive, negative, or
              no adjustment to their payments under the traditional Medicare program.
              3
               In 2018, Physicians who received 25 percent of their Medicare payments through an
              Advanced APM and who saw 20 percent of their Medicare patients through an Advanced
              APM qualified for a lump sum incentive payment on top of any bonus payments they may
              have earned through the model. CMS identifies physicians potentially eligible for an
              incentive payment using the Advanced APM’s provider participation or affiliated
              practitioner list. CMS assesses physicians on these lists to determine whether they meet
              the applicable threshold to qualify for an incentive payment. This threshold varies by year.
              See 42 U.S.C. § 1395l(z)(2).
              4
               Specifically, a two-sided risk track of OCM and the certified electronic health record
              technology track of CJR are Advanced APMs for the purposes of QPP. CMS announced
              BPCI Advanced in January 2018 and the first episodes began in October 2018.




              Page 22                               GAO-19-156 Medicare Episode-Based Payment Models
Appendix II: Effect of Mandatory Episode-
Based Payment Models on Advanced
Alternative Payment Model Participation




To identify stakeholder views on the effect of mandatory episode payment
models on providers’ ability to participate in Advanced APMs, we
reviewed relevant laws and regulations and interviewed stakeholders—
CMS officials, select hospitals that participated in CJR, provider
organizations representing provider types that participate in Medicare
episode-based payment models, and Medicare experts. We selected CJR
hospitals that varied in size and based on whether their participation was
voluntary or mandatory. We selected Medicare experts based on related
published research and recommendations from other stakeholders.
Because we used nonprobability sampling, our interviews are not
generalizable.

Provider organizations and experts we interviewed said that the one
mandatory episode-based payment model to date—CJR—provided
opportunities for physicians in the selected metropolitan statistical areas
(MSA) to participate in the Advanced APM track of QPP when they might
not otherwise have had such opportunities. Specifically, physicians not
located in the CJR model MSAs and specialists who treat conditions
beyond primary care have limited opportunities to participate in Advanced
APMs, according to stakeholders we interviewed. For example, one
provider organization noted that, in their experience, many physicians
would like to participate in Advanced APMs, but there aren’t models
available for the patients or conditions they treat. In addition, because
hospitals in mandatory CJR MSAs are required to participate in an
Advanced APM, the opportunities for physicians to partner with those
hospitals to meet volume thresholds increases.

We also heard from some stakeholders that mandatory participation in an
Advanced APM may preclude some physicians from voluntarily
participating in other Advanced APMs in which they could potentially be
more successful. Stakeholders provided the example of a group of
physicians that began participation in total knee or hip replacement
episodes through BPCI models 1, 2, or 4 as part of a participating hospital
that was also located in mandatory areas under the CJR model. The
participating hospitals received an exemption from mandatory
participation in the CJR model until September 2018. 5 A provider
organization that we interviewed told us that the transition into mandatory
CJR model participation for these hospitals would preclude physicians

5
 BPCI model 1 – 4 episodes have precedence for payment and participation over CJR
episodes and CJR episodes have precedence over BPCI Advanced episodes.




Page 23                               GAO-19-156 Medicare Episode-Based Payment Models
Appendix II: Effect of Mandatory Episode-
Based Payment Models on Advanced
Alternative Payment Model Participation




that practiced in these hospitals from leveraging their early BPCI
experience in BPCI Advanced, including those physicians that did well in
the earlier BPCI models. According to the CJR evaluation contractor,
roughly 30 hospitals were affected by this restriction.




Page 24                               GAO-19-156 Medicare Episode-Based Payment Models
Appendix III: Characteristics of Episode-
                          Appendix III: Characteristics of Episode-Based
                          Payment Model Participants



Based Payment Model Participants

                          Tables 5 through 12 provide information on select characteristics of
                          providers participating in the six episode-based payment models
                          implemented by the Center for Medicare and Medicaid Innovation
                          (Innovation Center). The information is based on our analyses of Centers
                          for Medicare & Medicaid Services (CMS) model participation files and
                          hospital-level data, and model evaluation reports published by the
                          Innovation Center.


Bundled Payments for      The Innovation Center tested four Bundled Payments for Care
Care Improvement Models   Improvement (BPCI) models between 2013 and 2018 that tested distinct
                          voluntary episode-based payment approaches for paying various types of
1, 2, 3, and 4
                          providers for various clinical episodes. BPCI models 1, 2, and 4 targeted
                          hospitals. 1 Tables 5 through 7 provide information on hospitals that
                          participated in these models compared with all hospitals nationally that
                          were eligible to participate. BPCI model 3 targeted post-acute care
                          providers. Tables 8 provides information on skilled nursing facilities
                          (SNF), and table 9 provides information on home health agencies (HHA)
                          that participated in model 3 as of 2016 compared with all relevant
                          providers nationally that were eligible to participate. 2




                          1
                           In addition to hospitals, physician group practices could also participate in BPCI model 2;
                          however, we excluded them from our analysis because provider-level characteristics data
                          are unavailable and CMS had published only limited information on participant physician
                          group practices relative to non-participants. Specifically, the BPCI Models 2-4: Year 5
                          Evaluation and Monitoring Report included limited information on the distribution of
                          physician specialties in BPCI-participating practices and on the average number of
                          quarterly hip and knee replacement discharges for orthopedic surgeons at BPCI-
                          participating practices versus all Medicare-billing orthopedic surgeons nationwide.
                          According to the report, the median number of hip or knee replacement discharges for
                          orthopedic surgeons at BPCI-participating physician group practices per quarter ranged
                          between 8 and 10, compared to 4 to 5 discharges per quarter across all orthopedic
                          surgeons.
                          2
                           In addition to SNFs and HHAs, long-term care hospitals, inpatient rehabilitation facilities,
                          and physician group practices could also participate in BPCI model 3; however, we
                          excluded them from our analysis because CMS had not published information on
                          characteristics of these providers relative to providers nationally at the time of our review.




                          Page 25                                GAO-19-156 Medicare Episode-Based Payment Models
                                                               Appendix III: Characteristics of Episode-Based
                                                               Payment Model Participants




Table 5: Select Characteristics of Bundled Payments for Care Improvement (BPCI) Model 1, Model 2, and Model 4
Participating Hospitals and Hospitals Nationally, 2018

                                                                                               BPCI Model
                                                                   Model 1 hospitals         Model 2 hospitals      Model 4 hospitals         All hospitals eligible
Characteristic                                                               (N=21)a                  (N=421)b                (N=22)c                   (N=3,326)d
Number of beds, mean                                                                   339                   375                      515                          240
                                        e
Resident-to-bed ratio, mean                                                           0.15                  0.13                     0.13                         0.07
Disproportionate share percentage, meanf                                              0.24                  0.30                     0.32                         0.30
                                            g
Medicare days percent, mean                                                           0.40                  0.36                     0.29                         0.38
Bed size (#, %)
            Small (<82 beds)                                                       1 (5%)               29 (7%)                    1 (5%)                  828 (25%)
            Medium (82-319 beds)                                                  9 (43%)            196 (47%)                   7 (32%)                 1,653 (50%)
            Large (>319 beds)                                                   11 (52%)             196 (47%)                  14 (64%)                   845 (25%)
                                    e
Teaching hospital (#, %)
            Yes                                                                 12 (57%)             155 (37%)                   8 (36%)                   747 (23%)
            No                                                                    9 (43%)            266 (63%)                  14 (64%)                 2,579 (78%)
Safety-net hospital (#, %)f
            Yes                                                                    0 (0%)             41 (10%)                     2 (9%)                  356 (11%)
            No                                                                21 (100%)              380 (90%)                  20 (91%)                 2,970 (89%)
Urban/ rural location (#, %)
            Urban                                                             21 (100%)              396 (94%)                22 (100%)                  2,501 (75%)
            Rural                                                                  0 (0%)               25 (6%)                    0 (0%)                  825 (25%)
                                h
Ownership type (#, %)
            For profit                                                            3 (14%)             77 (18%)                   5 (23%)                   793 (24%)
            Nonprofit                                                           17 (81%)             274 (65%)                  13 (59%)                 1,651 (50%)
            Government                                                             0 (0%)               22 (5%)                    1 (5%)                  529 (16%)
            Other                                                                  1 (5%)             48 (11%)                   3 (14%)                   353 (11%)
Source: GAO analysis of Centers for Medicare & Medicaid Services (CMS) data. | GAO-19-156

                                                               Notes: Table examines hospitals that participated in at least one downside risk episode of BPCI
                                                               models 1, 2, and 4 for at least one quarter as identified in quarterly BPCI analytic files as of March
                                                               2018 available from www.cms.innovation.gov. Characteristics are derived from CMS’s Provider of
                                                               Service and Inpatient Prospective Payment System Impact files for 2018.
                                                               a
                                                                Characteristic data were available for 21 of the 24 total hospitals that had participated in BPCI model
                                                               1 for at least one quarter between April 2013 and December 2016, when model testing ended.
                                                               b
                                                                Characteristic data were available for 421 hospitals that had, according to CMS data, participated in
                                                               at least one BPCI model 2 downside risk episode for at least one quarter between October 2013 and
                                                               March 2018.
                                                               c
                                                                Characteristic data were available for 22 of the 23 total hospitals that had, according to CMS data,
                                                               participated in at least one BPCI model 4 downside risk episode for at least one quarter between
                                                               October 2013 and March 2018.
                                                               d
                                                                All hospitals eligible to participate include acute care hospitals paid under the Medicare Inpatient
                                                               Prospective Payment System, excluding hospitals located in Maryland. We further limited to hospitals




                                                               Page 26                                     GAO-19-156 Medicare Episode-Based Payment Models
Appendix III: Characteristics of Episode-Based
Payment Model Participants




that were in both the 2018 Provider of Service and 2018 Inpatient Prospective Payment System
Impact files.
e
 We defined teaching hospitals as those with an affiliated allopathic or osteopathic physician
residency program. Resident-to-bed ratio is another measure of teaching hospital status that reflects
the ratio of the number of physician residents to hospital beds; a higher ratio reflects a larger teaching
program.
f
Disproportionate share percentage refers to the Disproportionate Share Hospital Patient Percentage,
which CMS uses to identify hospitals that serve large numbers of low-income patients and determine
eligibility for certain Medicare payment adjustments. We defined safety-net hospitals as those with
disproportionate share percentage in the top quartile for all hospitals in 2018.
g
 Medicare percentage reflects the percentage of the hospital’s total inpatient days due to Medicare
beneficiaries in 2018.
h
 Government ownership includes hospitals owned or operated by the federal government or by a
state or local government. Other includes tribal hospitals and those with unspecified ownership.



Table 6: Select Characteristics of Bundled Payments for Care Improvement (BPCI)
Model 2 Participating Hospitals and Non-Participating Hospitals Nationally, 2016

                                                                                                       Non-participating
    Characteristics                                          Model 2 hospitals                               hospitalsa
    Number of admissions for all                                               3,004                                   1,598
    BPCI episodes, 2011, mean                                                (N=419)                               (N=2,774)
    Baseline spending for 32 of 48                                           $23,067                                  $21,690
    episodes, 2011, meanb                                                  (N varies)                               (N varies)
Source: The Lewin Group, CMS Bundled Payments for Care Improvement Initiative Models 2-4: Year 5 Evaluation and Monitoring
Annual Report, a report prepared for the Centers for Medicare & Medicaid Services (Falls Church, Va., October 2018).| GAO-19-156

Notes: Table reflects 2011 episode data for hospitals that initiated BPCI episodes from October 2013
through December 2016.
a
 Non-participating hospitals are any hospitals not participating in any BPCI initiative between October
2013 and December 2016 that were not from Maryland and that had available 2011 claims and 2013
Provider of Service data available for select characteristics.
b
 Baseline spending is defined as the mean standardized Medicare Part A allowed payments in 2011
for an inpatient stay plus relevant services provided during a 90-day post-discharge period. Figures
are based on 2011 Medicare claims for 32 of the possible 48 BPCI model 2 clinical episodes.
Analysis is limited to average payments for the 32 clinical episodes that had enough sample size to
conduct a difference in difference analysis. Mean baseline payments for model 2 participating
hospitals were higher than non-participants for 30 of the 32 episodes, but the differences varied by
episode. The number of participating and non-participating hospitals varied by episode.




Page 27                                               GAO-19-156 Medicare Episode-Based Payment Models
Appendix III: Characteristics of Episode-Based
Payment Model Participants




Table 7: Select Characteristics of Bundled Payments for Care Improvement (BPCI)
Model 4 Participating Hospitals and Non-Participating Hospitals Nationally, 2015

                                                                                                       Non-participating
    Characteristics                                            Model 4 hospitals                             hospitalsa
    Number of admissions for all BPCI                                            3,460                                 1,598
    episodes, 2011, mean                                                        (N=23)                             (N=2,971)
    Baseline spending for 2 of 48                                             $30,745                                 $30,221
    episodes, 2011, meanb                                                   (N varies)                              (N varies)
Source: The Lewin Group, CMS Bundled Payments for Care Improvement Initiative Models 2-4: Year 3 Evaluation and Monitoring
Annual Report, a report prepared for the Centers for Medicare & Medicaid Services (Falls Church, Va., October 2017).| GAO-19-156

Notes: Table reflects 2011 episode data for hospitals that initiated BPCI episodes from October 2013
through September 2015.
a
 Non-participating hospitals are any hospitals not participating in any BPCI initiative between October
2013 and September 2015 that were not from Maryland and that had available 2011 claims and 2013
Provider of Service data available for select characteristics.
b
 Baseline spending is defined as the mean standardized Medicare Part A allowed payments in 2011
for an inpatient stay plus relevant services provided during a 90-day post-discharge period. Figures
are based on 2011 Medicare claims for 2 of the possible 48 BPCI model 4 clinical episodes. Analysis
is limited to average payments for the 2 episodes that had enough sample size to conduct a
difference in difference analysis. Mean baseline payments for model 4 participating hospitals were
higher than non-participants for one of the model 4 episodes and lower for the other. The number of
participating and non-participating hospitals varied by episode.



Table 8: Select Characteristics of Bundled Payments for Care Improvement (BPCI)
Model 3 Participating Skilled Nursing Facilities (SNF) and SNFs Nationally, 2016

                                                      BPCI Model 3 SNFs                                           All SNFsa
    Characteristics                                             (N=864)                                          (N=14,166)
    Number of beds, mean                                                      122                                           113
    Part of a chain (#, %)                                           216 (52%)                                  3,162 (22%)
    Urban/Rural (#, %)
         Urban                                                       724 (84%)                                10,088 (71%)
         Rural                                                       140 (16%)                                  4,078 (29%)
    Ownership (#, %)
         For profit                                                  740 (86%)                                10,114 (71%)
         Non-Profit                                                  121 (14%)                                  3,432 (24%)
         Government                                                       3 (0%)                                    620 (4%)
    Number of admissions for all                                              136                                             97
    BPCI episodes, 2011, mean
    Episode spending                                  BPCI Model 3 SNFs                     Non-participating SNFsb
                                                              (N varies)                                   (N varies)
    Baseline spending for 11 of 48                                      $26,742                                       $25,706
    episodes, 2011, meanc
Source: The Lewin Group, CMS Bundled Payments for Care Improvement Initiative Models 2-4: Year 5 Evaluation and Monitoring
Annual Report, a report prepared for the Centers for Medicare & Medicaid Services (Falls Church, Va., October 2018).| GAO-19-156




Page 28                                               GAO-19-156 Medicare Episode-Based Payment Models
Appendix III: Characteristics of Episode-Based
Payment Model Participants




Notes: Table reflects 2011 episode data for SNFs that initiated BPCI episodes from October 2013
through December 2016.
a
 All SNFs include SNFs that participated in BPCI episodes from October 2013 through December
2016 and all other SNFs not participating in a BPCI initiative during that time, that were not from
Maryland, and that had available 2011 claims and 2013 Provider of Service data available for select
characteristics. We calculated values for all SNFs by weighting data for non-participating SNFs
reported in the BPCI Models 2-4: Year 5 Evaluation and Monitoring Report.
b
 Non-participating SNFs are any SNFs not participating in any BPCI initiative between October 2013
and December 2016 that were not from Maryland and that had available 2011 claims and 2013
Provider of Service data available for select characteristics.
c
 Baseline spending is defined as the mean standardized Medicare Part A allowed payments in 2011
for an inpatient stay plus relevant services provided during a 90-day post-discharge period. Figures
are based on 2011 Medicare claims for 11 of the possible 48 BPCI model 3 clinical episodes.
Analysis is limited to average payments for the 11 clinical episodes that had enough sample size to
conduct a difference in difference analysis. Mean baseline payments for model 3 participating SNFs
were higher than non-participants for all 11 episodes, but the differences varied by episode. The
number of participating and non-participating SNFs varied by episode.



Table 9: Select Characteristics of Bundled Payments for Care Improvement (BPCI)
Model 3 Participating Home Health Agencies (HHA) and HHAs Nationally, 2016

                                                     BPCI Model 3 HHAs                                             All HHAsa
    Characteristics                                             (N=116)                                            (N=9,885)
    Number of employed nurses in                                                29                                                 9
    HHA, mean
    Part of a chain (#, %)                                            85 (73%)                                  3,195 (32%)
    Urban/Rural (#, %)
         Urban                                                        91 (78%)                                  7,974 (81%)
         Rural                                                        25 (22%)                                   1911 (19%)
    Ownership (#, %)
         For profit                                                   94 (81%)                                  7,552 (76%)
         Non-Profit                                                   22 (19%)                                  1,721 (17%)
         Government                                                       0 (0%)                                    612 (6%)
    Number of admissions for                                                  374                                           104
    BPCI episodes, 2011, mean
    Episode spending                                 BPCI Model 3 HHAs                     Non-participating HHAsb
                                                              (N varies)                                  (N varies)
    Baseline spending for 3 of 48                                         $8,739                                        $8,531
    episodes, 2011, meanc
Source: The Lewin Group, CMS Bundled Payments for Care Improvement Initiative Models 2-4: Year 5 Evaluation and Monitoring
Annual Report, a report prepared for the Centers for Medicare & Medicaid Services (Falls Church, Va., October 2018).| GAO-19-156

Notes: Table reflects 2011 episode data for HHAs that initiated BPCI episodes from October 2013
through December 2016.
a
 All HHAs include HHAs that participated in BPCI episodes from October 2013 through December
2016 and all other HHAs not participating in a BPCI initiative during that time, that were not from
Maryland, and that had available 2011 claims and 2013 Provider of Service data available for select
characteristics. We calculated values for all SNFs by weighting data for non-participating HHAs
reported in the BPCI Models 2-4: Year 5 Evaluation and Monitoring Report.




Page 29                                               GAO-19-156 Medicare Episode-Based Payment Models
                                                                  Appendix III: Characteristics of Episode-Based
                                                                  Payment Model Participants




                                                                  b
                                                                   Non-participating HHAs are any HHAs not participating in any BPCI initiative between October 2013
                                                                  and December 2016 that were not from Maryland and that had available 2011 claims and 2013
                                                                  Provider of Service data available for select characteristics.
                                                                  c
                                                                   Baseline spending is defined as the mean standardized Medicare Part A allowed payments in 2011
                                                                  for an inpatient stay plus relevant services provided during a 90-day post-discharge period. Figures
                                                                  are based on 2011 Medicare claims for 3 of the possible 48 BPCI model 3 clinical episodes. Analysis
                                                                  is limited to average payments for the 3 clinical episodes that had enough sample size to conduct a
                                                                  difference in difference analysis. Mean baseline payments for model 3 participating HHAs were
                                                                  higher than non-participants for 2 of the 3 episodes, but the differences varied by episode. The
                                                                  number of participating and non-participating HHAs varied by episode.




Oncology Care Model                                               In 2016, the Innovation Center began the Oncology Care Model (OCM) to
                                                                  test a voluntary episode-based payment approach for paying physician
                                                                  group practices for chemotherapy episodes. Table 10 provides
                                                                  information on physician group practices that participated in OCM and all
                                                                  practices nationally that were eligible to participate.

Table 10: Select Characteristics of Oncology Care Model (OCM) Participating Physician Group Practices and Practices
Nationally, 2017

                                                                                                                                        OCM practices               All eligible practicesa
 Characteristic                                                                                                                              (N=190)                             (N=2,148)
 Number of physicians per practice (all specialties), mean                                                                                                 42                                  14
 Number of oncology specialists per practice, mean                                                                                                         20                                   6
 Percentage of practices affiliated with an academic medical center                                                                                     16%                                   6%
 Percentage of episodes for beneficiaries eligible for both Medicare and Medicaid, mean                                                                 12%                                   14%
 Practice located in urban market                                                                                                                Nearly all                              No data
 Episode volume per practice, mean                                                                                                                     1,629                                  424
                                                   b
 Total spending per episode, mean                                                                                                                  $27,386                               $25,897
                                           c
 Practice market share, mean                                                                                                                            39%                                   24%
Source: Abt Associates, First Annual Evaluation of the Oncology Care Model: Baseline Period, a report prepared for the Centers for Medicare & Medicaid Services, Feb. 1, 2018. | GAO-19-156

                                                                  Notes: Most characteristics reflect data for January 2014 – June 2015 for physician practices
                                                                  participating in OCM and the national universe of OCM eligible practices as of 2017. The percentage
                                                                  with an academic affiliation reflects data for only 2015. Additional information on OCM participating
                                                                  physician group practices and the universe of practices are available from the source evaluation
                                                                  report.
                                                                  a
                                                                   The national universe includes physician practices that were eligible for OCM based on program
                                                                  participation rules and that were not substantially dissimilar from OCM practices. The source
                                                                  evaluation report identified groups of physicians eligible to participate in OCM based on tax
                                                                  identifiable billing units, which may not map perfectly to an entire physician group practice. We use
                                                                  the term “physician group practice” for ease of reporting.
                                                                  b
                                                                   Total spending per episode includes mean Medicare Part A, Part B, and Part D payments for
                                                                  episodes in the baseline period.
                                                                  c
                                                                   Market share is based on the total proportion of cancer related claims billed by the practice within the
                                                                  geographic market in which it is located. Market share is a measure of competition.




                                                                  Page 30                                                GAO-19-156 Medicare Episode-Based Payment Models
                          Appendix III: Characteristics of Episode-Based
                          Payment Model Participants




Comprehensive Care for    In 2016, the Innovation Center began the Comprehensive Care for Joint
Joint Replacement Model   Replacement (CJR) model to test a retrospective episode-based payment
                          approach for paying hospitals for hip and knee replacement. CJR
                          participation was mandatory for hospitals in 67 metropolitan statistical
                          areas for the first two performance years, 2016 and 2017. Starting in
                          performance year 3—2018—the Innovation Center made participation
                          voluntary for hospitals in 33 of the 67 areas and for hospitals in
                          mandatory areas designated as rural or low-volume. Table 11 provides
                          information on hospitals that participated in CJR at the end of
                          performance year two (when participation was entirely mandatory),
                          hospitals at the beginning of performance year three (when participation
                          became voluntary for some hospitals), and all hospitals nationally that
                          were eligible to participate. Table 12 examines characteristics of the
                          approximately 424 hospitals (of the approximately 800 total participating
                          hospitals) for whom participation became voluntary in year three, by
                          whether they opted to remain in CJR or drop-out of the model.




                          Page 31                               GAO-19-156 Medicare Episode-Based Payment Models
                                                               Appendix III: Characteristics of Episode-Based
                                                               Payment Model Participants




Table 11: Select Characteristics of Comprehensive Care for Joint Replacement (CJR) Model Participating Hospitals Before
Participation Change, CJR Participants After Participation Change, and Hospitals Nationally, 2018

                                                                                                CJR hospitals
                                                                                                             CJR hospitals after
                                                                        Original CJR hospitals,            participation change,
                                                                                  end of year 2              beginning of year 3                     All hospitals
Characteristic                                                                        (N=790)a                          (N=463)b                        (N=3,253)c
Number of beds, mean                                                                              302                          330                              243
                                        d
Resident-to-bed ratio, mean                                                                       0.10                        0.12                             0.06
Disproportionate share percentage, meane                                                          0.33                        0.33                             0.30
                                            f
Medicare days percent, mean                                                                       0.33                        0.34                             0.37
Bed size (#, %)
            Small (<82 beds)                                                                 129 (16%)                   45 (10%)                       783 (24%)
            Medium (82-319 beds)                                                             364 (46%)                 215 (46%)                      1,630 (50%)
            Large (>319 beds)                                                                297 (38%)                 203 (44%)                        840 (26%)
Teaching hospital (#, %)d
            Yes                                                                              228 (29%)                 142 (31%)                        741 (23%)
            No                                                                               562 (71%)                 321 (69%)                      2,512 (77%)
Safety-net hospital (#, %)e
            Yes                                                                              136 (17%)                   75 (16%)                       351 (11%)
            No                                                                               654 (83%)                 388 (84%)                      2,902 (89%)
Urban/ rural location (#, %)
            Urban                                                                           790 (100%)                463 (100%)                      2,453 (75%)
            Rural                                                                               0 (0%)                      0 (0%)                      800 (25%)
                                g
Ownership type (#, %)
            For profit                                                                       173 (22%)                 115 (25%)                        699 (22%)
            Nonprofit                                                                        409 (52%)                 249 (54%)                      1,623 (50%)
            Government                                                                       116 (15%)                   49 (11%)                       522 (16%)
            Other                                                                             92 (12%)                   50 (11%)                        409(13%)
Source: GAO analysis of Centers for Medicare & Medicaid Services (CMS) data. | GAO-19-156

                                                               Notes: CJR participation was mandatory for hospitals in 67 randomly selected metropolitan statistical
                                                               areas for the first two performance years 2016 and 2017. This table presents hospitals that
                                                               participated in CJR at the end of the second performance year, beginning of the third performance
                                                               year, and eligible hospitals nationally in 2018. Characteristics are derived from CMS’s Provider of
                                                               Service and Inpatient Prospective Payment System Impact files for 2018.
                                                               a
                                                               Approximately 800 total hospitals were participating in CJR at the end of year 2. Characteristic data
                                                               were available for 790 hospitals.
                                                               b
                                                               Approximately 465 total hospitals were participating in CJR at the beginning of year 3, according to
                                                               CMS data. Characteristic data were available for 463 hospitals.
                                                               c
                                                                All hospitals eligible to participate include acute care hospitals paid under the Medicare Inpatient
                                                               Prospective Payment System, excluding hospitals located in Maryland. We further limited to hospitals
                                                               that were in both the 2018 Provider of Service and 2018 Inpatient Prospective Payment System
                                                               Impact files.




                                                               Page 32                                     GAO-19-156 Medicare Episode-Based Payment Models
Appendix III: Characteristics of Episode-Based
Payment Model Participants




d
 We defined teaching hospitals as those with an affiliated allopathic or osteopathic physician
residency program. Resident-to-bed ratio is another measure of teaching hospital status that reflects
the ratio of the number of physician residents to hospital beds; a higher ratio reflects a larger teaching
program.
e
 Disproportionate share percentage refers to the Disproportionate Share Hospital Patient Percentage,
which CMS uses to identify hospitals that serve large numbers of low-income patients and determine
eligibility for certain Medicare payment adjustments. We defined safety-net hospitals as those with
disproportionate share percentage in the top quartile for all hospitals in 2018.
f
Medicare percentage reflects the percentage of the hospital’s total inpatient days due to Medicare
beneficiaries in 2018.
g
 Government ownership includes hospitals owned or operated by the federal government or by a
state or local government.. Other includes tribal hospitals and those with unspecified ownership.




Page 33                                     GAO-19-156 Medicare Episode-Based Payment Models
                                                               Appendix III: Characteristics of Episode-Based
                                                               Payment Model Participants




Table 12: Select Characteristics of Comprehensive Care for Joint Replacement (CJR) Model Hospitals After Participation
Change, by Participation Status, 2018

                                                                                                         CJR hospitals that could voluntarily opt-in for year 3
                                                                                                                              (N=411)b
                                                                   Mandatory CJR hospitals at                       Hospitals that                 Hospitals that
                                                                          beginning of year 3                            opted-in                   dropped-out
Characteristic                                                                      (N=377)a                               (N=86)                        (N=325)
Number of beds, mean                                                                              341                          285                             257
                                        c
Resident-to-bed ratio, mean                                                                       0.12                         0.06                            0.09
Disproportionate share percentage, meand                                                          0.34                         0.25                            0.34
                                            e
Medicare days percent, mean                                                                       0.34                         0.34                            0.33
Bed size (#, %)
            Small (<82 beds)                                                                   28 (7%)                    17 (20%)                       84 (26%)
            Medium (82-319 beds)                                                             176 (47%)                    39 (45%)                      149 (46%)
            Large (>319 beds)                                                                173 (46%)                    30 (35%)                       92 (29%)
Teaching hospital (#, %)c
            Yes                                                                              118 (31%)                    24 (28%)                       84 (26%)
            No                                                                               259 (69%)                    62 (72%)                      241 (74%)
Safety-net hospital (#, %)d
            Yes                                                                               73 (19%)                      2 (2%)                       60 (19%)
            No                                                                               304 (81%)                    84 (98%)                      265 (81%)
Urban/ rural location (#, %)
            Urban                                                                           377 (100%)                   86 (100%)                    325 (100%)
            Rural                                                                               0 (0%)                      0 (0%)                          0 (0%)
                                f
Ownership type (#, %)
            For profit                                                                       103 (27%)                    12 (14%)                       58 (18%)
            Nonprofit                                                                        195 (52%)                    54 (63%)                      159 (49%)
            Government                                                                        40 (11%)                     9 (11%)                       67 (21%)
            Other                                                                             39 (10%)                    11 (13%)                       41 (13%)
Source: GAO analysis of Centers for Medicare & Medicaid Services (CMS) data. | GAO-19-156

                                                               Notes: CJR participation was mandatory for hospitals in 67 randomly selected metropolitan statistical
                                                               areas (MSA) for the first two performance years 2016 and 2017. Effective January 1, 2018, a final
                                                               rule was issued making provider participation voluntary at the beginning of the third performance year
                                                               for 33 of the 67 MSAs and for hospitals designated as rural or low-volume. 82 Fed. Reg. 57,066 (Dec.
                                                               1, 2017). Participation remained mandatory for hospitals in the remaining 34 MSAs that were not
                                                               designated as rural or low-volume. This table presents CJR participating hospitals as of February
                                                               2018 after participation changed in the third performance year. Characteristics are derived from
                                                               CMS’s Provider of Service and Inpatient Prospective Payment System Impact files for 2018.
                                                               a
                                                               Mandatory CJR hospitals are those in the 34 MSAs that were not designated as rural or low-volume.
                                                               Characteristic data were available for 377 of 378 mandatory hospitals.
                                                               b
                                                                Voluntary CJR hospitals include hospitals for which participation became voluntary, including
                                                               hospitals in the 33 voluntary MSAs and those in mandatory MSAs designated as rural or low-volume.
                                                               Voluntary hospitals had a one-time opportunity to opt-in to continue participation in CJR for the




                                                               Page 34                                       GAO-19-156 Medicare Episode-Based Payment Models
Appendix III: Characteristics of Episode-Based
Payment Model Participants




remaining performance years. Opt-in hospitals are those that elected to remain in the model. Drop-
out hospitals are those that did not opt to remain in the model. Characteristic data were available for
411 of approximately 424 voluntary hospitals.
c
 We defined teaching hospitals as those with an affiliated allopathic or osteopathic physician
residency program. Resident-to-bed ratio is another measure of teaching hospital status that reflects
the ratio of the number of physician residents to hospital beds; a higher ratio reflects a larger teaching
program.
d
 Disproportionate share percentage refers to the Disproportionate Share Hospital Patient Percentage,
which CMS uses to identify hospitals that serve large numbers of low-income patients and determine
eligibility for certain Medicare payment adjustments. We defined safety-net hospitals as those with
disproportionate share percentage in the top quartile for all hospitals in 2018.
e
 Medicare percentage reflects the percentage of the hospital’s total inpatient days due to Medicare
beneficiaries in 2018.
f
Government ownership includes hospitals owned or operated by the federal government or by a state
or local government. Other includes tribal hospitals and those with unspecified ownership.




Page 35                                     GAO-19-156 Medicare Episode-Based Payment Models
Appendix IV: Example of a Model Tested
              Appendix IV: Example of a Model Tested Under
              Both Mandatory and Voluntary Participation
              Requirements


Under Both Mandatory and Voluntary
Participation Requirements
              This appendix provides additional information on the characteristics of
              Comprehensive Care for Joint Replacement (CJR) model participants and
              stakeholder perspectives. For its first two performance years, CJR model
              participation was mandatory for hospitals in 67 metropolitan statistical
              areas, but participation became voluntary for half of those areas starting
              in the third performance year. Therefore, the CJR model can be used to
              highlight some of the advantages of both approaches to participation.

              According to officials at the Center for Medicare and Medicaid Innovation
              Center (Innovation Center) and the third-party contractor responsible for
              evaluating the model, the CJR model’s mandatory participation provided
              a more diverse and representative group of participant hospitals than
              would have been possible with a voluntary model. The third-party
              contractor noted that this allowed it to observe the results of CJR
              implementation in a wide variety of hospitals and markets. 1 Our analysis
              of CJR participant characteristics also shows variation among participant
              hospitals. For example, 136 of the original CJR participants, or 17
              percent, were safety-net hospitals—hospitals that care for a large share
              of low-income patients—compared to 11 percent of hospitals nationally. It
              is useful for CMS to be able to evaluate how the model affects safety-net
              hospitals, as a November 2016 Mathematica study noted that safety-net
              hospitals faced a number of challenges when transitioning to value-based
              care. Specifically, many patients at safety-net hospitals have complex
              clinical and social needs that require greater coordination with medical
              and other social services, which makes it difficult for safety-net providers
              to perform well on measures of quality of care that are a component of
              episode-based care models. 2 The Innovation Center can get a sense of
              how safety-net hospitals are affected before making the determination of
              whether the model should be recommended for expansion.

              The CJR model participants may still not be representative of all hospitals
              due to selection bias that resulted from the Innovation Center’s choice to
              test the model in larger markets with higher-than-average historical
              spending. We found that the initial participant hospitals during the two

              1
               The Lewin Group, CMS Comprehensive Care for Joint Replacement Model: Performance
              year 1 Evaluation Report, a report prepared for the Centers for Medicare & Medicaid
              Services (Falls Church, Va.: August 2018).
              2
               L. Felland, P. Cunningham, A. Doubeday, C. Warren, Final Report: Effects of the
              Affordable Care Act on Safety Net Hospitals, a report prepared at the request of the
              Department of Health and Human Services (Washington, D.C.: Mathematica Policy
              Research, November 2016).




              Page 36                              GAO-19-156 Medicare Episode-Based Payment Models
Appendix IV: Example of a Model Tested Under
Both Mandatory and Voluntary Participation
Requirements




mandatory performance years were larger and more often located in
urban areas compared to hospitals nationally. As such, evaluation
findings are likely to only be generalizable to other similarly-sized larger
markets.

When CJR became voluntary for all hospitals in 33 of the 67 metropolitan
statistical areas, as well as rural and low-volume hospitals in the other 34
metropolitan statistical areas, there was a corresponding change in the
characteristics of model participants. 3 Of the approximately 424 hospitals
eligible for voluntary participation, 86 voluntarily continued their
participation in the remaining performance years. Our analysis shows that
these hospitals were larger than those hospitals that dropped-out (mean
bed size of 285 vs. 257, respectively) and became less diverse. For
example, of the 62 safety-net hospitals eligible for voluntary participation,
only 2 voluntarily remained in the model. This suggests that smaller
hospitals and safety-net hospitals, when given flexibility to determine
whether they can meet the model’s requirements, will not voluntarily
choose to participate or will end participation if they are able to do so.

Innovation Center and evaluation contractor officials said that they will
need to adapt their methods for evaluating the cost and quality impacts of
CJR in future performance years to account for some providers having
voluntary participation. According to these officials, the evaluation of
voluntary CJR participants will be similar to methods used when
evaluating other voluntary models, such as the Bundled Payments for
Care Improvement models. According to stakeholders, evaluation results
from later years that include both the mandatory and voluntary
participants will be less generalizable than results from the first two years
due to sample changes and selection biases.




3
 Of the 67 MSAs that previously had mandatory participation, CMS selected the 34 MSAs
with the highest average historical spending to continue to have mandatory participation
(except for low-volume and rural hospitals). 82 Fed. Reg. 57,066, 57,073 (Dec. 1, 2017).




Page 37                             GAO-19-156 Medicare Episode-Based Payment Models
Appendix V: GAO Contact and Staff
                  Appendix V: GAO Contact and Staff
                  Acknowledgments



Acknowledgments


                  Jessica Farb, (202) 512-7114 or farbj@gao.gov
GAO Contact
                  In addition to the contact named above, Gregory Giusto, Assistant
Staff             Director; Rebecca Abela, Analyst-in-Charge; Todd Anderson; Krister
Acknowledgments   Friday; Kimberly Lloyd Perrault; and Jennifer Rudisill made contributions
                  to this report.




(102446)
                  Page 38                             GAO-19-156 Medicare Episode-Based Payment Models
                         The Government Accountability Office, the audit, evaluation, and investigative
GAO’s Mission            arm of Congress, exists to support Congress in meeting its constitutional
                         responsibilities and to help improve the performance and accountability of the
                         federal government for the American people. GAO examines the use of public
                         funds; evaluates federal programs and policies; and provides analyses,
                         recommendations, and other assistance to help Congress make informed
                         oversight, policy, and funding decisions. GAO’s commitment to good government
                         is reflected in its core values of accountability, integrity, and reliability.

                         The fastest and easiest way to obtain copies of GAO documents at no cost is
Obtaining Copies of      through GAO’s website (https://www.gao.gov). Each weekday afternoon, GAO
GAO Reports and          posts on its website newly released reports, testimony, and correspondence. To
                         have GAO e-mail you a list of newly posted products, go to https://www.gao.gov
Testimony                and select “E-mail Updates.”

Order by Phone           The price of each GAO publication reflects GAO’s actual cost of production and
                         distribution and depends on the number of pages in the publication and whether
                         the publication is printed in color or black and white. Pricing and ordering
                         information is posted on GAO’s website, https://www.gao.gov/ordering.htm.
                         Place orders by calling (202) 512-6000, toll free (866) 801-7077, or
                         TDD (202) 512-2537.
                         Orders may be paid for using American Express, Discover Card, MasterCard,
                         Visa, check, or money order. Call for additional information.

                         Connect with GAO on Facebook, Flickr, Twitter, and YouTube.
Connect with GAO         Subscribe to our RSS Feeds or E-mail Updates. Listen to our Podcasts.
                         Visit GAO on the web at https://www.gao.gov.

                         Contact:
To Report Fraud,
                         Website: https://www.gao.gov/fraudnet/fraudnet.htm
Waste, and Abuse in
                         Automated answering system: (800) 424-5454 or (202) 512-7700
Federal Programs
                         Orice Williams Brown, Managing Director, WilliamsO@gao.gov, (202) 512-4400,
Congressional            U.S. Government Accountability Office, 441 G Street NW, Room 7125,
Relations                Washington, DC 20548

                         Chuck Young, Managing Director, youngc1@gao.gov, (202) 512-4800
Public Affairs           U.S. Government Accountability Office, 441 G Street NW, Room 7149
                         Washington, DC 20548

                         James-Christian Blockwood, Managing Director, spel@gao.gov, (202) 512-4707
Strategic Planning and   U.S. Government Accountability Office, 441 G Street NW, Room 7814,
External Liaison         Washington, DC 20548




                            Please Print on Recycled Paper.