AHA Responds to CMS' Episode Based Payment Model Request for Information

August 17, 2023

The Honorable Chiquita Brooks-LaSure
Centers for Medicare & Medicaid Services

Submitted Electronically

Re: Request for Information; Episode Based Payment Model

Dear Administrator Brooks-LaSure,

On behalf of our nearly 5,000 member hospitals, health systems and other health care organizations, our clinician partners — including more than 270,000 affiliated physicians, 2 million nurses and other caregivers — and the 43,000 health care leaders who belong to our professional membership groups, the American Hospital Association (AHA) appreciates the opportunity to provide feedback on the Episode Based Payment request for information.

The AHA applauds the Centers for Medicare & Medicaid Services’ (CMS’) continued efforts to reform reimbursement and develop innovative payment models to incentivize efficiency and improved outcomes. Our members support the U.S. health care system moving toward the provision of more accountable, coordinated care and are continuing to redesign delivery systems to increase value and better serve patients. We believe that episode-based payment models could help further these efforts to transform care delivery through financial accountability and improved care coordination.

Over the last 13 years, many of our hospital and health system members have participated in a variety of bundled payment models developed by the Center for Medicare and Medicaid Innovation (CMMI) including Bundled Payments for Care Improvement (BPCI), BPCI Advanced (BPCI-A), and Comprehensive Care for Joint Replacement (CJR). Model design elements like participation criteria, clinical episodes, payment methodologies, metrics and incentives have changed over the course of time. As CMS looks to evolve episode-based payment and develop potential new models, our feedback centers on several core principles:

  • Transparency. Models’ methodology, data and design elements should be transparently shared with all potential participants. Proposed changes should be vetted with stakeholders.
  • Flexible Model Design. Bundled payment model design should be flexible, incorporating features such as voluntary participation, the ability to choose individual clinical episodes, the ability to add components/waivers and options for participants to leave the model(s).
  • Risk Adjustment. Models should include adequate risk adjustment methodologies and account for social needs. This will ensure they do not inappropriately penalize participants treating the sickest, most complicated and underserved patients. The risk adjustment methodologies should be updated as necessary through the model in coordination with participants and the broader stakeholder community.
  • Resources to Support Initial Investment. Upfront investment incentives should be provided to support organizations in their transition to episode-based payment. For example, to be successful in such models, hospitals, health systems and provider groups must invest in additional staffing and infrastructure to support care delivery redesign and outcomes tracking. This includes hiring additional staff (like community health workers, care coordinators and analysts) as well as investing in information technology (IT) (software to track outcomes, electronic health record (EHR) interfaces, etc.).
  • Waivers to Address Barriers to Clinical Integration and Care Coordination. Models should waive the applicable fraud and abuse laws that inhibit care coordination to enable participating hospitals to form the financial relationships necessary to succeed. They should also provide maximum flexibility to identify and place beneficiaries in the clinical setting that best serves their short- and long-term recovery goals. This entails waiving Medicare program regulations that frequently inhibit care coordination and work against participants’ efforts to ensure that care is provided in the right place at the right time
  •  Appropriate On Ramp. Model participants should have an adequate on ramp or glidepath to transition to episode-based payment models. They must have adequate time to implement care delivery changes (integrating new staff, changing clinical workflows, implementing new analytics tools, etc.) and review data prior to initiating the program.
  • Balancing Risk Versus Reward. Models should also balance the risk versus reward in a way that encourages providers to take on additional risk but does not penalize those that need additional time and experience before they are able to do so. A glidepath approach should be implemented, gradually migrating from upside only to downside risk.
  • Establishing Guardrails to Ensure Hospitals Aren’t Competing Against Their Own Best Performance. Models should provide guardrails to ensure that participants do not have to compete against their own best performance and have incentives to remain in models for the long-term.
  • Adequate Model Duration. Models should be long enough in duration to truly support care delivery transformation and assess impact on outcomes. Episode-based models will take time to demonstrate impact on outcomes. Historically, many have been too short and/or have had multiple, significant design changes even within the designated duration, making it difficult for participants to self-evaluate and change course when necessary.
  • Timely Availability of Data. Model participants should have readily available, timely access to data about their patient populations. We would encourage dedication of resources from CMS (staff and technology) to provide program participants with more complete data as close to real-time as possible.

Specific feedback on RFI components follows.


The AHA agrees that incentives must be aligned to support the integration of primary and specialty care, and to support communication, collaboration and coordination across the care continuum. We recognize that bundled payment models must be developed in a manner that encourages care delivery transformation and incentivizes participation from the breadth of providers who touch the patient. As such we recommend the following.

Voluntary Participation. We encourage CMS to ensure that episode-based payment models are voluntary. This means that many organizations may not be of an adequate size or in a financial position to support the investments necessary to transition to mandatory bundled payment models. Requiring them to take on risk for large, diverse bundles of episodes, may require more financial risk than they can bear. This is especially true given the historic financial pressures that hospitals and health systems continue to face. Indeed, according to Kauffman Hall’s June National Hospital Report, while operating margins appear to have stabilized from historic losses, they are still well below historical norms and were under projections.1

Additionally, a Government Accountability Office report found that mandatory participation could negatively impact patient care and financial sustainability if participants are not able to leave the model. It also found that mandatory participation could impact organizations’ ability to support other voluntary models for which they may be better equipped.2

Further, much of the discussion about mandatory participation has been predicated on the high rates of drop out from historical models. However, instead of pursuing mandatory participation, we encourage CMS to address those model design features that led participants to withdraw from historical episode-based payment models in the first place. For example, many decisions to leave were due to index pricing concerns – specifically the ratchet effect where index prices were based on previous years’ performance, thus requiring organizations to compete against their own best performance (see additional information in the payment section).

Aligning Incentives to Increase Integration. Bundled payments can align incentives for providers — hospitals, post-acute care providers, physicians and others — and encourage them to work together to improve the quality and coordination of care. As such, we encourage CMS to provide flexibilities for gainsharing arrangements in future models whereby hospitals could work with physician group practices (PGPs) and post-acute providers to develop mutual accountability and shared risk.

Supporting IT Interoperability. Interoperability between participants and downstream providers is one crucial area that supports care coordination and timely communication. Models should support the entire continuum, from health systems that include downstream episode initiators and practices, to stand-alone hospitals wanting to partner with PGPs to support patients pre- and post-discharge, to other arrangements. Where possible, all organizations caring for the same patient in these models should use the same EHR or have application interfaces that ensure patient information is flowing freely and transparently. This may be challenging in certain scenarios where the participant may need to partner with a provider who lacks the resources to purchase and maintain a sophisticated EHR. As such, CMS should consider expanding Safe Harbor protections (i.e., Stark and Anti-Kickback) for hospitals and health systems to extend access to their EHRs out to others who also fill patient care needs in an episode-based payment model.

Regulatory Relief. AHA continues to call for maximum regulatory relief for all providers participating in alternative payment models. The waiver of certain Medicare program regulations is essential to participants’ ability to coordinate care and ensure that it is provided in the right place at the right time. Participants should have maximum flexibility to identify and place beneficiaries in the clinical setting that best serves their short- and long-term recovery goals. Providing such waivers is also commensurate with the level of risk and accountability that CMS is asking participants to assume as it shifts the burden of risk further away from the Medicare program onto providers. Specifically, we urge CMS to routinely waive:

  • hospital discharge planning requirements that prohibit hospitals from specifying or otherwise limiting the information provided on post-hospital services;
  • the skilled nursing facility three-day rule;
  • the home health homebound rule;
  • the inpatient rehabilitation facility (IRF) ‘60% Rule;’
  • the IRF ‘Three-hour Rule;’ and
  • the long-term care hospital ‘25% Rule.’

In addition, the application of these and similar fee-for-service regulations in an episode-payment scenario clouds the findings on the efficacy of that model, thereby limiting the potential for lessons learned through the model and its usefulness if, in the future, applied to a broader population of patients and providers.


We generally agree with the considerations CMS outlined for developing clinical episodes for future models, including reviewing clinical homogeneity, spending variability, episode volume, quality impact and episode overlap alignment. We encourage close coordination with the provider community in developing future clinical episodes, to garner feedback on clinical appropriateness, duration and services to be included. Below are specific areas of feedback for clinical episodes.

Allow Voluntary Selection of Individual Episodes Instead of Clinical Episode Categories. It is vitally important for participants to have the ability to select individual clinical episodes, as opposed to requiring participants to take on risk for large, diverse bundles of episodes. This becomes even more important if CMS chooses to pursue any mandatory models.

For example, some of the service line groups in BPCI-A included unrelated conditions. The transition to these disparate clinical groups in model year 4 may have contributed to a significant drop of 32% in the number of hospital participants.3 For example, an organization wanting to participate in the sepsis bundle under BPCI-A was required to support all conditions under “Medical and Clinical Care.” This included conditions such as cellulitis, chronic obstructive pulmonary disease, bronchitis, asthma, renal failure, simple pneumonia, and respiratory infections and urinary tract infections, many of which have no clinical connection to sepsis. This requirement created barriers to participation for organizations that were well positioned to support some but not all episodes within a service line.

Focus on MS-DRG Surgical/Medical Episodes. In general, we encourage CMS to focus bundles on acute surgical and medical episodes and exclude chronic conditions. Acute episodes provide greater opportunities in terms of larger volume, less clinical variability and less model overlap. Chronic conditions may introduce more overlap with other programs like accountable care organizations (ACOs) and likely would require longer durations for episode length and greater complexity in terms of confounding variables.

Update Exclusion Criteria. We encourage CMS to exclude unrelated conditions from bundles. These include but are not limited to pre-scheduled inpatient/outpatient services that occur during the episode’s timeline (e.g., glaucoma surgery), unrelated trauma services and critical care transport. For example, for rural and geographically remote areas, critical care transport often requires high-cost air ambulance services, which may inappropriately and adversely impact these organizations’ cost savings achievements if included. Additionally, we urge CMS to explore revised outlier methodologies to account for patients with unforeseen conditions, such as high-cost trauma or emergent services, or complications from unrelated comorbidities (more feedback on this is in the payment methodology section).


We encourage CMS to develop the models in a manner that supports transition to risk, including the specific recommendations outlined below.

Incentivizing Hospital Participation. A study of performance of hospitals and physician group practices (PGPs) participating in bundled payment models found that only hospitals achieved cost savings for both surgical and medical episodes.4 Furthermore, for medical episodes, hospitals demonstrated significant reductions in length of stay at skilled nursing facilities. The study counteracts arguments that PGPs are better suited to provide cost savings in value-based programs. Indeed, hospitals are well positioned to work with post-acute care facilities that may have overlap in patients served and may also be better positioned to assume financial risk. Therefore, it is vitally important that hospitals be fully incentivized to participate in these models.

Providing Transparent, Real-time Data. Historically, the lack of transparent, real-time data has created confusion on trigger events, eligibility for episodes and program participation. We encourage CMS to provide real-time data to conveners and episode initiators.


Hospitals and health systems share CMS’ deep commitment to advancing health equity within their organizations and in the communities they serve. We appreciate CMS seeking ways to promote innovative health equity approaches through its CMMI programs. Hospitals approach the critically important work of health equity and addressing health-related social needs (HRSNs) recognizing that while they may be starting from different points, advancing health equity is not just a one-time activity. Rather, it is a continual process that involves engaging with internal and external stakeholders to build understanding and trust, using data to identify where disparities exist, identifying root causes, deploying interventions to address those causes and measuring progress.

Below we offer our perspective on several health equity and HRSN-related issues raised in the RFI.

Accounting for HRSNs in Payment and Quality. The AHA supports the concept of CMMI models accounting for the impact of HRSNs in both payment and quality measurement methodologies. Indeed, we believe doing so would complement CMS’ commitment to advancing health equity. Recently, CMS has taken important steps forward in recognizing the complex interplay between provider performance and HRSNs by developing new methodologies to incentivize high quality, efficient care for underserved and historically marginalized communities. For example, the Medicare Shared Savings Program, the Hospital Value-Based Purchasing (VBP) program and Skilled-Nursing Facility VBP program now include a health equity adjustment (HEA) that awards providers bonus points based on a combination of their quality performance and the extent to which they treat underserved patients.

We encourage CMS to consider applying approaches like the HEA in its various models. At the same time, CMS should consider the use of a full range of approaches to account for HRSNs in quality measurement and payment — including direct incorporation of HRSN-related variables into risk adjustment models where appropriate. CMMI could, for example, consider using the health equity adjustment’s underserved variables in setting episode-initiator benchmarks, including dual-eligible status (DES), Area Deprivation Index (ADI), and Medicare Part D Low-Income Subsidy. We also encourage CMS to explore the use of the CDC Social Vulnerability Index.

The AHA acknowledges that any potential proxy for social risk or health-related social needs has tradeoffs. For example, DES has the significant benefit of being consistently recorded in Medicare administrative data, and relatively easy to tie back to individual hospitals. There also is a body of research showing the link between DES and other measures of social drivers, such as income. At the same time, DES tends to reflect those patients who face the most significant social needs. Furthermore, Medicaid eligibility criteria can vary across states, which means it may be a more comprehensive reflection of underserved populations for some hospitals than for others. Similarly, the main strength of the ADI is that it is attempts to create a multi-dimensional picture of the social drivers of health in a community. It draws on multiple data sources — including Medicare administrative data and census data — and uses 17 indicators of social risk to develop a single score for a geographic region. At the same time, because ADI is calculated at a census-block level, it has the potential to obscure differences within a particular census block. For example, the ADI for a community could look average, but parts of the community may face enormous structural barriers to accessing health care and other supportive resources that lead to better outcomes.

Ultimately, CMMI policies around how to account for HRSNs should be informed by careful analysis and, where possible, analyses that estimate their impact on model participants. We encourage the agency to conduct such analyses with as much transparency around methods, data sources and results as possible.

Aligning and Sharing Data. To identify and address health inequities, CMMI model participants need accurate and actionable demographic and HRSN-related data to stratify performance measures and track progress. CMS and CMMI can play a vital role in supporting this work by taking steps to align and share health equity and HRSN-related data with model participants. Specifically, we recommend that CMMI do the following.

  • Promote aligned and standardized approaches to collecting, analyzing and exchanging demographic and HRSN data. This includes promoting a consistent approach across CMS itself and other federal agencies and programs. Given the breadth of health equity issues, and the wide range of stakeholders affected by it, CMS can help ensure that all stakeholders use consistent definitions and standards.
  • Share with model participants existing data to which CMS or other governmental agencies may already have access before adding new data collection and reporting requirements. For example, to the extent CMS is collecting demographic and social risk data during the time of enrollment in Medicare, the agency should explore ways of improving its accuracy and determine whether the data could be linked to quality measure data for hospitals and other health care providers. These steps could help provide additional data for CMS’ efforts to identify disparities in performance and outcomes, while reducing the need for additional data collection by hospitals and other providers.


The AHA appreciates CMS’ interest in feedback on quality measurement approaches for CMMI models. Given CMMI’s mandate to lower costs while preserving and improving quality, CMMI’s models need well designed quality measures that are accurate, meaningful, feasible to collect and report, and do not require excessive administrative burden. In response to the issues raised in the RFI, AHA offers several recommendations below.

Measure Alignment Across Models. In general, the AHA supports alignment of measures across CMMI models, especially in instances where models are measuring the same topics. Given that many CMMI model participants engage in more than one model, an aligned approach to measurement helps ensure a consistent approach to incentivizing improved performance across programs while also lowering administrative burden for participants. Measure alignment can sometimes be achieved through adopting the same measure across multiple programs, and we encourage CMS to do so when practical. However, even when it is not possible to use identical measures, CMS should work to ensure directional alignment of definitions and methodologies.

The AHA also cautions CMS against “force fitting” measures into models simply for the sake of achieving alignment. Sometimes, the differences in the programmatic goals and designs of CMMI models means that using precisely the same measures in each program is not always possible or even desirable. The AHA has previously expressed misgivings about the Merit-based Incentive Program’s APM Payment Pathway (APP) measure set and CMS’ policy of requiring the measure set to be used in every model to qualify for the APP. For example, while the APP measure set includes a depression screening measure, it is not always clear how such screening and follow up are relevant to bundled payment models that focus on procedural inpatient care. As a result, the measure may result in administrative burden that outweighs its value in improving care.

Transparency of Measure Specifications. The AHA urges CMS to increase the transparency of the measure specifications used in various CMMI models. Our members have expressed concern that while they know what measures CMS may be using in particular models, they often do not have enough information about what populations are included in the measure, specific details about risk adjustment methodologies and variables, measurement periods and other details that are vital to understanding how CMMI implements particular measures.

Recently, CMS took the step of adding CMMI model measures to its CMS Measure Inventory Tool (CMIT). The AHA appreciates this step and believes there is value in a holistic view of the measures used in CMMI models. However, the information available in the CMIT provides high-level specifications and lacks some of the details described above. We would encourage CMS to build on the information in the CMIT to ensure model participants have the information they need to accurately interpret, implement and track performance on the measures in its models.

Use of Patient-reported Outcome Measures (PROMs). CMS has expressed interest in the broader use of PROMs in CMMI models and in other CMS programs. In concept, the AHA believes PROMs hold promise in providing meaningful insights into patients’ experience of care and outcomes that matter in their daily lives. For example, many PROMs reflect whether patients are regaining day-to-day physical function within a period after a procedure, such as the ability to walk certain distances without discomfort.

However, we urge CMS to adopt a gradual, stepwise approach to implementing PROMs in CMMI models and other programs. This includes making the reporting of PROMs voluntary or rewarding only bonus for successfully undertaking PROM reporting. Experience from the implementation of the total hip arthroplasty/total knee arthroplasty (THA/TKA) PROM in both the CJR model and the Inpatient Quality Reporting (IQR) program has raised significant concerns about the high level of administrative effort and resources needed to collect PROM data. PROM measures usually require at least two patient surveys — one to establish a baseline and a follow up within a set time to measure changes from the baseline. Hospitals have reported that patient follow up is often very challenging, making it difficult to meet CMS’ standards for data completeness. Furthermore, it is not yet clear what a reasonable threshold for data completeness looks like across providers because PROM implementation is still relatively novel on a national scale.

In other words, while PROMs have long-term value in CMS programs, both CMS and health care providers need time to develop the infrastructure to support a sustainable implementation of PROM reporting. Voluntary reporting through CMMI models would provide an important opportunity to gain experience, test measurement approaches and share best practices about successful PROM implementation.

Multi-Payer Alignment. The AHA supports multi-payer alignment to ensure consistency across payers in episode composition, payment policies, outcomes metric methodology, and target thresholds. We encourage CMS to work with Medicaid and commercial payers to establish consistent bundled payment models. Aligning payers will also support further transition to value-based care, by bringing all stakeholders to the table and leveraging episodes as a steppingstone or bridge to population-based models.


Some of the biggest challenges in the BPCI, BPCI-A and CJR programs have been related to payment methodology and structure. Below are recommendations to improve payment methodologies for future models.

Providing Appropriate Risk Adjustment for Clinical Complexity and Health Related Social Needs. An article from JAMA analyzing hospitals leaving the CJR program in 2018 found that the majority of those opting out had higher proportions of non-white and Medicaid patients.5 The article states that these hospitals likely dropped out “since they were more likely to sustain financial losses by remaining in the program” due to higher prevalence of complications and post-acute care needs. This is of deep concern to us. Therefore, we urge CMS to develop models in a manner that incentivizes participation from organizations serving underserved communities. Ensuring adequate risk adjustment methodologies that account for social needs and clinical complexity, as mentioned above, is one policy that would help in this goal.

Leveraging Appropriate Breadth of Historical Claims Data for Setting Target Prices. As part of the extension for the CJR program, CMS updated its methodology for calculating index prices. Specifically, the timeframe for historical claims was adjusted from three years to one year. This continues to be of concern to us. As we previously commented, broader data sets help stabilize target prices and moderate for unforeseen variation, such as in volume or clinical complexity.6 Therefore, we encourage CMS to use longer time periods in calculating target prices, such as three years.

Updating High-cost Spending Caps. High-cost spending caps are necessary to protect hospitals from incurring undue penalties from unexpected and severe complications. For example, the CJR model originally capped individual episode costs at two standard deviations above the mean. However, CMS later changed the cap to the 99th percentile, which was too high and did not capture the prevalence of severe complications. Therefore, we urge CMS to set high-cost spending caps at two standard deviations above the mean.

Create Policies to Help Ensure Organizations Do Not Have To Compete Against Their Own Best Performance. An article from JAMA identified the issue with the spending benchmark methodology in current episode payment models, whereby participants are essentially penalized for efficiency.7 Specifically, benchmarks generally adjust based on historical performance, so if an organization lowers spending, their benchmark will be lowered in future years. As the organization continues to lower spending in an attempt to hit their continually decreasing benchmark, eventually they will not be able to achieve the benchmark. Organizations have cited this as a primary reason for leaving bundled payment programs. CMS has made efforts to enact guardrails to help ensure participants do not have to compete against their own best performance, such as benchmarks that include an adjustment for prior savings and introducing regional trends into benchmarks. “Efficiency floors” are another possibility, which help ensure that organizations are not penalized for reducing spending. In the JAMA article referenced above, an efficiency floor would provide a threshold of spending reductions after which point organizations would not face penalties or could opt out of participation. We strongly encourage CMS to include policies like these in future models, as well as continue researching this issue in general.

Stop-loss Limits for Repayments. We urge CMS to include in models stop-losses that limit participants’ overall repayment responsibility. These should be implemented in a gradual glide-path fashion with additional protections provided for certain participants, such as rural hospitals and lower volume facilities.

Incentivizing Post-acute Participation. Post-acute care (PAC) providers, including long-term care hospitals, inpatient rehabilitation hospitals, skilled nursing facilities and home health agencies, play a key role in ensuring long-term medical and functional outcomes following discharge from an acute-care hospital. Therefore, incentivizing coordination with these providers should be a critical component of models. The intensity of services offered among post-acute care providers can vary greatly, along with the associated cost. However, meaningful differences in outcomes, such as return to daily activities, fall reduction, readmission avoidance and even mortality, and hence savings to Medicare, are not apparent until months or even years later.

Utilizing the most efficient downstream provider is the goal, but “efficiency” includes a measure of outcomes achieved, not only cost. As such, models should incorporate measures of functional outcomes to ensure that they are not incentivizing the use of less costly providers without regard to long-term outcomes. We also urge CMS to ensure it provides waivers to enable hospitals’ flexibility to place beneficiaries in the setting that best serves short- and long-term recovery goals. Specifically, we encourage CMS to waive the post-acute care transfer policy when beneficiaries are discharged from an acute care hospital to post-acute facilities and organizations that commit to coordinating with hospital partners for episode- based payment models.


As CMS explores ways to better integrate specialty care and ACOs in episode-based payment models, we have several recommendations regarding model overlap, detailed below.

Reduce Barriers for Hospitals to Transition to APMs and ACOs. As CMS evaluates ways to integrate ACOs in episode-based payment models and increase specialty integration, CMS should implement policies to remove barriers to transitioning to risk-based models. For example, CMS should eliminate its designation of ACOs as either low- or high-revenue. The agency has used this label as a proxy measure, for example, to determine if an organization is supporting underserved populations and/or if the organization is physician led in order to qualify for Advance Investment Payments (AIPs). Yet, there is no valid reason to conclude that this delineation, which measures an ACO’s amount of “captured” revenue, is an accurate or appropriate predictor of whether it treats an underserved region. In fact, analysis suggests that critical access hospitals, federally qualified health centers and rural health centers are predominantly classified as “high-revenue,” disqualifying them from obtaining AIPs that would be of a huge benefit in their ability to take on higher risk. Further, both low- and high-revenue ACOs are working to address health equity as part of their care transformation work; assistance investing in these efforts would help across the board. Advanced APM Incentive payments should also be extended to support non-fee-for-service programs, including meal delivery programs, transportation services, and digital tools and care coordinators, each of which promote population health.

We appreciate your consideration of these issues. Please contact me if you have questions or feel free to have a member of your team contact Jennifer Holloman, AHA’s senior associate director of policy, at (202) 6262-320 or jholloman@aha.org.



Ashley Thompson
Senior Vice President Policy

Cc: Elizabeth Fowler
      Director, CMMI