Exploring Episode-Based Approaches for Medicare Performance Measurement, Accountability and Payment Final Report. Next Steps: Future Research and Demonstration Work


There is an absence of solid empirical work related to and few real-world applications of episode-based approaches that provide guidance on how best to construct and apply episodes of care in the context of performance measurement and/or payment policy.  The work done within this project was exploratory in nature and represents only a first step in a much larger process to flesh-out episode of care-based approaches to performance measurement and payment.  Our exploration highlighted a number of issues and gaps in the knowledge base, where additional research studies and/or testing in the form of small pilot studies or demonstrations could further advance Medicare's capabilities to apply episodes of care in various ways to drive improvements in quality and cost-efficiencies.  

Although not an exhaustive list, additional research that could be considered includes:

On a limited basis, explore how to define episodes of care:  HHS could select a limited (e.g., 5-10) number of high volume/high cost clinical conditions, and explore how to define episodes using different build outs per a building block approach (e.g., hospital-based only including hospital and physicians, ambulatory and hospital providers, etc.).  The purpose of this work would be to test the face validity of different episode constructs with physicians and institutional providers.  Providers would be asked to consider the various constructs as they apply to the various functionalities that HHS is exploring, such as aligning measurement activities, profiling physicians, building financial incentives, and bundling payments.  This work could help flag potential problems with various approaches and help engage providers in the development process.  As part of the work to define how to construct episodes, consideration will need to be given to how to distinguish different types of episodes, such as chronic episodes with acute exacerbations, strictly chronic episodes, and strictly acute episodes (among others).  Such distinctions might be important depending on the actual application of the episodes, such as whether the episodes are being used for performance measurement or for payment.

  • Construct a detailed measure mapping across provider types and settings (within conditions) and identify gaps:  Prior to the application of performance measurement within an episode context, it will be important to understand what could be drawn from existing measures portfolios that could be applied within an episode construct for various types of clinical conditions, through a detailed measure mapping exercise.  This work would identify areas that could be measured in the near term using an episode-of-care-based approach, and highlight measurement gaps that could be addressed through HHS' existing five-year initiative to develop performance measures across all settings of care in which Medicare beneficiaries receive services. - Such a mapping would provide policy guidance to the development process, by providing a framework for measure development.  The mapping could also be shared with providers to obtain their feedback on the next generation of measurement that is more holistic in its orientation.
  • Define what constitutes an accountable unit for an episode of care:  There is limited understanding of and consensus around what would define an organization/system of care that could be held accountable for care delivered within an episode of care.  In more highly integrated systems of care in which providers are already linked organizationally and financially, accountabilities may be easier to define; however, these types of systems represent a small fraction of all providers participating in the FFS Medicare program.  In most cases, there are either loose configurations or an absence of explicitly defined relationships between providers.  Work could be done to test the feasibility of forming virtual groupings of providers that could be held accountable for care delivered within an episode.  This type of study could examine, for a small number of conditions across a handful of local health care markets, where Medicare beneficiaries are receiving care (mapping of care within a community) and then could use this information to construct virtual provider groupings.  Examining implicitly defined care arrangements (actual care seeking patterns) in a FFS world may help shed light on whether there is a complete absence of groupings of providers within a community or if patterns that emerge that could be used to form the basis of an accountable unit.  

This work would allow HHS to test the feasibility of being able to expand out beyond the minority of highly integrated delivery system to all types of settings/locations the notion of an accountable group of providers who could be held responsible for performance and/or payment purposes.  Given variation in types of inter-relationships and connections between providers in a local health care market, the proposed work would explore whether virtual groups are a viable concept.  This work could explore with providers differences between patient-driven (empirical analysis of actual care seeking patterns) versus provider-driven (how providers see themselves as related within a community) patterns of care to define the virtual group.  

  • Test impact of varying attribution methods:  For a limited set of conditions and using various building block episode constructions, HHS could construct episodes of care and assign the episodes to various providers under alternative attribution methods.  This work would comprise both empirical analyses of the results of alternative attribution approaches and tests provider acceptance of various assignments with providers.  This testing work should consider a mix of different types of clinical conditions that will involve more/fewer number of providers, to illustrate the complexities and challenges that may arise. -  The work could also explore the feasibility of mapping Medicare physicians to physician groups for the purposes of testing the feasibility of using the group as the unit of accountability.
  • Design and test episode-based performance measurement with providers:  HHS could identify a small number of clinical areas, define what constitutes an episode, identify associated performance measures applicable to the episode, generate episode-based performance scores, and attribute the episode scores to both integrated and virtual teams of providers.  The test output could be used to test provider acceptance of information packaged this way with groups of providers and to solicit their feedback about how to design such reports and how this information would be used by them.  This work could explore the implications of various attribution methods using both narrow and broad episode constructions, per a building block approach.  
  • Test alternative severity adjustment methods applied in the context of episodes of care:  Severity adjustments will be required to level the playing field for both performance measurement (i.e., in assessing clinical outcomes) and in establishing payment that are fair.  While severity adjustment has been used in the context of predicting annual expenditures for Medicare beneficiaries (i.e., the HCCs) and in constructing clinical outcome measures of performance (e.g., risk-adjusted mortality rates), little work has occurred to examine how an episode of care as the unit of analysis might be adjusted to account for differences in the severity of patient episodes that would be assigned to various accountable unit(s).  Our analyses found that applying the HCCs to episodes constructed from commercially available grouper software had little predictive power in estimating the expenditures for an episode (R2 of 0.03).  It will be important to test alternative approaches to severity adjustment of episodes of care at the patient level, assessing the predictive power of the alterative models in the context of payment and performance measurement applications.  This modeling work would need to consider variations by each type of episode, as different factors will be in play for diabetes versus hip fracture as an example.  
  • Conduct a series of qualitative discussions with physicians related to structuring linked pay for performance incentives:  As a means to start linking the behaviors of providers who care for patients, HHS could engage physicians and hospitals in commenting on various scenarios that explore the creation of joint financial incentive payments--such as with the SNFs and rehabilitation facilities that are routinely used for post-acute care.  These qualitative discussions could also explore issues that might be involved in holding physicians responsible for quality metrics across related types of care.
  • Generate simulation models of episode-based payments:  HHS could use existing Medicare claims data to conduct a series of empirical simulations of alternative bundled payment configurations.  The simulation modeling should explore the level of variation across providers in the types of episodes for which they would be responsible and what the implications might be in terms of bearing the financial risk.  This would be especially germane for small providers who would lack the ability to absorb large financial risks, and may highlight a need to create re-insurance or outlier payment mechanisms to mitigate risk.  Simulation modeling could serve to illuminate where risk would occur and help to identify strategies that would mitigate undue risks to providers.  This work could inform the design of additional bundled payment demonstrations.  
  • Test the feasibility of a bundled payment to providers who are not organized under a common financial entity:  Through a series of qualitative discussions, HHS could obtain feedback from providers to understand the practical issues that would arise in an environment where a bundled payment exists for an episode and yet the providers caring for the patient during this episode are not operating under a single financial entity.
  • Expand out current bundling demonstrations, such as the Acute Care Episodes (ACE) demonstration:  HHS could develop additional bundled payment pilot projects, by expanding the clinical areas covered by the ACE demonstration as well as broadening out the set of provider settings within the demonstration.  For example, a demonstration could explore bundling of ambulatory care only episodes or bundling episodes that include ambulatory and one or more other settings for a discrete set of conditions.
  • Conduct further empirical analyses of care delivered to beneficiaries outside their primary state of residence:  Because approximately 12 percent of care for the conditions we examined involved care delivered outside the beneficiary's primary state of residence, more work is required to understand how much of this care represents referral care, a snow-bird migration effect, or typical day-to-day cross boarder care for residents who live close to state boundaries.  This work would focus on highlighting challenges that would be involved in linking together providers into "virtual" units of accountability when they are not necessarily geographically proximate. 
  • Conduct analyses to understand reasons for geographic variation:  The observed variation in the number of episodes and the payments per episode signal a need for additional work to understand the sources of variation.  Variations due to differences in practice patterns would be ripe for applying policies to dampen down on unnecessary variation.  Variations due to differences in population characteristics (e.g., sickness level) will need to be considered in any episode-type application, to prevent unintended consequences such as avoiding caring for sicker patients.
  • Test the development of composite measures of performance within episode types: As various performance measures are compiled across an episode, these measures could be summarized into related groupings or composite measures to strengthen the ability to reliably measure performance and to winnow down the amount of information that is potentially shared with consumers. Various approaches are available to create summary measures, including an opportunities model (i.e., what proportion of the time was recommended care delivered within the population), an appropriate care model (i.e., what proportion of the time was all the recommended care delivered to patients with a particular condition) or something more complex in which there is an explicit weighting of the measures within the composite measure to reflect the differential importance of various care processes. This type of study could also explore what is the public health impact of providing various services and the ways in which physicians should focus their time and resources to achieve maximum benefit. Within an episode construct, HHS could convene physician panels to weight the various measures in terms of their clinical importance and consider how to handle patients with multiple comorbidities related to prioritizing measures.

View full report


"report.pdf" (pdf, 1.02Mb)

Note: Documents in PDF format require the Adobe Acrobat Reader®. If you experience problems with PDF documents, please download the latest version of the Reader®