Risk adjustment for retrospective episode-based payment

| Article

Many healthcare payors and providers consider “episodes of care” to be an appealing approach to measure provider performance and reward successful outcomes. An episode of care is any clinical situation with relatively measurable start and end points and with clear patient outcomes, such as procedures, hospitalizations, acute outpatient care, and some treatments for cancer and behavioral health conditions. Recently, momentum has been building for “Retrospective Episodebased Payment” (REBP) as one method to reward providers who consistently deliver high quality and/or favorable costs for specific episodes.1,2 REBP identifies one or more providers who are in the best position to affect the clinical outcomes and costs associated with an episode of care (herein referred to as Principal Accountable Providers, or PAPs). REBP then assesses (through retrospective analysis of claims data) the outcomes achieved and costs incurred during each episode, on average, over a specific period of time (e.g., quarterly). The PAP(s) for each episode are assessed based on their average performance across episodes they treated over that period, and can be rewarded, penalized, or neither accordingly.3

For episode-based performance measurement and payment, including REBP, to function correctly (and fairly), it is critical to account appropriately for clinical factors that affect the expected cost of delivering a specific episode of care for a specific patient. An effective approach should account for the fact that clinical factors affect the cost of different types of episodes to different degrees. An effective approach should also function correctly for dozens of different types of episodes, some of which may occur infrequently. For these reasons, current approaches to population-based risk adjustment (e.g., Johns Hopkins University’s Adjusted Clinical Groups, 3M’s Clinical Risk Groups, UCSD’s Chronic Illness and Disability Payment System) are not adequate for the task.

In this paper, we discuss the challenges posed by accounting for clinical factors in episodes of care and describe a scalable methodology we believe helps fairly measure episode performance even for relatively infrequent, or “low-volume,” episodes.

Context for clinical exclusions and risk adjustment

REBP relies on accurately, fairly, and consistently measuring the average medical spend for which a PAP is responsible when delivering an episode of care (from here on referred to as “episode spend”). We suggest several mechanisms that may be employed to achieve this goal:

  • Targeted inclusion of spend in the episode: Include only medical spend judged to be relevant to the episode in the calculation of episode spend. For example, during an asthma acute exacerbation episode, an inpatient admission related to the asthma acute exacerbation would be included, while an inpatient admission because of a broken leg would not.
  • Exclusion of episodes for business reasons: Take into account only those episodes for which the available claims information is comparable and complete. For example, episodes should be excluded for the following reasons: payment or eligibility rules (e.g., inconsistent enrollment, dual eligibility when only one payor’s claims data is available), select provider characteristics (e.g., certain provider types with special or exceptional payment rules, PAP not identifiable), or missing or exceptional claims information (e.g., long hospitalizations, missing DRG, incomplete claims).
  • Exclusion of episodes of certain patients: Include only those episodes that involve comparable patients. Exclude episodes of patients who, for example, are long-termcare residents, left against medical advice, died, or fall within certain age groups (unless the episode is defined around patients with these characteristics).
  • Exclusion or winsorizing of episodes with very high episode spend: Exclude or winsorize episodes with very high episode spend, which may indicate a patient had a unique or unusual event. 

Download the full article (PDF)

  1. Michael Bailit and Margaret Houy, “Key Payer and Provider Operational Steps to Successfully Implement Bundled Payments.” HCI3 Issue Brief, May 2014. See also: “Bundled payments: Moving from pilots to programs.” Health Affairs  (blog), June 2, 2014.
  2. Tom Latkovic, “The trillion dollar prize: Using outcomes-based payment to address the US healthcare financing crisis.” McKinsey Quarterly , March 2013, mckinsey.com.
  3. Adi Kumar, Tom Latkovic, and Daniel Tsai, “It’s time for episode-based health care spending,” Harvard Business Review  (blog), October 2013.