Joshua A. Rolnick, Joshua M. Liao, and Amol S. Navathe

Bundled payments have become a prominent value-based payment arrangement. The US recently concluded a five-year national programme, the Bundled Payments for Care Improvement (BPCI) initiative, prior to launching a similarly designed and ongoing successor programme. In both, episodes start with a hospitalization and extend up to 90 days after discharge. All hospital and post-hospitalization services occurring within this episode window (e.g. nursing home stays, home health services, office visits, and hospital readmissions) are grouped together for the purposes of holding hospitals and physician groups financially accountable for episode spending.

Our recent paper sheds new light on the impact of bundled payments on patients with common medical conditions such as heart failure and chronic obstructive pulmonary disease. Examining the first three years of BPCI, we found that hospitals achieved modest savings by reducing the amount of nursing home care used in favor of increased home health services. These results provide the first account of hospitals changing practice patterns in response to bundled payments for medical conditions.

In generating these results, we grappled with a critical, yet often overlooked, policy question: how should episodes be constructed and assigned to bundled payment participants and non-participants?

Though it may seem technical, the issue is in fact highly salient to policy and can have a surprisingly large impact on how episode performance is measured. At the core of the issue is the fact that episodes are not directly observable clinical events. Instead, they must be constructed by first identifying a trigger (in the case of BPCI, hospitalization) then following out for the defined episode duration (in the case of BPCI, 90-days after hospital discharge). This process is usually straightforward for surgical episodes, as patients tend to have one very clearly defined procedure (e.g., joint replacement, coronary artery bypass).

In contrast, applying episode construction rules is more complex for medical condition episodes because patients may have multiple, frequent events that could serve as episode triggers. In the case of BPCI, patients could experience repeated hospitalizations over a short period of time, creating the need to sort hospitalizations into initial triggering admissions and non-triggering readmissions. Doing so when patients seek care at multiple hospitals further complicates the process.

These choices matter. Is the hospitalization that triggers an episode always the earliest one? The latest? When episodes can be assigned to multiple hospitals, should the one with the longest experience in the bundled payment program take precedence? The answers can determine which clinicians and hospitals are financially responsible for episode care, and could even change the apparent success or failure of a bundles program. These choices are also unavoidable. Whether intentionally or not, policymakers must make these choices through bundled payment program rules. 

In our study’s main analysis, we used a “naturally occurring” method that assigned the episode to the earliest hospitalization—in other words, we used patterns of care that existed independent of bundled payments to guide episode construction before evaluating the effects using those patterns. By creating episodes and assigning them to participants and nonparticipants without concern for the program, we were able to measure performance with as little bias as possible. 

We contrasted this approach in a secondary analysis to one in which we assigned episodes using “program precedence”, i.e., rules that preferentially attributed episodes to hospitals participating in bundled payments. In effect, what this meant was that when hospitalizations clustered in time across different hospitals for a given patient, episodes were over-attributed to participant hospitals, which in turn, had inflated readmissions rates and greater total spending. This over-attribution artificially washed away savings we would have otherwise observed with program participation. Notably, the new BPCI-Advanced program uses precedence rules and to our knowledge previous studies implemented BPCI’s precedence rules when evaluating BPCI. In other words, the stakes are high here.

Our example underscores that if bundled payment programs are to achieve their intended purposes, policymakers must design policy in ways that address potential bias and ensure fairness to participants. Investigators and policymakers evaluating bundled payment programs need to pay careful attention to the “guts” of how episodes are defined, accounting for them and transparently presenting the approach used. Failure to do so risks distorting the apparent effectiveness of bundled payments as a payment model and dampening enthusiasm from both policymakers and would-be participants.   

Joshua A. Rolnick is a physician at the Corporal Michael J. Crescenz VA Medical Center. 

Joshua M. Liao is an Assistant Professor of Medicine at the University of Washington. 

Amol S. Navathe is an Assistant Professor of Medicine and Health Policy at the University of Pennsylvania. 

Competing interests: Dr. Rolnick reported consulting fees from Tuple Health Inc. Dr. Liao reports textbook royalties from Wolters Kluwer. Dr. Navathe reported receiving grants from Hawaii Medical Service Association, Anthem Public Policy Institute, the Healthcare Research and Education Trust, Cigna, and Oscar Health; personal fees from Navvis and Company, Navigant Inc., and Agathos, Inc.; personal fees and equity from NavaHealth; equity from Embedded Healthcare; speaking fees from the Cleveland Clinic; serving as a board member of Integrated Services Inc. without compensation, and an honorarium from Elsevier Press, none of which are related to this opinion piece.