Two-Year Evaluation of Mandatory Bundled Payments for Joint Replacement

Background: In 2016, Medicare implemented the Comprehensive Care for Joint Replacement Model (CJR), a national mandatory bundled payment model for lower extremity joint replacement (LEJR) in randomly selected metropolitan statistical areas (MSAs). Hospitals in selected MSAs receive bonuses or pay penalties based on LEJR spending through 90 days post-discharge. Evaluations of the first year of CJR found modest decreases in spending, though not always statistically significant. As CJR matures, it is unclear whether these savings become larger and whether selection against riskier patients becomes evident. We compared changes in outcomes in the first two years of the CJR program between MSAs randomized to the new bundled payment model (treatment MSAs) and control MSAs.

Methods: We conducted difference-in-differences analyses using Medicare claims from 2015-2017, comparing LEJR episodes in 75 treatment MSAs randomized to mandatory participation in CJR with episodes in 121 control MSAs, before vs. after CJR implementation in April 2016. Our primary outcomes were institutional spending per LEJR episode (primarily hospital and post-acute facility payments), rates of post-surgical complications, and the proportion of patients with higher spending risk (a measure of patient selection). Secondary outcomes included total spending per episode in a random 20% sample of episodes, use of post-acute care and length of stay in post-acute facilities. We fit linear regression models at the LEJR episode-level controlling for patient and procedure characteristics as well as quarter, hospital and MSA random effects, with the interaction between the post-intervention period and an indicator for treatment MSA as the parameter of interest.

Results: In 2015-2017 there were 280,161 LEJR procedures in 803 hospitals in treatment MSAs and 377,278 procedures in 962 hospitals in control MSAs. After initiation, institutional spending per LEJR episode declined more in treatment than in control MSAs (-$812 or -3.1% relative to the treatment group baseline; p<0.001). This differential decrease in institutional spending grew over the 18-month period of CJR implementation from -$541 in April-September 2016 (95% CI -754, -328) to -$860 in April-September 2017 (95% CI -1075, -645). Total episode spending differentially decreased by $1,084 (95% CI -1409, -760), or a 3.6% differential reduction. The differential reduction was driven largely by a 5.9% relative decrease in the fraction of episodes with any institutional post-acute care. There was also a differential reduction of 1.7 days in length of stay (95% CI -2.0, -1.4) for post-acute facilities among those using any institutional post-acute care in treatment MSAs. The program had no effect on complication rates (p=0.67) or the fraction of LEJR procedures performed on higher-risk patients (p=0.81). Estimates were similar using fixed effects, generalized linear models or logistic regression for their respective outcomes.

Conclusions: In its first two years, the CJR program resulted in a modest reduction in spending per LEJR episode without increasing complication rates or leading to widespread selection against higher-risk patients. The decreased spending came nearly exclusively from lower use of inpatient post-acute care. Savings grew over period, raising the possibility that CJR could lead to greater spending reductions as hospitals adapt to the new model.