Perverse Incentives and Savings Forgone: Year Three of Medicare’s Experiment with Bundled Payments

Monday, June 24, 2019: 10:30 AM
Wilson A - Mezzanine Level (Marriott Wardman Park Hotel)

Presenter: Andrew Wilcock

Co-Authors: Michael Barnett; Ateev Mehrotra; David Grabowski; J. McWilliams

Discussant: Michael E. Chernew

CMS initiated the Comprehensive Care for Joint Replacement Model (CJR) in April 2016, a five-year randomized trial where just under 200 metropolitan statistical areas (MSAs) were randomized to a mandatory bundled payment model for inpatient lower extremity joint replacements (LEJR) of the hip and knee. Over its first two-years, CJR modestly lowered LEJR episode spending (inpatient + 90 days post discharge) among participating hospitals without evidence of risk selection or deleterious effects on complications. In year three (2018), CMS chose to partially dismantle CJR due to regulatory and patient safety concerns, reducing the set of MSAs originally mandated to participate from 67 to 34. In addition, starting in January of 2018 CMS removed total knee arthroplasty from the inpatient only list, allowing orthopedic surgeons to deliver these procedures to fee-for-service Medicare patients as outpatients, further reducing the volume of LEJRs under CJR risk. Little is known about what impact these changes had on the performance of hospitals originally mandated to participate in the CJR model.

In this paper we evaluated episode spending, complications, patient selection and volume changes among hospitals originally mandated to participate in CJR (“treatment hospitals”) compared with hospitals originally mandated to no payment changes (“control hospitals”). Our focus was the adjusted differential change in our outcomes by year three, compared with changes by year one or two. In sub-analyses we explored changes in three hospital groups: hospitals mandated to remain in CJR in year three (“mandatory”), hospitals that chose to exit the program in year three (“dropouts”), and hospitals that volunteers to stay in the program in year three when they could have exited (“volunteers”). We included outpatient total knee arthroplasty episodes into our study sample.

By year three, we found that the differential spending effect of CJR among the original treatment hospitals was 40% smaller compared to year two (-520 vs -847, p<.001). Among these original treatment hospitals, 53% were mandated to remain in CJR, while 11% volunteered to stay in program, and 36% chose to dropout. The reduction in savings appears driven primarily by two mechanisms. Among mandatory hospitals, there was a smaller shift of total knee replacements into the outpatient setting (15% vs 26% in control hospitals). Then, among dropout hospitals the differential reduction in episode spending reduced by 60% from year 2 to year 3 (-281 vs -679, p<.001).

The savings of CJR observed in year 2 were markedly attenuated by year 3. Hospitals that were mandated to stay in CJR shifted less knee volume from the inpatient to outpatient setting, indicating that by only focusing on inpatient procedures, CJR may have created a perverse incentive to retain inpatient volume in the face of a broad secular change towards less-expensive outpatient procedures. Further, making the program voluntary allowed a number of hospitals to dropout of the program and this further reduced savings.