RESUMEN
The Bundled Payments for Care Improvement (BPCI) initiative established four models to test whether linking payments for an episode of care could reduce Medicare payments while maintaining or improving quality. Evaluations concluded that model 2, the largest, generally lowered payments without reducing quality for the average beneficiary, but these global results could mask adverse findings among vulnerable subpopulations. We analyzed changes in emergency department visits, unplanned hospital readmissions, and all-cause mortality within ninety days of hospital discharge among beneficiaries with one or more of three vulnerable characteristics-dementia, dual eligibility for Medicare and Medicaid, and recent institutional care-in 105,458 beneficiary episodes in the period October 2013-December 2016. The results for twelve types of medical and surgical BPCI episodes were evaluated relative to results in matched comparison groups. Our findings suggest that BPCI model 2 did not adversely affect care quality for beneficiaries with vulnerabilities. While this conclusion does not discourage the further development of bundled payment models, policy makers should support ongoing research to ensure that vulnerable populations are not adversely affected by these approaches.