Challenges in Evaluating Bundled Payments: BPCI and beyond Daver C. Kahvecioglu, CMMI June 25, 2017 1
Outline of Presentation Overview of BPCI (Models 2-4) Evaluation Intended consequences of bundled payments Lower payments and improve quality? Internal validity challenges External validity challenges Unintended consequences of bundled payments Additional measurement challenges Conclusions 2
Overview of BPCI Models 2-4 Evaluation Possibly the largest evaluation of bundled payments both in scope and size Flexibility Extremely large dimensionality (bundle of bundles) 3 Models X 6 Episode Initiator Types X 48 clinical episodes X 3 bundle lengths X rolling entry and exit Separate matched comparison groups needed for many of these combinations New comparison groups needed in each quarter due to rolling entry and exit 3
Challenges: Internal Validity Minimize selection bias in estimated impacts on cost and quality At the participant level At the beneficiary/episode level BPCI evaluation Matched comparison episode initiators Difference-in-differences estimation Regression-based risk adjustment Absent randomization, no guarantee that unobserved differences are not biasing the results 4
Challenges: External Validity How generalizable are the results for the U.S. at large? Generally an issue for all voluntary models Will want to estimate impacts on the subgroups of interest such as Episode Initiators with low use of post-acute care in the baseline if sufficient sample size adds to the dimensionality 5
Unintended Consequences of Bundled Payments Unintended consequences of bundled payments Shifts in patient mix? Shifts in costs? Unwarranted increase in volume? Decline in quality? Need to define and track many metrics – adds to the dimensionality Arbitrage: Possibility of exploiting imperfect target prices 6
Measurement Challenges Adequate sample size difficulty reliably measuring important characteristics Availability of data functional assessment data not available for all beneficiaries measured at the same time after discharge Beneficiary survey partially remedies this, but still no baseline data on the beneficiary 7
Conclusions For credible and reassuring evidence on bundled payments, it is important to Identify adequate comparison groups (Episode Initiators) for voluntary models Adjust for the heterogeneity of patients even within specific clinical episodes Ensure there are no unintended consequences Generalizability to the U.S. system at large will be limited for voluntary models 8
Recommend
More recommend