HCI3 Improving Incentives Issue Brief: Designing the BPCI for Success

CMMI Bundled Payments for Care Improvement Requires Design Changes to Ensure Pilot Success

Objective: Examine the variation in costs within certain episodes of care to better quantify the risks and benefits to providers participating in the CMMI Bundled Payments for Care Improvement (BPCI).

Methods: We performed a retrospective analysis of claims data using the SAS-based freeware application developed by Brandeis University and the Health Care Incentives  Improvement Institute (HCI3) for the BPCI initiative. We used seven Hospital Referral Cluster (HRC) datasets comprised of Medicare Fee For Service claims across the United States. The study population consisted of 67,746 major joint replacement surgeries, 34,912 percutaneous coronary intervention (PCI) procedures, 20,435 Heart Failure hospitalizations, and 55,362 Pneumonia hospitalizations between January 1, 2008 and December 31, 2009. We examined the variation in average episode costs by (1) comparing episode costs for major joint replacement and percutaneous coronary intervention (PCI) MS-DRGs, split by Principal ICD-9 Procedure and Diagnosis codes respectively; (2) analyzing the magnitude of variation in heart failure episode costs as the episode time windows are increased from 30 days to 90 days to 180 days; and (3) studying the effects of outlier episodes on pneumonia costs. In addition, we also evaluated the impact of excluding Indirect Medical Education (IME), Disproportionate Share (DSH) and Capital add-on amounts from the episode (chronic, acute, and procedural) bid price by analyzing a range of episodes triggered at 10 Academic Medical Centers (AMCs).

Results: For episodes triggered by Joint Replacement DRGs 469 and 470, we observed significantly different average episode costs for Total Hip Replacements, Partial Hip Replacements, and Total Knee Replacements within the same DRG. Similarly, we found differences in average episode costs for PCIs based on the underlying etiology. Patients undergoing PCI had much lower average episode costs if their underlying condition was stable coronary artery disease (CAD) compared to patients with more acute diagnoses, such as acute myocardial infarction (AMI) or Cardiac Dysrhythmias. We observed that in patients hospitalized for heart failure (CHF), the variability in episode costs increases as the episode time window is extended, primarily due to variability in post acute care costs, and that a relatively small number of patients could account for a substantial increase in average costs due to readmissions and long-term care. We also observed that Pneumonia patients that die during the episode time window have much lower average episode costs. Finally, the impact of reducing readmissions within the context of the BPCI would lead to reductions in mission related add-on payments to academic medical centers, and the reductions would increase from $200,000 if the readmit reduction is 10% to $500,000 if the reduction is 25%.

Policy Implications: Depending on the DRG-based episode selected, the current proposed design for the BPCI can create an opportunity for providers to beat or come over their bid price based on factors outside their control that have very little, if anything, to do with clinical management. Building a price by principal procedure or principal diagnosis code (depending on the episode of interest) reduces the variability due to patient mix and creates more clinical homogeneity around the pricing. BPCI applicants should understand the source of the added variability related to longer episode time windows and consider asking for stop-loss when episodes exceed the bid price by a specified amount. Similarly, variability in episode costs, due to deaths or high outlier episodes, could be mitigated by excluding deceased patients and capping high cost outliers when calculating the target price. Creating a more stable bid price and flagging the number of patients who die or are recognized as outliers, will reduce perverse incentives and the likelihood of gaming in the implementation period. In addition, some allowance for the expected loss in mission-related add-ons for AMCs would reduce the current perverse incentive in the BPCI that would penalize these hospitals for reducing readmissions.