Evaluation of the Maryland Quality-Based Reimbursement Program: Does Paying for Performance Lead to Greater Improvements in Quality than Paying for Reporting?

Abstract

Background: In value-based purchasing models for health care, sometimes called pay-for-performance (P4P), providers’ compensation is linked to performance on selected quality-of-care measures. In 2008, Maryland implemented the all-payer Quality Based Reimbursement (QBR) Initiative for acute inpatient care, which linked hospital payments to performance in clinical quality measures for heart attack, heart failure, pneumonia, and surgical care infection prevention (SCIP). QBR uses a budget-neutral redistribution of DRG-based revenue from poor-performing hospitals to high-performing hospitals in the form of adjustments to future DRG payment rates based on process-of-care (POC) quality indicators measured in the previous year.

Purpose: This study evaluated the effect of the QBR program on POC quality indicators that were included in the QBR reimbursement formula and spillover effects on POC indicators that were excluded.

Methodology/Approach: Using quarterly POC quality indicators from 2007-2011 from the CMS Hospital Compare website, we estimated difference-in-differences, comparative time series econometric models of the impact of QBR on individual POC measures, condition-specific composite measures, achievement of benchmarks, and quarterly changes. We compared all 47 Maryland acute-care hospitals to 54 Pennsylvania hospitals that were not subject to a systematic P4P program during the study period. The models controlled for hospital and county characteristics, with fixed effects for quarter. The regression coefficient for the interaction of the Maryland/Pennsylvania and pre/post indicator explanatory variables indicated the effect of QBR.

Results: We found no systematic effect of QBR on POC quality indicators used in the reimbursement formula nor spillover effects on indicators that were not included in the formula. We noted from box plots of trends for each indicator that low-performing Maryland hospitals seemed to be improving more than low-performing Pennsylvania hospitals, so we used logistic and quantile regressions with additional outcome measures (likelihood of exceeding common benchmarks, quarterly change) to determine if this apparent pattern was systematically associated with QBR, but were unsuccessful. The few significant indicators of a QBR effect were more likely caused by random variation in significance than by a systematic effect.

Conclusion: Previous research on revenue-redistribution P4P programs has been mixed—about as many studies showing favorable outcomes as unfavorable outcomes—so there is no consistent evidence to support the value of these program. Our study is consistent with previous studies that found no effects, although we think that the hypothesis that low performing hospitals might benefit from further investigation.

Practice Implications: This study adds to the evidence suggesting that revenue-redistribution P4P programs based on POC measures do not produce sufficient benefit to justify their cost.