The effects of Medicare’s Hospital Value-Based Purchasing program

The Patient Protection and Affordable Care Act initiated Hospital Value-Based Purchasing (HVBP) for all Acute Care Hospitals, making HVBP the first national P4P program in the United States.  Starting in FY 2013, Medicare reimbursement to Acute Care Hospitals is adjusted based on hospitals’ performance on incentivized measures. Beginning with clinical process measures and patient experience (FY 2013), the program has expanded to include additional patient outcome and efficiency measures in subsequent years. The magnitude of the program incentives increases gradually, from 1% of revenue in FY2013 to 2% by FY 2017.

This study evaluates the effect of HVBP on quality of care and patient experience during its first three years of implementation (FY 2013-FY2015). We exploit the fact that hospitals that do not receive reimbursement under the Medicare Inpatient Prospective Payment System (IPPS) – including Critical Access Hospitals, Acute Care Hospitals in Maryland, and Veteran’s Administration hospitals – are not exposed to HVBP. These hospitals are used as the comparison group for the IPPS hospitals that are exposed to HVBP.

Our study outcomes are the performance domains that are incentivized under HVBP: clinical quality, patient experience, patient outcomes (including 30-day mortality for heart attack, heart failure, and pneumonia as well as hospital acquired conditions), and efficiency (measured as 30-day episode costs). All study outcomes are measured at the hospital level.

Using data from up to eight years prior to HVBP and three years after, we perform a difference-in-differences analysis, comparing performance on incentivized clinical process and patient experience measures between hospitals exposed and not exposed to the program. Our model specification includes hospital fixed effects, and year fixed effects. Models are estimated separately for each performance domain. We use matching estimators — matching HVBP hospitals to non-HVBP hospitals on lagged outcomes — to ensure parallel trends between the treatment and comparison groups.  Extensive sensitivity analysis is performed in which we vary the pre-intervention time interval, the modeling of pre-intervention trends, and the composition of the comparison group. We also use quantile regression to evaluate heterogeneity in the effects of HVBP across the quality distribution.