Pay for Performance in the US: Did Early Adopting Hospitals Perform Better?

The Hospital Value-Based Purchasing (HVBP) program is a nationwide pay-for-performance program that creates incentives for hospitals to provide higher quality care for Medicare beneficiaries. The evidence that it has had a positive impact on outcomes is limited although advocates argue that it takes time for hospitals to make care meaningfully better. Given that HVBP was modeled closely after the voluntary Hospital Quality Incentive Demonstration (HQID) from 2003 to 2009, these hospitals have now been under financial incentives for quality for nearly a decade.  Therefore, we hypothesize that these “early adopters” perform better under HVBP than the “late adopters” – the rest of the U.S. hospitals who joined pay-for-performance when it became obligatory in 2011. Our study aims to identify the differences between the early and late adopting hospitals in terms of 30-day risk-adjusted mortality, clinical process scores, and financial bonuses under HVBP. As a robustness check, we estimated the differences in financial readmission penalties and patient experience scores, which were not part of the HQID.

Using the 100% Medicare inpatient claims file from 2013 (n = 783,575), we identified Medicare patients hospitalized in the U.S. with acute myocardial infarction, congestive heart failure, and pneumonia. We then calculated 30-day mortality rates and readmission rates for these conditions using the CMS-risk adjustment methodology.  We used the American Hospital Association survey to identify hospital characteristics, linked with Hospital Compare data to obtain individual hospital performance on clinical processes, patient experience, and whether they received financial bonuses or penalties under HVBP.  Among the 2,933 hospitals participating in HVBP in 2013, 232 were early adopters. Using exact matching, we identified 994 late adopters with the same observed characteristics (size, region, ownership, teaching status, urban-rural locality, having an ICU, and safety-net status).  We then estimated regression models to compare the early and late adopters on the three outcome measures. We observed no difference in 30-day mortality rates. We did note differences in clinical process scores (+1.45 percentage points, p = 0.069) and found that early adopters were significantly more likely to receive a bonus under the HVBP program (OR 1.65, p = 0.001). In robustness check, we found no evidence that early adopters were more adept at avoiding readmission penalties although they appeared to have somewhat higher patient experience scores (+0.58 percentage points, p=0.079).  Further research using both our 100% Medicare inpatient claims and Hospital Compare data from 2009 to 2014 will be performed (before ASHEcon) to estimate difference-in-differences models for all outcomes. We will also determine whether the better performance of early adopters is driven by: 1) unobserved characteristics that made them more likely to self-select into HQID or 2) a competitive advantage given prior experience with financial incentives under HQID.

We conclude that early adopting hospitals seem to perform better in terms of receiving financial bonuses. However, there were no differences in patient mortality.  As experience under HVBP continues, these findings suggest that we should expect at best modest improvements in hospital quality, even over the long run.