Measuring The Quality of Hospital Quality Assessment: Implications for Reference Pricing

The contention that the U.S. health care system simultaneously provides too much low-value care and too little high-quality care lies at the heart of many delivery system reform initiatives.  While these reforms differ in scope and design, they share a common feature in that they create more direct linkages between provider reimbursement and measures of quality.   This paper aims to estimate the causal relationship between these “report card” measures on patient outcomes.  To compare similar patients across hospitals in the same market, we build on the empirical approach outlined in Doyle et al. (2012) and exploit ambulance company preferences as an instrument for hospital assignment.

 Within this framework, we proceed in two steps.  The first is to consider the relationship between existing quality measures and outcomes, controlling for patient selection.  Doing so, we find great dispersion across measures.  First, we find that composite process quality scores are weakly associated with patient outcomes, casting doubt on these measures as a reliable proxy for overall quality.  Next, we find that readmission-based measures are at best weakly predictive of patient-level readmission risk, with no observed relationship between these measures and patient-level mortality outcomes at 30 days and 1 year.  At the same time, we find a strong correlation between risk-standardized hospital mortality rates and patient readmission and mortality outcomes. 

Second, we consider the appropriateness of hospital reference pricing and other quality-based delivery system reforms.  Once selection is corrected, if patient outcomes are comparable across hospitals of a fixed quality level, and if quality levels are reliably estimated, then there is no reason to pay more to more intensive/expensive hospitals.  We examine the advisability of reference pricing by estimating outcome regressions as a function of both quality measures and Medicare spending.  Here, we find that conditional on quality, there remains a strong positive correlation between hospital spending for initial hospitalizations and subsequent patient survival.   However, we also find an offsetting negative association between post-discharge spending and survival on the marginal cases that make up our instrument.    On balance, these returns offset each other, resulting in a finding of no association between total one-year  spending and patient outcomes – a common result in the literature that we confirm using a measure of hospital-level end-of-life spending (which includes both acute and non-acute spending).  These results suggest that tying inpatient reimbursement solely to quality measures may impede patient outcomes in some systems.  On the other hand, our results also point to considerable “efficiency slack” in the post-acute setting that may be improved through more broadly defined delivery system reforms such as Accountable Care Organizations and bundled payment programs.