Targeted Quality Measurement of Health Care Providers

Motivated by the problem of measuring the quality of cancer care delivered by providers across the US, we present a framework for institutional quality measurement which addresses the heterogeneity of the public they serve. For this, we conceptualize the task of quality measurement as a causal inference problem, decoupling estimands from estimators, making explicit the assumptions needed for identification, and helping to target flexible covariate profiles that can represent specific populations of interest. We propose methods for layered case-mix adjustments that combine weighting and regression modeling approaches in a sequential manner in order to reduce model extrapolation and allow for provider effect modification. We evaluate these methods in an extensive simulation study and highlight the practical utility of weighting methods that warn the investigator when case-mix adjustments are infeasible without some form of extrapolation that goes beyond the support of the data. Specifically, our constrained optimization approach to weighting constitutes a diagnostic of sparse or null data for a given provider relative to the target profiles. In our study of cancer care outcomes, we assess the performance of oncology practices for different profiles that correspond to different types of patients that may receive cancer care. We describe how the proposed methods may be particularly important for high-stakes quality measurement, such as public reporting or performance-based payments. They may also be important for individual patients seeking practices that provide high-quality care to patients like them. Our approach applies to other settings besides health care, including business and education, where instead of cancer practices, we have companies and schools.