Evaluating medical providers in terms of patient health disparities: a statistical framework

Statistical methods for provider profiling are designed to measure healthcare quality and create incentives for low-quality providers to improve. While most existing healthcare quality metrics assess the overall treatment delivered to providers' entire patient populations, widespread racial and sociodemographic disparities in health outcomes highlight the need to evaluate providers' healthcare quality across patient subgroups. Disparity measure development at the healthcare provider level is a nascent research area, and existing measures have lacked the statistical rigor and clear definitions of performance benchmarks that are characteristic of traditional provider profiling methodology. In response to these gaps, we propose a formal disparity model for provider evaluations, which can be used to construct null hypotheses and test statistics for concepts such as systemic disparities, equality in performance, and equity in performance. For each of these topics, we define the appropriate performance benchmark in terms of statistical parameters, and describe its relationship with the existing literature. These arguments also shed light on the long standing debate regarding social risk adjustments in assessments of overall healthcare quality. Finally, we develop an assessment chart to easily visualize disparity patterns and identify the most culpable providers. These methods are demonstrated through analyses of racial disparities in access to transplantation among End-Stage Renal Disease patients.