Targeted Quality Measurement of Health Care Providers

Abstract

Assessing the quality of cancer care administered by US health providers poses numerous challenges due to meaningful heterogeneity in patient populations. Patients undergoing oncology treatment exhibit substantial variation in disease presentation among other crucial characteristics. In this paper, we present a framework for institutional quality measurement that addresses this patient heterogeneity. Our framework follows recent advancements in health outcomes research, conceptualizing quality measurement as a causal inference problem. This approach enables us to use flexible covariate profiles to target specific patient populations of interest. We use different clinically relevant covariate profiles and evaluate methods for case-mix adjustments. These adjustments integrate weighting and regression modeling approaches in a progressive manner in order to reduce model extrapolation and allow for provider effect modification. We evaluate these methods in an extensive simulation study, comparing their performance in terms of point estimates and estimated rankings. We highlight the practical utility of weighting methods that can generate stable weights when covariate overlap is limited and alert investigators when case-mix adjustments are infeasible without some form of extrapolation that goes beyond the support of the observed data. In our study of cancer-care outcomes, we assess the performance of oncology practices for different profiles that correspond to important types of patients who may receive cancer care. We describe how the methods examined may be particularly important for high-stakes quality measurement, such as public reporting or performance-based payments. These methods have the potential to help inform individual patient health care decisions and contribute to progress toward more personalized quality measurement.