Creation and evaluation of composite measures of physician practice quality using aggregated health insurance claims.

摘要

This study uses mixed-methods to study composite measures of physician practice quality. First, using qualitative interview data, this study describes the opinions of stakeholders related to the utility and development of composite measures. Also, using aggregated health insurance claims, this study investigates a number of important properties of composite measures. First, the study investigates whether multiple measures of physician practice quality conform to a latent factor structure that can be used to create composite measures. Second, the study will also test the extent to which different techniques used to calculate composite measures lead to differences in physician practice ratings. Third, the study will also test the stability of physician practice quality measurement and the extent to which composite measures may improve that stability.;The findings raise fundamental questions about the potential utility of composite measures. Particularly, based on the data available in this setting, a single unidimensional measure of physician practice quality should not be created, significantly limiting the potential uses of composite measures. This would suggest that composite developers in other settings should use similar factor analytic approaches to determine if a single unidimensional measure might be appropriate. However, if unidimensional measures cannot be created, the potential utility of composite measures is greatly diminished. Also, composite measures address very few of the concerns stakeholders had related to physician quality measurement, and they did very little to address a specific potential application of composite measures related to data lags. Although composite measures likely still hold some utility related to improving decision making by consumers and health plans, more time should be focused on addressing more fundamental issues with quality reports that cannot be addressed by composite measures.