Comparing accuracy of risk-adjustment methodologies used in economic profiling of physicians

This paper examines the relative accuracy of risk-adjustment methodologies used to profile primary care physician practice efficiency. Claims and membership data from an independent practice association health maintenance organization (HMO) were processed through risk-adjustment software of six different profiling methodologies. The Group R2 statistic was used to measure, for simulated panels of HMO members, how closely each methodology's cost predictions matched the panel's actual costs. All but one methodology explained at least 50% of panel cost variance with panels as small as 25 patients. Group R2 performance tended to be better when high-cost cases were included rather than excluded from the analyses.