The case for case-mix adjustment in practice profiling. When good apples look bad.

OBJECTIVE: To assess the influence of patient characteristics on practice profiling. Using the example of specialty referrals by primary care physicians, we evaluated the impact of adjusting for patient characteristics (age/sex vs case mix) on the estimation of practice variation, the identification of outlier practices, and the evaluation of potential predictors of variation. DESIGN AND
SETTING: We applied several measurement strategies to a retrospective cohort of patients (N = 37,830) within 52 physician practices in a large staff-model health maintenance organization during a 1-year period. OUTCOME MEASURES: We calculated unadjusted referral rates and adjusted standardized referral ratios for each physician. Using these, we determined coefficients of variation and statistical "outlier status."
RESULTS: Adjustment for patient characteristics decreased the observed variation in referral profiles, with a decrease of more than 50% in the coefficient of variation. Three quarters of the physicians identified as statistical outliers with use of an age/sex-adjusted measure were no longer identified as such with use of an case-mix-adjusted measure. Several key predictors of unadjusted referral rate (including physician age, practice tenure, site of practice, and extent of laboratory test ordering) dropped out of regression models when the outcome variable was adjusted for patient characteristics.
CONCLUSION: Failure to adjust for case mix in physician practice profiles may lead to overestimates of variation and misidentification of outliers. To the extent that unadjusted practice profiles are used for decisions about education, sanctions, or employment, physicians may be subject to inequitable decisions and actions. Misinformation about the causes and extent of practice variation may also lead to misdirection of scarce resources for quality improvement efforts.