The reproducibility of a method to identify the overuse and underuse of medical procedures.

BACKGROUND To assess the overuse and underuse of medical procedures, various methods have been developed, but their reproducibility has not been evaluated. This study estimates the reproducibility of one commonly used method. METHODS We performed a parallel, three-way replication of the RAND-University of California at Los Angeles appropriateness method as applied to two medical procedures, coronary revascularization and hysterectomy. Three nine-member multidisciplinary panels of experts were composed for each procedure by stratified random sampling from a list of experts nominated by the relevant specialty societies. Each panel independently rated the same set of clinical scenarios in terms of the appropriateness of the relevant procedure on a risk-benefit scale ranging from 1 to 9. Final ratings were used to classify the procedure in each scenario as necessary or not necessary (to evaluate underuse) and inappropriate or not inappropriate (to evaluate overuse). Reproducibility was measured by overall agreement and by the kappa statistic. The criteria for underuse and overuse derived from these ratings were then applied to real populations of patients who had undergone coronary revascularization or hysterectomy. RESULTS The rates of agreement among the three coronary-revascularization panels were 95, 94, and 96 percent for inappropriate-use scenarios and 93, 92, and 92 percent for necessary-use scenarios. Agreement among the three hysterectomy panels was 88, 70, and 74 percent for inappropriate-use scenarios. Scenarios involving necessary use of hysterectomy were not assessed. The three-way kappa statistic to detect overuse was 0.52 for coronary revascularization and 0.51 for hysterectomy. The three-way kappa statistic to detect underuse of coronary revascularization was 0.83. Application of individual panels' criteria to real populations of patients resulted in a 100 percent variation in the proportion of cases classified as inappropriate and a 20 percent variation in the proportion of cases classified as necessary. CONCLUSIONS The appropriateness method is far from perfect. Appropriateness criteria may be useful in comparing levels of appropriate procedures among populations but should not by themselves be used to direct care for individual patients.