ESTIMATING PHYSICIAN COSTLINESS - AN EMPIRICAL BAYES APPROACH

Outpatient data obtained from the general medicine practice of an urba n, health care facility are used to provide an application of empirica l Bayes techniques in the estimation of physician ''costliness.'' The results illustrate that application of the simplest empirical Bayes es timation procedure can provide more reasonable estimates of physician' s utilization of resources than a standard estimation procedure. Empir ical Bayes estimates are shown to adjust for potential instability in standard estimates that may arise from either a physician treating a s mall number of patients or an inappropriate case-mix adjustment. Using simulation, it is demonstrated that the empirical Bayes procedure can provide overall better estimates using fewer data than the standard p rocedure. This application, although somewhat limited in scope, should provide impetus for increased utilization of the numerous Bayesian an d empirical Bayes techniques that currently exist in the statistical l iterature and pertain to small area estimation techniques.