Evaluating physician efficiency in hospitals: A multivariate analysis of best practices

This paper reports on a six-month study of the clinical efficiency of 35 physicians in a single hospital. Both technical and scale efficiency are analyzed using Data Envelopment Analysis (DEA) and a multi-factor Tobit analysis is conducted to see which variables are associated with higher levels of physician performance. DEA identifies 24 inefficient physicians. The slack associated with these physicians supports the hypothesis that a substantial amount of money could be saved if every physician practiced medicine as efficiently as the most competent physicians. The Tobit analysis revealed two categories of technically efficient physicians: those who (1) belong to a health maintenance organization or (2) specialize by diagnostic related groups. Estimates of most productive output scale suggest that locally decreasing returns to scale set in at higher output levels for physicians who treat higher proportions of high severity patients. The results illustrate how blending DEA with censored regression can sharpen an analysis of best practices.