Can we measure hospital quality from physicians' choices?

Abstract

In this paper, we propose an alternative methodology for ranking hospitals based on the choices of Medical School graduates over hospital training vacancies. Our methodology is therefore a revealed preference approach. Our methodology for measuring relative hospital quality has the following desirable properties: a) robust to manipulation from hospital administrators; b) conditional on having enough observations, it allows for differences in quality across specialties within a hospital; c) inexpensive in terms of data requirements, d) not subject to selection bias from patients nor hospital screening of patients; and e) unlike other rankings based on experts' evaluations, it does not require physicians to provide a complete ranking of all hospitals. We apply our methodology to the Spanish case and find, among other results, the following: First, the probability of choosing the best hospital relative to the worst hospital is statistically significantly different from zero. Second, physicians value proximity and nearby hospitals are seen as more substitutable. Third, observable time-invariant city characteristics are unrelated to results. Finally, our estimates for physicians' hospital valuations are significantly correlated to more traditional hospital quality measures.