Juan M. Contreras, Beomsoo Kim, and Ignez M. Tristao
We examine the "learning by doing" hypothesis in medicine using a longitudinal census of laser in situ keratomileusis (LASIK) eye surgeries collected directly from patient charts. LASIK surgery has precise measures of presurgical condition and postsurgical outcomes. Unlike any other surgery, the impact of unobservable conditions on outcomes is minimal. Learning by doing is identified through observations on surgical outcomes over time for each doctor. Our unique data set overcomes some of the major measurement problems in health outcomes, and enhances the possibility of identifying the impact of learning by doing separate from other effects. Our results do not support the hypothesis that doctors' individual learning improves outcomes, but we find strong evidence that experience accumulated by doctors as a group in a clinic significantly improves outcomes.