This study proposes an econometric approach to account for quality heterogeneity in hospital efficiency analysis. Specifically, we consider quality as a latent variable and adapt the control function approach to estimate the stochastic frontier model. Our approach does not require any parametric assumptions on the distributions of inefficiency and error terms. Moreover, it allows for exploiting the information on latent quality to construct a measure to rank the relative quality of hospitals. Using a real hospital dataset, we found that the proposed measure of relative quality, which is derived endogenously from our model, can capture various dimensions of hospital quality.