Previous Page  |  Table of Contents  |  Next Page

Gautam Gowrisankaran, Principal Investigator

Estimating Hospital Quality using a Bayesian Selection Model

This project is estimating the quality of care provided by different hospitals. The main innovation of the research is to control for patient selection using Bayesian econometric methods. The primary sample contains data on all pneumonia patients treated at Los Angeles County, California hospitals from 1989-92. The problem of selection is that patients who are more severely ill than average may choose to seek treatment at high quality hospitals. In this framework, the conceptual experiment that reveals quality is hospital-specific mortality rates following random patient assignment. This work controls for patient selection into hospital using distance as the main exogenous explanatory variable. Given the assumption that a patient's mortality is not affected by how far the patient is from alternative hospitals, the estimation will infer a hospital to be of high quality if patients residing near that hospital have a low residual mortality, regardless of where they seek treatment. Because the outcome variable, mortality, is dichotomous, standard linear econometric techniques cannot be used to transform this insight into estimation. The Bayesian simulation methods developed here fill that role. The methodology developed here exploits the similarity between this model and the conventional linear simultaneous equations model, were the latent utilities for choice and mortality observed. In addition to handling the latent variable problem, Bayesian inference makes it possible to address the motivating policy questions directly, by providing marginal posterior distributions for any functions of interest.

Computation of the primary model has now been completed. Conditional on the data set, the posterior distribution for the parameters of the model has a number of interesting substantive implications. There is substantial variance to the posterior distribution of quality of most individual hospitals. Nonetheless, there appears to be two key relations between hospital characteristics and quality. Hospitals that are small-fewer than 150 beds-have lower mortality than larger hospitals, and private for-profit hospitals have lower mortality rates than public, private teaching and private not-for-profit hospitals. There is also strong evidence that the level of unobserved severity of illness differs across hospitals. A high unobserved severity of illness is found to be positively correlated with estimated hospital quality.

In ongoing work, the researchers are constructing a separate sample of all heart disease patients treated at Los Angeles County, California hospitals during the same sample period. A similar econometric procedure is being used to analyze this sample. The results from the heart patients will potentially allow for comparison with results from pneumonia, a different, but equally important disease. It can then be examined to what extent the results are generalizable. This would ultimately provide much stronger results regarding the determinants of the quality of hospital care.

This information is available in alternative formats upon request by individuals with disabilities. Please send email to alt-format@msi.umn.edu or call 612-624-0528.

HOME | QUESTIONS | FEEDBACK
Employment | Events | Links | People | Programs | Publications | Support | Welcome