Medical evidence evaluation using a multi-stage Bayesian approach

Abstract

Analyses of errors in medical diagnoses have generally been confined to the application of single-stage Bayesian models. While this approach is illuminating, it limits the scope of analyses to those inherent in the equipment and/or the test. However, the ability of a physician to correctly interpret the results produced by equipment, or a test, could also have a profound effect on the accuracy of diagnosis. Using a single-stage Bayesian approach underestimates the likelihood of error. In this paper, we show that by using a multi-stage Bayesian model we can incorporate the ability of a physician to interpret test results accurately. By combining the physician's ability with the accuracy of the equipment or test, it is possible to obtain a better estimate of the probability of making mistakes with medical diagnoses.