On uncertainty in medical testing.

There is confusion in the medical decision-making literature about how to handle uncertainty in medical tests. In this article, the authors consider the situation in which there is uncertainty about the pretest probability of a disease in a patient as well as uncertainty about the sensitivity and specificity of a diagnostic test for that disease. They discuss how to calculate posttest probabilities of a disease under such uncertainty and how to calculate a distribution for a posttest probability. They show that given certain independence assumptions, uncertainty about these parameters need not complicate the calculation of patient positive predictive values: One can simply use the expected values of the parameters in the standard Bayesian formula for posttest probabilities. The discussion on how to calculate distributions for positive predictive values corrects a common and potentially important error.

Duke Scholars

Author Robert L. Winkler Fuqua School of Business