Computerized Bayesian ADE Assessment

Abstract

We describe a Bayesian method for doing causality assessment of suspected adverse drug reactions (ADRs). The method uses the specific findings in a case to transform a prior into a posterior probability of drug causation. The approach balances evidence appropriately, does not artificially constrain the weight that any piece of evidence can carry, and is completely open-ended — there is no limit to the number of case details that can be assessed. We provide a very brief overview of two ways for implementing the method (a spreadsheet program and an expert system) and describe how Bayesian assessment might fit into an integrated and coherent system to predict ADR incidence based on data from postmarketing surveillance.