Bayesian Hierarchical Pattern Mixture Models for Comparative Effectiveness of Drugs and Drug Classes Using Healthcare Data: A Case Study Involving Antihypertensive Medications

Abstract

Comparative effectiveness research aims, in part, to provide evidence most relevant to clinical decision making. One decision relevant to hypertensive patients is which therapeutic drug class is the most safe and effective. In addition, once a drug class has been chosen it would be useful to know whether there are differences in effectiveness between drugs within class. Randomized trials are unlikely to provide sufficient evidence for answering these questions. We therefore propose a modeling approach that can be used to address the questions using administrative databases. We propose a Bayesian hierarchical model, where drugs are nested within their corresponding class. We account for the type of missing data that are common in these databases using a pattern mixture model. The methodology is illustrated using data from a comparative effectiveness study of antihypertensive medications.