Hierarchical Bayesian Benefit–Risk Modeling and Assessment Using Choice Based Conjoint

Abstract

Assessment of benefit–risk for treatments is usually complex and involves trade-offs between multiple, often conflicting, preferences. Although discrete choice experiments are increasingly used in health outcomes research to assess trade-offs in preferences, their usage in benefit–risk assessment so far is fairly limited. This is primarily because of the high cognitive burden to assess multiple attributes and the requirement for a large pool of respondents. The hierarchical Bayes benefit–risk model using choice based conjoint that we propose drastically reduces the cognitive burden as each respondent only needs to evaluate a small fraction of preference questions, each of which compares a single pair of attributes. This method also leverages the Bayesian framework to borrow strength for analysis with a limited number of respondents. This article illustrates both a simulated experiment and a pilot experiment incorporating experts’ preferences in oncology. Ultimately, patients are the most important voice in the benefit–risk balance. Therefore, we propose an augmented model to obtain a more precise estimate of benefit–risk preferences based on patients’ characteristics.