Decision-Making in Medicare Prescription Drug Plans: A Generative AI Approach to Consumer Behavior Analysis

Abstract

This manuscript introduces the topic of consumer decision-making in the Medicare prescription drug market, where an annual enrollee decision process is the first stage of a two-stage sequential treatment that informs long-term insurance plan choice and actual utilization and claims. We examine the potential for generative AI to analyze decision-making in this context, which is of interest in consumer behavior more generally and in healthcare areas such as marketing, where both low-stakes and high-stakes purchase decisions are made by information-limited consumers. Knowledge potentially generated by this study could be of interest to all stakeholders in the Medicare Part D program. In summary, we present two broad contributions to this study. Methodologically, we demonstrate the use of generative deep learning models for inferring consumer preferences and heterogeneity from observational data in a specific consumer products market with implications for evaluation and public policy. Moreover, the approach presents a potential non interventionist method for determining individual or subpopulation-specific treatment effects from uncontrolled big data. At a more specific industry level, this study considers decision-making in the high-stakes healthcare market. We illustrate that low-income adults, who may have health complications in addition to age-related problems, can suffer disbenefits from consumer misinformation. We view this as an important and often overlooked area of policy and management research.