Machine Learning in Health Economics: Modeling Costs, Outcomes, and Policy Decisions

Abstract

Health economics increasingly faces complex challenges in balancing rising costs, improving outcomes, and informing policy decisions under constrained resources. Traditional econometric models often struggle with nonlinearities, high-dimensional data, and heterogeneous treatment effects. Machine learning (ML) provides a promising alternative by enabling more accurate predictions of healthcare costs, health outcomes, and cost-effectiveness metrics. This paper explores how ML methods—such as random forests, gradient boosting, and neural networks—can enhance economic evaluation frameworks. Using a representative dataset of 100,000 patients over a 5-year horizon, we demonstrate that ML models reduce prediction error by up to 35% compared to generalized linear models and provide improved identification of high-cost patients. By integrating cost and outcome predictions into incremental cost-effectiveness ratio (ICER) and net monetary benefit (NMB) frameworks, ML significantly alters policy decisions, particularly under budget-constrained scenarios. Findings suggest that ML can not only improve efficiency in modeling but also shape fairer and more effective health policies.