Multivariate Bayesian Additive Regression Trees for Cost-Effectiveness Analysis in Health Economics

Mateus Maia Marques (University of Glasgow)

Wednesday 2nd April 14:00-15:00 Maths 311B

Abstract

Bayesian additive regression trees (BART) have emerged as a highly effective nonparametric method for regression. Motivated by cost-effectiveness analyses in health economics—where it is crucial to jointly model healthcare costs and health-related quality of life—we propose a novel multivariate extension of BART for regression analyses with multiple dependent outcomes. Our framework accommodates both continuous and binary outcomes and addresses key limitations of existing multivariate BART approaches by allowing outcome-specific ensembles of trees while preserving dependencies across responses. For continuous outcomes, the model can be viewed as a nonparametric analogue of seemingly unrelated regression; for binary outcomes, it generalises the multivariate probit model. We introduce flexible, interpretable prior specifications and describe MCMC algorithms for posterior inference. The proposed methods are implemented in the R package subart. We demonstrate their effectiveness through simulation studies and an empirical case study in health economics, evaluating the cost-effectiveness of a novel trauma care intervention, also incorporating propensity scores for causal interpretation.

Add to your calendar

Download event information as iCalendar file (only this event)