Mental Health
Heterogeneous Treatment Effects of CBT on Post-MI Depression and Survival: A Machine Learning Approach on ENRICHD Trial Yuta Takemura* Yuta Takemura Takemura Takemura Department of Neuropsychiatry, The University of Tokyo Hospital, Japan
Depression and low perceived social support (LPSS) are associated with poor prognosis after acute myocardial infarction (MI). The ENRICHD trial demonstrated that cognitive behavioral therapy (CBT) improved psychosocial symptoms but failed to reduce all-cause mortality or recurrent MI in the overall population. In this post-hoc analysis of the ENRICHD trial, we investigated the heterogeneous treatment effect (HTE) to determine if specific subgroups derive CBT benefit regarding depressive symptoms or survival. We applied two distinct machine learning algorithms: Causal Survival Forest (CSF) to estimate the conditional average treatment effect (CATE) on event-free survival (all-cause mortality and recurrent MI), and Causal Forest (CF) to estimate CATE on 6-month depressive symptom reduction (Beck Depression Inventory). If HTE was detected, participants were classified into high- or low-benefit groups based on the estimated CATE (defined as
