Fatal cardiovascular events after first acute ST-segment elevation myocardial infarction remain difficult to predict when psychosocial and lifestyle information is omitted. Following the published CRAS-MI methodology and ethics approval IR.NIMAD.REC.1397.295, we analyzed 1,730 modeling records from a multicenter cohort with annual follow-up over 3 years. The endpoint was fatal major adverse cardiovascular events, defined as cardiovascular death from coronary artery or cerebrovascular disease. The dataset contained 79 fatal outcomes (4.6%); mean age was 56.2 +/- 9.9 years, mean body mass index was 26.6 +/- 4.1 kg/m2, and 82.3% were male. A TabNet classifier used clinical, laboratory, lifestyle and psychosocial predictors, derived interaction terms, tuned class weighting, repeated stratified five-fold cross-validation and isotonic calibration. Out-of-fold evaluation yielded AUROC 0.836 (95% CI 0.799-0.877), Brier score 0.037 (0.030-0.044), sensitivity 0.709 (0.612-0.814), and specificity 0.795 (0.776-0.814). Feature importance showed that clinical measures, sleep efficiency and psychosocial variables carried complementary prognostic signal.