Exploratory Mixed-Data Phenomapping and Risk-Enrichment Frameworks for Long-Term Mortality After Acute Coronary Syndrome: A Single-Center Cohort Study

Background and Objectives: Long-term prognosis after acute coronary syndrome (ACS) may reflect multidomain baseline characteristics. We evaluated outcome-independent mixed-data phenotypes and post hoc additive frameworks for all-cause mortality. Materials and Methods: Of 3801 consecutive patients hospitalized with ACS during 2017–2024, 3561 formed the survival cohort after 240 patients lacking mortality status or valid follow-up were excluded. Gower–PAM phenomapping included 3491 patients after 70 patients with incomplete clustering variables were excluded. Cluster number was selected among k = 2–4 using prespecified outcome-independent silhouette, minimum-size, and bootstrap-stability criteria. AAR-3, CARE-4, and ACEF were compared in a common complete-case cohort of 3492 patients using an age-inclusive clinical reference model. Results: During a median follow-up of 3.35 years, 407 deaths occurred in the survival cohort. Three Gower–PAM phenotypes were selected (average silhouette width, 0.417), with crude mortality proportions of 12.1%, 4.2%, and 19.2% in phenotypes A, B, and C, respectively (global log-rank p < 0.001). The phenotype factor improved the primary age-inclusive model (joint p < 0.001), although only phenotype B differed from phenotype A after adjustment. The joint phenotype effect was attenuated after expanded clinical adjustment (p = 0.673), and agreement with k-means sensitivity analyses was low. In the common cohort, the first-year adjusted HRs were 1.78 per 1-point increase in AAR-3 (95% CI, 1.46–2.19), 2.08 per 1-point increase in CARE-4 (95% CI, 1.76–2.46), and 2.66 per 1-unit increase in ACEF (95% CI, 2.14–3.30). The corresponding >365-day landmark HRs were 1.56 (95% CI, 1.32–1.85), 1.66 (95% CI, 1.44–1.91), and 2.39 (95% CI, 1.95–2.92), respectively. Full-follow-up C-indices were 0.793 (95% CI, 0.772–0.814) with AAR-3, 0.815 (95% CI, 0.795–0.834) with CARE-4, and 0.812 (95% CI, 0.792–0.833) with ACEF, compared with 0.770 (95% CI, 0.747–0.792) for the age-inclusive reference model. None materially improved the model that already contained age, hemoglobin, eGFR, and LVEF continuously. Conclusions: Outcome-independent phenomapping identified exploratory, method-dependent patterns rather than ordered biological risk classes. The post hoc additive frameworks were internally associated with mortality but should not be regarded as validated prediction or treatment-directing tools. Prospective multicenter external validation is required before clinical application.