Clinical, structural and psychological phenotypes after myocardial infarction: a multi-domain cluster analysis

Objective — to identify clinically interpretable multi‑domain phenotypes after myocardial infarction (MI) from clinical, angiographic, echocardiographic and psychological variables, and to compare patient‑reported outcomes (PROs) and short‑term clinical outcomes across the phenotypes. Materials and methods. This single‑centre registry included 831 consecutive patients with acute MI. A multi‑domain dataset was available for 546 patients. After harmonisation and record‑level quality checks, 287 patients with complete Depression Anxiety Stress Scale‑21 (DASS‑21) scores formed the phenotyping cohort. K‑means clustering was performed on standardised clinical, coronary, left ventricular (LV) and DASS‑21 features. Outcomes were not used to derive clusters. Phenotypes were compared using Kruskal–Wallis, Pearson c2 or Fisher exact tests, as appropriate. PRO analyses used measure‑specific non‑missing denominators. Results. Three phenotypes were identified. P1 (n=62) was an older, predominantly female hypertensive/intermediate‑distress profile. P2 (n=63) was an adverse‑remodelling/high‑distress profile. P3 (n=162) was a younger, predominantly male preserved‑LV/low‑distress profile with the highest prevalence of current smoking. P2 had the lowest median LV ejection fraction (48%) and the highest LV end‑diastolic and end‑systolic volume indices (81 and 40 mL/m2). Its median DASS‑21 stress, anxiety and depression scores were 20, 20 and 18 points, respectively (all p<0.001). Global health visual analogue scale and the 12‑item Kansas City Cardiomyopathy Questionnaire (KCCQ‑12) were numerically lowest in P2, but these comparisons were exploratory because only 9 — 31 observations per phenotype were available for individual measures. At 6 months, the broad registry‑defined adverse‑event composite occurred in 87%, 75% and 88% of P1, P2 and P3, respectively (p=0.043). Four deaths occurred in total. Conclusions. Multi‑domain clustering revealed distinct structural and psychosocial profiles after MI. The findings support integrated assessment of LV remodelling, psychological distress and perceived health, but should be regarded as hypothesis‑generating until replicated in an independent cohort with more complete longitudinal PRO and outcome data.