[Cardiovascular age prediction based on large language models and prognostic value analysis in acute coronary...

Objective: To investigate the application value of cardiovascular age prediction based on large language model for prognostic assessment in patients with acute coronary syndrome (ACS). Methods: This retrospective cohort study included 47 249 ACS patients who were admitted to the Department of Cardiology at the General Hospital of Northern Theater Command and underwent percutaneous coronary intervention between November 2015 and May 2023. Cardiovascular age was predicted using a large language model for cardiovascular critical care developed by our institution, incorporating patients' demographic characteristics, medical history, and laboratory data. The difference between cardiovascular age and chronological age was calculated as the Age-Gap. Based on previous studies, patients were categorized into the high Age-Gap group (Age-Gap≥8 years) and the low Age-Gap group (Age-Gap<8 years). The primary endpoint was major adverse cardiovascular and cerebrovascular events (MACCE), including cardiac death, myocardial infarction, stroke, and target vessel revascularization. Secondary endpoints included the individual components of MACCE and all-cause mortality. Pearson correlation analysis was used to assess the correlation between cardiovascular age and chronological age. Cox proportional hazards models were used to evaluate the association between Age-Gap and clinical outcomes, and restricted cubic spline analyses were applied to explore potential nonlinear relationships between Age-Gap and study endpoints. Results: The chronological age was (61.1±10.4) years, the cardiovascular age was (69.5±10.8) years. There were 31 775 patients in the high Age-Gap group and 15 474 in the low Age-Gap group. Cardiovascular age was positively correlated with chronological age (r=0.87, P<0.000 1). The follow-up duration was 39.6 (24.0, 59.6) months. The incidence of MACCE was significantly higher in the high Age-Gap group than that in the low Age-Gap group (12.99% (4 127/31 775) vs. 7.86% (1 216/15 474), P<0.000 1). After adjustment for age, sex, smoking status, and other confounders, Cox regression analysis showed that patients in the high Age-Gap group had a significantly increased risk of MACCE compared with those in the low Age-Gap group (HR=1.47, 95%CI 1.37-1.58, P<0.000 1). Higher risks were also observed for cardiac death (HR=2.72, 95%CI 2.36-3.15, P<0.000 1), myocardial infarction (HR=1.77, 95%CI 1.46-2.14, P<0.000 1), stroke (HR=1.24, 95%CI 1.05-1.46, P=0.010 9), and all-cause mortality (HR=2.01, 95%CI 1.81-2.23, P<0.000 1). Restricted cubic spline analyses demonstrated a nonlinear association between Age-Gap and the risks of MACCE, cardiac death and all-cause mortality, with a marked increase in risk as Age-Gap increased (all P-nonlinear<0.000 1; all P-overall<0.000 1). Conclusions: The application of large language model to predict cardiovascular age in ACS patients is feasible, and the Age-Gap exhibits a significant correlation with the risk of MACCE and mortality in this patient population. The cardiovascular age predicted by large language model offers a novel perspective for individualized risk assessment in ACS patients.