Electrocardiographic age derived from routine ECG parameters and subsequent mortality: Guangzhou Biobank Cohort Study (Preprint)

BACKGROUND Chronological age does not fully capture heterogeneity in cardiovascular risk. Electrocardiograms (ECGs) provide a widely available, non-invasive source of age-related electrical and physiological information. We developed an ECG-derived age model, evaluated its associations with mortality, and assessed its predictive value relative to a conventional-risk-factor-derived heart-age marker. OBJECTIVE To develop and internally validate an ECG-age model based on standard ECG parameters, evaluate its association with all-cause and cardiovascular mortality, and compare its 10-year mortality discrimination with that of Heart Foundation Heart Age (HFHA). METHODS We analysed baseline ECG parameters from the Guangzhou Biobank Cohort Study, a southern Chinese population-based cohort of older adults. ECG age was developed using machine-learning algorithms and validated in a 30% held-out sample. Associations with all-cause and cardiovascular disease (CVD) mortality were assessed using Cox proportional hazards regression after excluding participants with baseline CVD. Ten-year mortality prediction was evaluated using the area under the receiver operating characteristic curve (AUC) and compared with HFHA. RESULTS Among 17,090 participants (mean age 60.6 years [SD=7.2]; 74.2% women) followed from 2003–08 through 2023, LightGBM showed the best age-prediction performance (mean absolute error=4.91 years; R²=0.302). Higher ECG age was associated with increased risk of all-cause mortality (hazard ratio [HR] per SD 1.67, 95%CI 1.62–1.71) and CVD mortality (1.67, 1.61–1.73). For 10-year mortality prediction, AUCs for ECG age versus HFHA were 0.758 versus 0.755 for all-cause mortality and 0.799 versus 0.806 for CVD mortality. Adding ECG age to HFHA improved discrimination for all-cause mortality (ΔAUC 0.006, 95%CI 0.003–0.010) and CVD mortality (0.010, 0.002–0.017). CONCLUSIONS ECG age was associated with all-cause and cardiovascular mortality and showed mortality discrimination similar to the Heart Foundation Heart Age marker in this internally validated cohort. External validation is required before clinical interpretation or use.