Risk stratification of coronary artery bypass patients using an artificial intelligence electrocardiogram-derived age

To assess the prognostic value of artificial intelligence electrocardiogram-derived (AI ECG) age in predicting outcomes following isolated CABG. We used preoperative ECGs (within 30 days from surgery) from 13,808 isolated CABG patients to calculate AI ECG-derived age using convolutional neural networks. The age gap was calculated as AI ECG-derived age minus chronological age. The associations of age gap with baseline comorbidities, operative outcomes, and long-term survival were analyzed using multivariable regression models. Median chronological age was 68 (60-74) years, and the AI ECG-derived age was 67 (61-72) years. Mean age gap was -1±8 year (range -39 to +35 years). In 44% of patients, the AI ECG-derived age was older than the chronological age (positive age gap of 6(±5) years), and in 21.4% this gap was greater than 5 years. Patients with age gap>5 years were more likely to have renal failure, congestive heart failure, history of myocardial infarction, higher BMI, and lower ejection fraction. Postoperatively, they were at higher risk of atrial fibrillation (OR=1.15(95% confidence interval [95% CI] 1.05-1.30), p=0.042), prolonged ventilation (OR=1.2 (95% CI 1.1-1.5), p=0.047), blood transfusion (OR=1.15(95% CI 1.1-1.30), p=0.017), postoperative creatinine(B-coefficient=+0.15 units, p<0.001), and hospital stay (B-coefficient=+0.7 days, p=0.001). Long-term survival was lower in patients with age gap>5 years compared to those with gap≤5 (Hazard Ratio=1.4(95% CI 1.2-1.5), p<0.001). Higher age gap identifies a cohort of CABG patients with increased comorbidities and advanced physiological age. Advanced physiological age identified by AI ECG is independently associated with worse operative outcomes and long-term mortality.