AI-ECG age ≥8 years older than chronological age predicts 45% higher mortality risk; ≥8 years younger predicts 12% lower risk in cardiovascular and acute care patients.
Does AI-ECG-derived biological age predict all-cause mortality in cardiovascular and acute care patients?
AI-ECG-derived biological age significantly improves mortality risk stratification beyond chronological age in cardiovascular and acute care patients.
Absolute Event Rate: 0% vs 0%
Abstract Background Artificial intelligence (AI) algorithms applied to standard 12-lead ECGs offer promising methods for biological age prediction which may provide insights beyond chronological age (1). While prior research has demonstrated an age-independent association between ECG age and mortality in general populations (2), its prognostic value in patients with established cardiovascular disease (CVD) or in those seeking acute medical care remains unknown. Methods We applied a validated open-source AI algorithm, trained on 1,558,415 patients, to our ECG database containing 234,945 ECGs from 74,175 unselected patients treated at our cardiology inpatient, outpatient clinics, and medical emergency department between 2000 and 2021. After excluding 24,879 patients due to insufficient ECG quality or missing clinical information, we analyzed 48,950 patients. Patients were classified into three groups based on age deviation between AI-ECG age and chronological age (Δ-age): ≥8 years younger, ≥8 years older, and within ±8 years. Cox proportional hazards models, adjusted for age and gender, examined associations with all-cause mortality. Results Overall, AI-ECG age was strongly correlated with chronological age (r = 0.72, Figure 1). In the total cohort, patients with negative Δ-age (estimated to be ≥8 years younger) had lower mortality risk (HR: 0.88, CI: 0.84–0.92, p 0.001), while those with positive Δ-age (≥8 years older) had higher risk (HR: 1.45, CI: 1.37–1.52, p 0.001) (Figure 2). These findings were consistent across subcohorts. A Cox model including age, gender and Δ-age achieved AUC of 0.80 (CI: 0.80-0.81, p 0.001) in 10-year ROC analysis. Net reclassification index (NRI) analysis showed 19.4% improved risk classification (p 0.001) when including Δ-age. Conclusion AI-ECG-estimated biological age provides valuable prognostic insights for patients with cardiovascular diseases and those admitted to a medical emergency department. Its integration into clinical practice could enhance risk stratification and guide personalized treatment strategies in the future.Scatterplot of AI-ECG age vs. chron. age Age- and gender-adjusted survival curves
Pavluk et al. (Sat,) reported a other. AI-ECG age ≥8 years older than chronological age predicts 45% higher mortality risk; ≥8 years younger predicts 12% lower risk in cardiovascular and acute care patients.