Deep neural network-estimated electrocardiographic age as a mortality predictor

Area of Science:

• Cardiology
• Artificial Intelligence
• Medical Diagnostics

Background:

• The electrocardiogram (ECG) is a primary tool for assessing cardiovascular diseases.
• Existing ECG analysis primarily focuses on detecting abnormalities rather than overall cardiovascular health.
• The potential of AI to extract novel prognostic information from ECG data is underexplored.

Purpose of the Study:

• To investigate if artificial intelligence (AI)-predicted age from ECG (ECG-age) can serve as a measure of cardiovascular health.
• To determine the association between ECG-age and all-cause mortality.
• To validate the prognostic value of ECG-age in diverse patient cohorts.

Main Methods:

• A deep neural network was trained to predict patient age from 12-lead ECG data.
• The model was trained on the large CODE study cohort (n=1,558,415) and validated on external cohorts (ELSA-Brasil and SaMi-Trop).
• The difference between ECG-age and chronological age (ECG-age gap) was analyzed for its association with mortality.

Main Results:

• Patients with an ECG-age >8 years older than chronological age showed significantly higher mortality (HR 1.79, p<0.001).
• Patients with an ECG-age >8 years younger than chronological age exhibited significantly lower mortality (HR 0.78, p<0.001).
• The ECG-age gap remained a significant predictor of mortality even in individuals with outwardly normal ECGs.

Conclusions:

• AI-derived ECG-age is a robust predictor of cardiovascular health and mortality risk.
• ECG-age provides prognostic information beyond traditional ECG interpretation and patient chronology.
• AI-enhanced ECG analysis represents a promising advancement in cardiovascular risk stratification.