Artificial Intelligence-Electrocardiography to Predict Incident Atrial Fibrillation: A Population-Based Study

Area of Science:

• Cardiology
• Artificial Intelligence
• Predictive Analytics

Background:

• Artificial intelligence (AI) algorithms applied to electrocardiography (ECG) can detect concurrent atrial fibrillation (AF).
• The potential of AI-enabled ECG (AI-ECG) to predict future AF requires characterization.
• Performance comparison with existing risk scores like CHARGE-AF is needed in population-based samples.

Purpose of the Study:

• To evaluate AI-ECG as a predictor of future AF.
• To compare the performance of AI-ECG with the CHARGE-AF score.
• To assess the combined predictive value of AI-ECG and CHARGE-AF.

Main Methods:

• Utilized AI-ECG to calculate AF probability in participants without prior AF history from the Mayo Clinic Study of Aging.
• Employed Cox proportional hazards models to determine independent prognostic value and interactions between AI-ECG and CHARGE-AF score.
• Calculated C statistics for AI-ECG, CHARGE-AF score, and their combination.

Main Results:

• 1936 participants (median age 75.8 years) were analyzed.
• AI-ECG AF model output >0.5 predicted cumulative AF incidence of 21.5% at 2 years and 52.2% at 10 years.
• Both AI-ECG (HR 1.76) and CHARGE-AF (HR 1.90) independently predicted future AF; combined C statistic was 0.72.

Conclusions:

• AI-ECG model output and CHARGE-AF score independently predict incident AF.
• AI-ECG offers a potential single-test method for risk assessment.
• This AI-ECG approach bypasses the need for manual or automated clinical data abstraction.