Predicting Heart Failure From 12-Lead ECGs Using AI: A HeartShare/AMP-HF Pooled Cohort Analysis.

Area of Science:

• Cardiology
• Artificial Intelligence in Medicine
• Preventive Cardiology

Background:

• Artificial intelligence applied to electrocardiograms (ECG-AI) offers a scalable method for identifying individuals at risk for heart failure (HF).
• ECG-AI can guide preventive interventions by detecting underlying cardiac dysfunction.

Purpose of the Study:

• To evaluate if ECG-AI, detecting systolic and diastolic dysfunction, improves incident HF prediction beyond the PREVENT-HF clinical risk equation.
• Assessing the enhancement of HF risk prediction using ECG-AI combined with established clinical risk factors.

Main Methods:

• Pooled baseline clinical and ECG data from Framingham Heart Study, MESA, and CHS participants.
• Utilized validated ECG-AI algorithms (ECG-AI LEF for systolic, ECG-AI DD for diastolic dysfunction) and the PREVENT-HF equation.
• Evaluated risk discrimination and reclassification using Harrell's C-statistic and net reclassification improvement.

Main Results:

• Participants with positive composite ECG-AI screens showed a 10- to 20-fold higher risk of developing HF.
• Adding ECG-AI to PREVENT-HF significantly improved net reclassification of HF risk at 1, 3, 5, and 10 years.
• Net reclassification improvements ranged from 0.086–0.125 (PREVENT-HF 10% threshold) and 0.327–0.403 (PREVENT-HF 20% threshold).

Conclusions:

• The integration of ECG-AI with PREVENT-HF enhances the discrimination of near-term heart failure risk.
• ECG-AI holds potential for population-level HF risk stratification.
• This approach may facilitate the implementation of targeted preventive strategies for heart failure.