This study explores the use of a noise-adapted artificial intelligence (AI) algorithm applied to single-lead ECGs to predict the future risk of heart failure (HF) in individuals without a prior HF diagnosis.
Using data from over 248,000 individuals across three large cohorts in the U.S., U.K., and Brazil, the researchers found that a positive AI-ECG result was associated with a 3- to 7-fold increased risk of developing HF.
Notably, the AI model retained its predictive accuracy even when the ECG data was noisy—mirroring the kind of signals one might expect from portable or wearable ECG devices.
The algorithm's performance was at least comparable, and often superior, to two well-established clinical risk prediction tools: the PCP-HF and PREVENT scores.
The study also observed a dose-dependent relationship, where each 0.1 increase in the model's risk probability translated into a 27% to 65% higher chance of future HF.
Importantly, these associations remained consistent across geographic, demographic, and clinical subgroups, and even when adjusting for other cardiovascular risk factors and competing risks like death.
The authors propose that such a tool could be a scalable and accessible strategy for early identification of individuals at high risk for HF—especially in low-resource or non-clinical settings—if validated further in prospective, real-world scenarios.
【MORE】