Sleep ECG could predict heart failure risk

A recent study shows that the way the heart behaves during sleep can reveal a person's risk of serious heart conditions up to ten years ahead. Researchers used a single ECG lead placed on the chest while patients slept and fed the data into an artificial‑intelligence model. The model could predict future irregular heartbeats, strokes, heart failure, or death from any cause.

The team first trained the deep‑learning system on nearly 16,000 patients who had overnight ECG recordings as part of a Massachusetts General Hospital sleep study. They then tested it on two independent cohorts—about 10,000 patients from Emory University Hospital and 13,000 from Beth Israel Deaconess Medical Center—finding that each standard‑deviation increase in the AI risk score raised the risk of atrial fibrillation by 103–172 %, stroke by 19–40 %, heart failure by 69–133 %, and all‑cause mortality by 65–82 %. Only myocardial infarction risk increased in one group.

Adding the AI-derived score to existing risk models improved prediction accuracy. The study reports up to a 6 % gain for atrial fibrillation, 9 % for heart failure and 7 % for mortality, even after accounting for age, sex, diabetes, hypertension, smoking and sleep stage. The most informative signals came from the lightest sleep stage, where rapid autonomic changes amplify subtle heart‑rate irregularities.

Proactive cardiovascular screening

These findings suggest that sleep‑study ECG data, which are already collected for diagnosing insomnia and other disorders, could double as a screening tool for cardiovascular disease. If incorporated into routine polysomnography, clinicians could identify patients at elevated risk and intervene earlier. The researchers say further work is needed to assess whether this approach improves long‑term outcomes.