A research team supported by the US National Institutes of Health (NIH) has found that electrocardiograms (ECGs) recorded during sleep, when combined with sleep-stage information, can help identify individuals at higher risk of future cardiovascular events. The study, published in the journal Sleep, used deep learning to assess long-term cardiovascular risk and suggests that data collected during routine sleep studies could support clinical decision-making.

The researchers investigated whether ECG recordings obtained during polysomnography, a standard test for diagnosing sleep disorders such as insomnia and obstructive sleep apnoea, could help predict cardiovascular outcomes over 10 years. Although ECGs are routinely recorded during these tests, they are not commonly analysed to assess long-term cardiovascular risk.

The team developed a deep-learning model using single-lead ECG recordings alongside expert-annotated sleep-stage data. The model was fine-tuned using data from 15,809 patients at Massachusetts General Hospital in Boston. Its performance was then evaluated using two independent datasets comprising 9,810 patients from Emory University Hospital in Atlanta and 12,576 patients from Beth Israel Deaconess Medical Center in Boston. Patient outcomes were obtained from electronic health records.

The model categorised individuals into groups with different levels of long-term cardiovascular risk, with higher scores associated with greater risk. Its predictive value remained after accounting for established risk factors, including age, sex, body mass index, diabetes and hypertension, as well as sleep-related factors such as sleep apnoea severity, arousals and sleep efficiency.

The model demonstrated predictive ability for atrial fibrillation, heart failure and all-cause mortality. However, the researchers said further optimisation is needed to improve its predictions for myocardial infarction and stroke.

According to the research team, adding the model’s risk score improved the prediction of cardiac outcomes beyond established risk factors. The findings indicate that analysing ECG data already collected during sleep studies could help identify at-risk individuals earlier and support more proactive patient care. Further evaluation is needed to determine whether this approach improves conventional cardiovascular risk prediction in clinical practice.

The researchers also noted that the method relies on a single ECG lead rather than the conventional 12-lead setup. This could make data collection simpler and potentially support the use of low-cost wearable patches for overnight monitoring.

David Goff, acting director of the NIH’s National Heart, Lung, and Blood Institute (NHLBI), said the approach could potentially identify people at risk of cardiovascular disease years before symptoms develop. He added that evaluating its performance alongside traditional risk-prediction methods would be an important next step.