Heart health
Sleep-study ECG data may help flag future heart risk
An NIH-supported study found that ECG data already collected in sleep studies may help estimate later heart risk, pending further validation.
AI-assisted summary and sourcing. Official and research sources are linked below; this article is informational and is not personal medical advice.
CareNews
Key takeaways
- Researchers analyzed 38,195 patients who received hospital-based sleep studies at three U.S. centers.
- A higher model score added predictive information for atrial fibrillation, heart failure, and all-cause death after adjustment for common risk factors.
- The model did not add clinical benefit for stroke or heart-attack prediction and still needs prospective validation.
A new use for an overnight recording
A polysomnogram commonly records a single ECG channel while examining sleep disorders. NIH-supported researchers asked whether that existing heart signal could provide additional clues about later cardiovascular risk.
They paired overnight ECG recordings with professionally scored sleep stages in a deep-learning model. Massachusetts General Hospital records were used for development, then separate patient groups at Emory and Beth Israel Deaconess were used for testing.
Sources for this section: NIH News Release – Deep learning model using ECGs during sleep studies can predict cardiovascular outcomes, Sleep – Prediction of cardiovascular outcomes using electrocardiography from overnight polysomnography
What the record-based study found
The analysis included 38,195 hospital sleep-study patients. In the external test groups, higher model scores were associated with greater long-term risk of atrial fibrillation, stroke, heart attack, heart failure, and death from any cause after adjustment for common clinical and sleep factors.
Added prediction was clearest for atrial fibrillation, heart failure, and all-cause death. The model did not substantially improve prediction of stroke or heart attack beyond established clinical and sleep-related information. An association with later events is not a proven screening test.
Sources for this section: NIH News Release – Deep learning model using ECGs during sleep studies can predict cardiovascular outcomes, Sleep – Prediction of cardiovascular outcomes using electrocardiography from overnight polysomnography
Promising research, not routine care
If further validated, the approach might draw cardiovascular-risk information from a heart signal already recorded during a sleep study. It could complement, rather than replace, established assessment for people receiving this type of hospital testing.
The study cannot show that using the model prevents illness or improves outcomes. It analyzed past records from a specialized hospital sleep-study population, so prospective research is required before it could be used routinely.
Sources for this section: NIH News Release – Deep learning model using ECGs during sleep studies can predict cardiovascular outcomes, Sleep – Prediction of cardiovascular outcomes using electrocardiography from overnight polysomnography
Sources