A sleep study’s heart recording can flag heart risk 10 years ahead
☕ Includes a question for the family ↓
Sleep healthPart of a story we are following · 11 articles · See every update
If you’ve ever done an overnight sleep study, a small sensor was already tracking your heartbeat all night. Researchers found that an AI can read that recording and predict heart trouble up to 10 years ahead.
That’s good news for anyone who has had a sleep test. The heart data is already collected, so spotting risk early would take no extra test. Doctors don’t use this tool in routine care yet, but the results point to a clear way forward.
How the study worked
Researchers started with an AI trained to spot heart rhythm problems. They taught it using overnight heart recordings and sleep-stage data from thousands of patients at Massachusetts General Hospital. Then they tested it on patients at two other hospitals: Emory University Hospital (9,810 people) and Beth Israel Deaconess Medical Center. They tracked heart events over 10 years in medical records. The work was published Oct. 9, 2026, in Sleep. Everyone in it had been referred for a sleep study, and the outcomes came from billing codes in health records.
What they found
The AI’s score was strongly linked to future heart problems. That held even after the researchers accounted for age, sex, weight, smoking, high blood pressure, diabetes and other sleep measures. It held in both testing hospitals.
For each step up in the AI score, here is how much higher the risk was:
- Atrial fibrillation: about double or more (2.03 at Emory and 2.72 at Beth Israel).
- Heart failure: higher in both hospitals (1.69 and 2.33).
- Death from any cause: higher in both hospitals (1.82 and 1.65).
- Stroke: higher in both hospitals (1.19 and 1.40).
For heart attack, the link showed up in one hospital but not the other.
So the sleep recording carries real information about the heart. It picked up signals that age, weight, blood pressure and the usual sleep numbers did not already explain.
Why does a sleep test matter here?
Millions of sleep studies are done every year, and the heart recording is already part of each one. Until now it has rarely been analyzed beyond checking for obvious problems. Using it could mean people get preventive care sooner without anything extra.
What is atrial fibrillation?
It’s an irregular heartbeat. It often goes unnoticed until it causes a stroke, which is why catching it early is so valuable. It is also the heart problem where the AI score showed the strongest link in this study.
What you can do today
If you’ve had a sleep study, log in to your patient portal or call the sleep clinic and ask for a copy of your report. Look for any note about your heart rhythm. If something is written there, bring it to your next checkup so it becomes part of your health story.
Talk about it at home
During a sleep test, a sensor records your heartbeat all night. A computer program learned to spot tiny patterns in that recording. Those patterns showed up years before some people had heart trouble.
Question for the table
How could scientists check that the computer’s guesses are right, and not just lucky?
Keep exploring
Copies this article and its questions so you can paste them into ChatGPT, Claude, Gemini or the AI you use, and keep the conversation going.
Questions and answers
- What did the AI analyze?
- The single-lead heart recording (ECG) made during overnight sleep studies, combined with data on sleep stages.
- What did it predict?
- Risk over 10 years of atrial fibrillation, heart failure, stroke and death. It worked even after accounting for age, weight, smoking, blood pressure and diabetes.
- Can I get this test now?
- Not in routine care yet. The good part is that the heart recording is already collected during sleep studies.
- Where was it tested?
- It was trained at Massachusetts General Hospital and tested at Emory University Hospital and Beth Israel Deaconess Medical Center.





