Sleep recordings reveal five patient groups with different risks
The highest-risk group had more than double the lowest-risk group’s risk of death, while standard categories based on breathing interruptions had limited predictive value.
Edited by Boris Djordjevic · AgingResearch.News
In a retrospective observational cohort study, researchers analyzed 9,608 laboratory sleep recordings from 9,297 Cleveland Clinic patients, linked to medical records. They developed a computer model to learn patterns across signals recording brain activity, breathing, heart activity and blood oxygen. These patterns separated patients into five groups associated with different risks of death, cardiovascular disease and neurological disease.
In the clinical cohort, the highest-risk group had more than double the mortality risk of the lowest-risk group. The mortality analysis accounted for age and gender. Standard severity categories based on breathing interruptions during sleep had limited predictive value. In the independent Sleep Heart Health Study, the model also distinguished high- and low-risk participants despite lower-resolution recordings.
Why it matters
The work addresses whether sleep physiology can help assess future cardiovascular and neurological disease risk, a question relevant to healthy aging.
Caveats
This observational study cannot establish cause and effect or show that using the model improves care. Results from the main clinical cohort may not generalize to people who do not undergo sleep studies.
The paper
A foundation model for sleep-based risk stratification and clinical outcomes
IBM (United States); University of Washington
Nature Communications, Aug 3, 2026, Open access
Comments
Be the first to comment.