Researchers analyzed more than 10,000 full-night sleep studies using AI and linked the findings with long-term patient outcomes recorded in medical records.
An AI model has demonstrated that routine sleep studies contain valuable clinical information far beyond diagnosing sleep apnea, enabling more accurate prediction of cardiovascular, neurological, and mortality risks.
A new study published in Nature has explored whether artificial intelligence could extract additional health insights from this largely underused information.
“For decades we have distilled an overnight sleep study into a handful of summary measures,” study leader Dr Reena Mehra of the University of Washington said.
“AI gives us the opportunity to move beyond those summaries and learn from the full richness of sleep physiology.”
Researchers analyzed more than 10,000 full-night sleep studies using AI and linked the findings with long-term patient outcomes recorded in medical records.
Instead of focusing solely on breathing interruptions, the model examined multiple physiological signals collected during sleep to identify previously unrecognized health patterns.
The AI system classified patients into five clinically meaningful subtypes, each associated with distinct long-term health outcomes.
Individuals placed in the highest-risk category had twice the likelihood of death within five years compared with those in the lowest-risk group, a difference that was not identified through the conventional apnea-hypopnea index alone.
The highest-risk group also showed 65% higher odds of developing heart failure, 84% higher odds of experiencing a heart attack, 93% higher odds of cognitive impairment, and more than 200% higher odds of both atrial fibrillation and epilepsy when compared with the lowest-risk group.
Researchers also found that the AI model predicted outcomes consistently for both men and women, addressing a known limitation of the apnea-hypopnea index, which has historically performed better in men.
The findings were further validated in a separate nationwide cohort involving more than 6,000 patients.
Stay tuned for more such updates on Digital Health News
