AI uncovers hidden health risks in sleep study data

Researchers used artificial intelligence to analyze more than 10,000 overnight polysomnograms and identify five distinct patient subtypes with markedly different long‑term health outcomes. The highest‑risk group faced twice the odds of death within five years and substantially higher odds of heart failure, heart attack, cognitive impairment, atrial fibrillation and epilepsy compared with the lowest‑risk group—risks not captured by the traditional apnea‑hypopnea index. The AI model performed well for both men and women and was validated in an independent cohort of over 6,000 patients. The study, led by Dr. Reena Mehra of the University of Washington, was published in *Nature*. The findings suggest that AI can extract richer physiological information from routine sleep studies, potentially enabling earlier detection of serious conditions and more personalized care.