AI analysis of sleep data identifies hidden health risks, study finds

WASHINGTON, United States — Researchers at Washington University have used artificial intelligence to analyze routine sleep study data and identify health risks that conventional methods miss, according to a study published in Nature. The team fed more than 10,000 sleep studies alongside patients' long-term medical records into an AI model, which identified five patient subtypes with differing risk levels. Patients in the highest-risk group were twice as likely to die within five years compared with those in the lowest-risk group. The same group showed a 65% higher risk of heart failure, an 84% higher risk of heart attack and a 93% higher risk of cognitive decline, while the risks of atrial fibrillation and epilepsy rose by more than 200%. The standard apnea index used in clinical practice did not capture these distinctions. The model produced consistent results for both men and women, and its findings were validated in an independent database of more than 6,000 patients. Prof. Rina Mehra, who led the research team, said sleep studies have for decades relied on a limited set of indicators, and that artificial intelligence makes it possible to use the full range of physiological data collected during such studies.