Researchers have developed an artificial intelligence model that analyzes routine polysomnography data to predict long-term health risks. This model, utilizing data from the Cleveland Clinic’s STARLIT registry and a national cohort, identified five distinct patient risk subtypes with significantly different health outcomes. Patients categorized in the highest-risk group faced double the five-year mortality rate compared to those in the lowest-risk group, a finding not evident through traditional sleep apnea diagnostic methods.
Read More
The AI model effectively stratified patients into five risk categories, demonstrating that individuals in the highest tier experienced a 100 percent increase in five-year mortality risk compared to those in the lowest tier. This advancement outperformed the traditional Apnea-Hypopnea Index (AHI), capturing latent physiological features across brain, lung, muscle, and cardiac signals that standard metrics overlook. Notably, the new model provides equal predictive accuracy for cardiovascular, neurological, and mortality outcomes in both men and women, contrasting with the AHI’s historical bias favoring male populations.
Each year, between 1 to 4 million polysomnographies are conducted in the United States, primarily to assess sleep apnea. These tests generate extensive physiological data, yet clinicians have historically relied on a limited set of measures to evaluate sleep apnea severity. Reena Mehra, M.D., a professor at the University of Washington and the study's senior author, emphasized that AI allows for a more comprehensive understanding of sleep physiology beyond conventional summaries.
Developed by a multidisciplinary team as part of a ten-year research partnership between Cleveland Clinic and IBM, the model also successfully classified patient risk across an independent nationwide cohort. Jeffrey Rogers, Ph.D., the study’s corresponding author from Yale School of Medicine, stated that modern AI methods can uncover significant insights from a night’s sleep data, distinguishing clinically relevant patient groups with diverse long-term health risks.
Matheus Lima Diniz Araujo, Ph.D., a sleep researcher at Cleveland Clinic, highlighted the importance of sleep health, noting that approximately 70 million Americans suffer from chronic sleep disorders, affecting their well-being. The research signifies a shift towards utilizing routine sleep testing for more personalized care options.
The study's authors, including Erhan Bilal, Ph.D., and Carl Saab, Ph.D., acknowledged the necessity for further validation across diverse populations and collaboration among medical and technical experts to expand the application of this AI tool in clinical practice. As highlighted by Erhan Bilal, sleep studies offer critical insights into overall health that go beyond diagnosing sleep disorders.