AI Model Uncovers Hidden Health Risks in Routine Sleep Study Data, Doubling Mortality Prediction Accuracy
Key Insights
A novel AI model developed through the Cleveland Clinic-IBM Discovery Accelerator identifies hidden sleep patterns linked to heart disease (search), cognitive decline (search), and death from routine polysomnography data.
Patients in the highest-risk group had twice the five-year mortality risk compared to the lowest-risk group, a distinction not captured by the standard apnea-hypopnea index.
The model performed well for both men and women, addressing a known limitation of the apnea-hypopnea index which historically performs better in men.
An artificial intelligence model can extract previously unrecognized health risk information from standard overnight sleep studies, revealing patient subtypes with sharply different long-term outcomes that conventional clinical measures fail to detect, according to research published today in Nature Communications.
Developed through a multidisciplinary collaboration of sleep physicians, AI researchers, data scientists, and neuroscientists, the model identified hidden sleep patterns linked to higher odds of heart disease (search), cognitive decline (search), and death. Patients classified into the highest-risk group had twice the mortality risk over the next five years compared to those in the lowest-risk group — a distinction not captured by the apnea-hypopnea index, the standard clinical measure used to assess sleep apnea (search) severity.
"For decades we have distilled an overnight sleep study into a handful of summary measures," said Dr. Reena Mehra, professor of medicine at the University of Washington School of Medicine and the study's senior clinical author. "AI gives us the opportunity to move beyond those summaries and learn from the full richness of sleep physiology."
Leveraging the Full Richness of Sleep Physiology
Each year, an estimated 1 to 4 million polysomnograms are performed in sleep labs across the United States, typically to evaluate sleep apnea (search). While these studies collect rich data on each patient's brain activity, respiration, muscle function, and cardiac signals, clinicians have historically narrowed their focus to a small subset of that information to grade sleep apnea severity.
The research team, brought together through the Discovery Accelerator — a 10-year research partnership between Cleveland Clinic and IBM aimed at accelerating discovery in life sciences through AI and quantum computing — sought to change that paradigm.
Using data from the Cleveland Clinic Sleep Signals, Testing, and Reports Linked to Patient Traits (STARLIT) registry, the researchers grouped patients into five distinct risk categories based on physiological patterns detected by the AI model. The findings were independently confirmed in a nationwide patient cohort.
"Modern AI lets us recover much more of the information contained in a night's worth of sleep physiology, revealing clinically meaningful patient groups with very different long-term health risks," said Jeffrey L. Rogers, Ph.D., corresponding author, global research leader at IBM, and adjunct neurosurgery professor at the Yale School of Medicine.
Addressing Sex-Based Disparities in Risk Assessment
Notably, the AI model predicted outcomes well for both men and women, addressing a known limitation of the apnea-hypopnea index, which has historically performed better in men. This finding suggests the model may offer more equitable risk stratification across sexes.
The model detects latent physiological features invisible to the human eye and extracts prognostic biomarkers that help stratify risk for cardiovascular and neurologic disease and survival, opening the door to earlier and more personalized care.
"Nearly 70 million Americans live with chronic disorders of sleep and wakefulness, affecting daily functioning and overall health," said Matheus Lima Diniz Araujo, Ph.D., a sleep researcher at Cleveland Clinic. "This discovery offers a more personalized approach to sleep medicine, by potentially expanding the value of routine sleep testing and reinforcing the key role sleep plays in chronic disease."
Broader Implications for Routine Medical Testing
The findings carry implications beyond sleep medicine. The research suggests that other routine medical tests may contain substantially more physiological information than is currently being extracted in clinical practice.
"Sleep is increasingly recognized as a critical component of health, yet the physiological information captured during sleep remains largely underused," said Erhan Bilal, Ph.D., founder of Enkira, previously an IBM researcher and lead author of the study. "Because everyone sleeps, sleep studies offer a remarkable window into human health that extends far beyond the diagnosis of sleep disorders."
Dr. Mehra noted that as these methods continue to be validated in prospective studies, they have the potential to transform the sleep study "from primarily a diagnostic test into a richer source of information about an individual's future health and may accelerate discoveries about the relationships between sleep physiology and chronic disease."
Carl Saab, Ph.D., professor of biomedical engineering and Chief Scientist of Cleveland Clinic's Discovery Accelerator, emphasized the path forward: "The next step is to validate these findings in diverse populations and expand collaborations among medical and technical experts, industry partners and professional society stakeholders."
The study was supported by the Cleveland Clinic-IBM Discovery Accelerator Program and a grant from the National Heart Lung and Blood Institute (1R21HL170206-01).