AI Model Unlocks Hidden Health Risks in Routine Sleep Studies, Doubles Mortality Prediction Accuracy

AI Foundation Model Transforms Sleep Medicine

A multidisciplinary research team has developed an AI model that unlocks previously hidden health insights from routine sleep study data, according to findings published in Nature Communications1

1

2

2

3

3

. The AI foundation model analyzes polysomnography results to predict long-term health risks including cardiovascular events, cognitive decline, and mortality—revealing prognostic biomarkers that conventional measures completely miss. Developed through the Discovery Accelerator, a 10-year joint research partnership between Cleveland Clinic and IBM, the model demonstrates how AI-driven healthcare can extract substantially more physiologic information from routine medical tests than current clinical practice captures4

4

.

Standard Measures Miss Critical Health Signals

The research exposed a fundamental limitation in sleep medicine: the apnea-hypopnea index, used for decades to assess sleep apnea severity, failed to predict mortality in the study cohort1

1

. "For decades we have distilled an overnight sleep study into a handful of summary measures," said Reena Mehra, M.D., professor of medicine at the University of Washington 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"3

3

. Each year, an estimated 1 to 4 million polysomnograms are performed in the United States, typically to evaluate sleep apnea2

2

. While these studies collect rich data on patients' brains, lungs, muscles, and heart, clinicians have historically focused on a small subset of that information to grade sleep apnea severity4

4

.

Five Patient Subtypes With Distinct Health Trajectories

Using data from the Cleveland Clinic Sleep Signals, Testing, and Reports Linked to Patient Traits (STARLIT) Registry, researchers trained their AI model on 9,608 high-resolution polysomnography studies from 9,297 patients linked to electronic medical records1

1

. The model identified five patient subtypes with markedly different trajectories for mortality, major adverse cardiovascular events, atrial fibrillation, cognitive decline, and epilepsy1

1

. Patients categorized into the highest-risk group showed twice the five-year mortality risk compared to those in the lowest-risk category—a critical distinction invisible to conventional measures2

2

3

3

. The AI-derived risk groups showed strong, graded associations with clinical outcomes even after adjusting for demographics, comorbidities, and the apnea-hypopnea index itself1

1

.

Hidden Physiological Signals Decoded Across Multiple Systems

Unlike traditional machine learning models that depend on task-specific designs and require complete retraining for new tasks, this AI foundation model was trained on large, diverse datasets, allowing it to adapt to new datasets or tasks with minimal additional training1

1

. The model extracts and analyzes hidden physiological signals across neural, ocular, muscular, cardiac, oxygenation, and ventilatory signals from routine sleep study data1

1

. "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 Rogers, Ph.D., the corresponding author and professor adjunct, neurosurgery, Yale School of Medicine3

3

. The approach uses AI to detect latent physiologic features invisible to the human eye and extract prognostic biomarkers that help stratify risk for cardiovascular and neurologic disease4

4

.

Validation and Equal Performance Across Sexes

To validate the model's robustness, researchers applied a simplified two-group version to data from the Sleep Heart Health Study, where it successfully reproduced associations with mortality and heart failure in both men and women1

1

. This represents a significant advance, as the apnea-hypopnea index has historically performed better in men2

2

3

3

. The findings were independently confirmed in a nationwide patient cohort, demonstrating the model's generalizability4

4

.

Implications for Personalized Medical Interventions

"Sleep is foundational to health and wellness," said Matheus Lima Diniz Araujo, Ph.D., a sleep researcher at Cleveland Clinic. "Nearly 70 million Americans live with chronic disorders of sleep and wakefulness, affecting daily functioning and overall health. 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"3

3

. The findings directly contribute to American Thoracic Society priorities for developing scalable, objective measures that go beyond the apnea-hypopnea index to predict cardiovascular and neurological outcomes1

1

. By identifying high-risk patients earlier, the model opens the door to earlier and more personalized medical interventions2

2

.

Next Steps and Broader Applications

Carl Saab, Ph.D., professor of biomedical engineering and Chief Scientist of Cleveland Clinic's Discovery Accelerator, outlined future directions: "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"3

3

. Lead author Erhan Bilal, Ph.D., emphasized the broader potential: "Sleep is increasingly recognized as a critical component of health, yet the physiological information captured during sleep remains largely underused. Because everyone sleeps, sleep studies offer a remarkable window into human health that extends far beyond the diagnosis of sleep disorders. Our work shows how foundation models can begin to unlock the richness of these complex signals. And this is only the beginning"4

4

. The collaborative research team brought together sleep physicians, AI researchers, data scientists, and neuroscientists through the Discovery Accelerator partnership, demonstrating how interdisciplinary approaches can advance sleep medicine and AI-driven healthcare4

4

.