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

AI sleep model reveals health risks missed by standard apnea scores

By Dr. Sanchari Sinha Dutta, Ph.D.Reviewed by Lauren HardakerAug 6 2026 By decoding neural, cardiac, respiratory, and other signals hidden within routine sleep studies, the model identified high-risk patients that conventional apnea measurements failed to distinguish. Paper: A foundation model for sleep-based risk stratification and clinical outcomes. Image Credit: F01 PHOTO / Shutterstock Want to read later? Download your PDF copy by clicking here. Researchers have developed an artificial intelligence (AI)-based model that can extract hidden physiological signals from routine sleep test data to stratify patients according to long-term health risks. The study is published in the journal Nature Communications. Background Sleep is a regulated biological state essential for physical and mental well-being. Disrupted sleep can substantially affect quality of life and increase the risk of cardiovascular, neurologic, and psychiatric disease. Sleep disorders are highly prevalent worldwide. They represent a substantial global health burden. Sleep apnea is one of the most common sleep disorders and affects nearly one billion adults worldwide. Polysomnography is the gold standard test for diagnosing sleep disorders. However, the clinical interpretation of this test is often reduced to single summary measures of limited prognostic value, such as the apnea-hypopnea index (AHI). This index captures only limited information on sleep physiology and thus cannot fully evaluate sleep integrity. A multidisciplinary research team recently developed an AI-based model that can decode hidden sleep patterns associated with cardiovascular disease, cognitive impairment, and mortality risk. Model design The researchers developed an AI foundation model that can analyze full-night polysomnography results to identify hidden physiological patterns associated with long-term clinical outcomes. Unlike traditional machine learning models, which often depend on task-specific designs and require complete retraining for new tasks, foundation models are broad, general-purpose models trained on large, diverse datasets. This methodology allows these models to adapt to new datasets or tasks with minimal additional training. However, applying this methodology to polysomnography data is challenging, as sleep recordings are large, noisy, and highly individual-specific. These challenges highlight the need for developing innovative methods that can transform physiological signals recorded at different frequencies into uniform tokens for machine learning models. Here, researchers trained their foundation model using a unique resource of 10,000 high-resolution polysomnography studies from the Cleveland Clinic Sleep Signals, Testing, and Reports Linked to Patient Traits (STARLIT) Registry, which were linked to electronic medical records. After quality control, 9,608 studies from 9,297 patients were retained for clustering analyses. Key findings and significance By extracting and analyzing hidden physiological characteristics across neural, ocular, muscular, cardiac, oxygenation, and ventilatory signals from routine sleep recordings, the foundation model identified five embedding-derived patient groups with markedly different trajectories for mortality, major adverse cardiovascular events, atrial fibrillation, cognitive impairment, and epilepsy. Patients categorized into the highest-risk group showed more than double the mortality risk compared to those in the lowest-risk category. In contrast, the traditional AHI severity categories failed to predict mortality. For decades, the AHI has been widely used to define sleep-disordered breathing severity. Yet, the index failed to detect any association with mortality in the study cohort, indicating insufficient ability to stratify clinically meaningful risks. The AI-derived risk groups, on the other hand, showed strong, graded associations with clinical outcomes, even after adjusting for demographics, comorbidities, and the AHI itself. These findings suggest that the foundation model can capture more detailed pathophysiological signals of sleep associated with clinical outcomes beyond those captured by airway obstruction alone. The American Thoracic Society (ATS) has prioritized developing scalable, objective measures that go beyond the AHI to predict cardiovascular and neurological outcomes. The current findings directly contribute to these priorities by providing a scalable, externally validated framework that uncovers hidden physiologic signals with clear clinical implications. To further validate the model's robustness, researchers applied a simplified two-group version of the stratification approach to data obtained from the Sleep Heart Health Study (SHHS). The approach reproduced associations with mortality and heart failure in both men and women. In contrast, earlier SHHS analyses linked severe sleep-disordered breathing to heart failure only in men and found mortality associations primarily among middle-aged men with severe disease. The consistency of results across independent cohorts using different polysomnography protocols supports the approach's potential generalizability. Notably, SHHS participants in the high-risk group exhibited markedly short total sleep time, similar to the marked physiological disruptions observed in the highest-risk cluster in the STARLIT cohort, further supporting the biological plausibility and clinical relevance. The major strength of the study is the integration of multimodal polysomnographic signals with longitudinal electronic medical records from a large, demographically diverse clinical cohort. The linked records provided an average total clinical observation window of 14.5 years, including medical history before and follow-up after polysomnography. This integration improved the richness of the clinical data and allowed more comprehensive risk stratification than either source alone. The foundation model, however, was trained using technician-supervised objectives, including sleep stages, respiratory events, and oxygen desaturations. A possible influence of these objectives on the learned embedding cannot be fully ignored. Future research exploring self-supervised and disease-targeted objectives is therefore needed to reduce dependence on technician-defined labels. Other limitations include unavailable objective data on positive airway pressure adherence, incomplete medication records, possible residual confounding, and reliance on electronic health record diagnostic codes. The retrospective design also prevents causal conclusions, and further external validation and prospective clinical trials will be required before the model can be implemented in healthcare settings. Such studies will also need to determine whether the approach can accurately guide risk assessment and improve outcomes for individual patients.

AI Analyzes Sleep Data to Predict Cognitive Decline and Health

Summary: Researchers developed an artificial intelligence model capable of extracting hidden physiological signals from routine polysomnography data to predict long-term health risks. Analyzing data from Cleveland Clinic's STARLIT registry alongside a nationwide cohort, the AI model identified five distinct patient risk subtypes with vastly different health trajectories. Patients categorized into the highest-risk group faced double the five-year mortality risk compared to the lowest-risk group, a prognostic distinction invisible to conventional sleep apnea diagnostic measures. Key Facts * Two-Fold Mortality Risk Separation: The AI model stratified patients into five distinct risk categories, showing that individuals in the highest-risk tier had a 100 percent increase in five-year mortality risk compared to those in the lowest-risk tier. * Superiority Over Conventional AHI: The prognostic risk stratification succeeded where the traditional Apnea-Hypopnea Index (AHI) failed, capturing latent physiological features across brain, lung, muscle, and cardiac signals that standard metrics miss. * Sex-Balanced Predictive Accuracy: While the traditional AHI metric historically performs better in male populations, the new AI foundation model predicted cardiovascular, neurological, and mortality outcomes with equal high accuracy across both men and women. * Unlocking Underutilized Clinical Data: Demonstrates that routine polysomnograms, 1 to 4 million of which are conducted annually in the United States, contain vast amounts of unused prognostic data capable of driving early preventative healthcare. Source: Cleveland Clinic A novel AI model can use information collected during routine sleep studies to identify patients' long-term health risks, according to a new study published in Nature Communications. Developed by a multidisciplinary research team, the model uncovered hidden sleep patterns linked to risks including heart disease, cognitive decline and death. The findings also suggest that routine medical tests may contain substantially more physiologic information than current clinical practice extracts from them. In this case, AI identified meaningful signals in standard overnight sleep study data that are not captured by conventional summary measures alone. The research revealed clinically meaningful patient subtypes with sharply different long-term health risks. Patients in the highest-risk group had twice the mortality risk over the next five years compared to those in the lowest-risk group, a distinction that was not captured by the standard clinical measure used to assess sleep apnea severity, the apnea-hypopnea index. Each year, an estimated 1 to 4 million polysomnograms, or in-lab sleep studies, are performed in the United States, typically to evaluate sleep apnea. While these studies collect rich data on each patient's brains, lungs, muscles and heart, clinicians historically have focused on a small subset of that information to grade sleep apnea severity. "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." The model was developed by a collaborative team of sleep physicians, AI researchers, data scientists and neuroscientists brought together through the Discovery Accelerator, a 10-year joint research partnership between Cleveland Clinic and IBM aimed at advancing the pace of discovery in life sciences through AI and quantum computing. Using data from the Cleveland Clinic Sleep Signals, Testing, and Reports Linked to Patient Traits (STARLIT) registry, the researchers grouped patients into five risk categories. The model also predicted outcomes well for men and women, while the apnea hypopnea index has historically performed better in men. 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 Rogers, Ph.D., the corresponding author and professor adjunct, neurosurgery, Yale School of Medicine. "These findings demonstrate that routine medical tests can contain substantially more physiologic information than current clinical practice extracts from them." The model could also help researchers better understand how sleep impacts health outcomes. By looking beyond traditional measures, 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 disease, and survival, opening the door to earlier and more personalized care. "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." Carl Saab, Ph.D., a professor of biomedical engineering and Chief Scientist of Cleveland Clinic's Discovery Accelerator, said, "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." "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., 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. Our work shows how foundation models can begin to unlock the richness of these complex signals. And this is only the beginning." The research team included Erhan Bilal, Ph.D.; Matheus Lima Diniz Araujo, Ph.D.; Kristen Beck, Ph.D.; Catherine Heinzinger, D.O.; Samer Ghosn, B.S.; Nancy Foldvary-Schaefer, D.O.; Carl Saab, Ph.D.; Jeffrey Rogers, Ph.D.; and Reena Mehra, M.D. Key Questions Answered: Editorial Notes: * This article was edited by a Neuroscience News editor. * Journal paper reviewed in full. * Additional context added by our staff. About this AI and sleep research news Author: Alicia Reale Source: Cleveland Clinic Contact: Alicia Reale - Cleveland Clinic Image: The image is credited to Neuroscience News Original Research: Open access. "A foundation model for sleep-based risk stratification and clinical outcomes" by Erhan Bilal, Matheus Lima Diniz Araujo, Kristen L. Beck, Catherine M. Heinzinger, Samer Ghosn, Carl Y. Saab, Nancy Foldvary-Schaefer, Jeffrey L. Rogers & Reena Mehra. Nature Communications DOI:10.1038/s41467-026-75326-9 Abstract A foundation model for sleep-based risk stratification and clinical outcomes Clinical sleep studies capture multiple physiologic signals, yet interpretation is often reduced to single summary measures of limited prognostic value, such as the apnea-hypopnea index. We present a foundation model that learns rich representations of sleep physiology from more than 10,000 clinical sleep recordings linked to electronic medical records. Here we show that sleep physiology contains latent risk structure invisible to conventional metrics, identifying five patient risk groups with markedly different trajectories for mortality, cardiovascular, and neurological disease. The highest-risk group shows more than double the mortality risk of the lowest, whereas apnea-hypopnea index severity categories show limited predictive value. The framework generalizes to the independent Sleep Heart Health Study, distinguishing high- and low-risk patients despite lower-resolution data. We demonstrate that foundation models recover clinically meaningful risk information embedded in routine sleep recordings that conventional metrics systematically miss, providing a scalable path to precision sleep medicine.

AI model can use routine sleep study data to identify patients' long-term health risks

A novel AI model can use information collected during routine sleep studies to identify patients' long-term health risks, according to a new study published in Nature Communications. Developed by a multidisciplinary research team, the model uncovered hidden sleep patterns linked to risks including heart disease, cognitive decline and death. The findings also suggest that routine medical tests may contain substantially more physiologic information than current clinical practice extracts from them. In this case, AI identified meaningful signals in standard overnight sleep study data that are not captured by conventional summary measures alone. The research revealed clinically meaningful patient subtypes with sharply different long-term health risks. Patients in the highest-risk group had twice the mortality risk over the next five years compared to those in the lowest-risk group, a distinction that was not captured by the standard clinical measure used to assess sleep apnea severity, the apnea-hypopnea index. Each year, an estimated 1 to 4 million polysomnograms, or in-lab sleep studies, are performed in the United States, typically to evaluate sleep apnea. While these studies collect rich data on each patient's brains, lungs, muscles and heart, clinicians historically have focused on a small subset of that information to grade sleep apnea severity. For decades we have distilled an overnight sleep study into a handful of summary measures. AI gives us the opportunity to move beyond those summaries and learn from the full richness of sleep physiology." Reena Mehra, M.D., professor of medicine, University of Washington and study's senior clinical author The model was developed by a collaborative team of sleep physicians, AI researchers, data scientists and neuroscientists brought together through the Discovery Accelerator, a 10-year joint research partnership between Cleveland Clinic and IBM aimed at advancing the pace of discovery in life sciences through AI and quantum computing. Using data from the Cleveland Clinic Sleep Signals, Testing, and Reports Linked to Patient Traits (STARLIT) registry, the researchers grouped patients into five risk categories. The model also predicted outcomes well for men and women, while the apnea hypopnea index has historically performed better in men. 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 Rogers, Ph.D., the corresponding author and professor adjunct, neurosurgery, Yale School of Medicine. "These findings demonstrate that routine medical tests can contain substantially more physiologic information than current clinical practice extracts from them." The model could also help researchers better understand how sleep impacts health outcomes. By looking beyond traditional measures, 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 disease, and survival, opening the door to earlier and more personalized care. "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." Carl Saab, Ph.D., a professor of biomedical engineering and Chief Scientist of Cleveland Clinic's Discovery Accelerator, said, "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." "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., 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. Our work shows how foundation models can begin to unlock the richness of these complex signals. And this is only the beginning." Source: Cleveland Clinic Journal reference: Bilal, E., et al. (2026). A foundation model for sleep-based risk stratification and clinical outcomes. Nature Communications. DOI: 10.1038/s41467-026-75326-9. https://www.nature.com/articles/s41467-026-75326-9

AI Identifies Previously Unrecognized Health Insights in Routine Sleep Studies | Newswise

Interdisciplinary research team develops foundation model that identifies patient groups with markedly different long-term health risks Newswise -- August 3, 2026, 5:00 AM EDT: A novel AI model can use information collected during routine sleep studies to identify patients' long-term health risks, according to a new study published in Nature Communications. Developed by a multidisciplinary research team that included Cleveland Clinic, the model uncovered hidden sleep patterns linked to risks including heart disease, cognitive decline and death. The findings also suggest that routine medical tests may contain substantially more physiologic information than current clinical practice extracts from them. In this case, AI identified meaningful signals in standard overnight sleep study data that are not captured by conventional summary measures alone. The research revealed clinically meaningful patient subtypes with sharply different long-term health risks. Patients in the highest-risk group had twice the mortality risk over the next five years compared to those in the lowest-risk group, a distinction that was not captured by the standard clinical measure used to assess sleep apnea severity, the apnea-hypopnea index. Each year, an estimated 1 to 4 million polysomnograms, or in-lab sleep studies, are performed in the United States, typically to evaluate sleep apnea. While these studies collect rich data on each patient's brains, lungs, muscles and heart, clinicians historically have focused on a small subset of that information to grade sleep apnea severity. "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." The model was developed by a collaborative team of sleep physicians, AI researchers, data scientists and neuroscientists brought together through the Discovery Accelerator, a 10-year joint research partnership between Cleveland Clinic and IBM aimed at advancing the pace of discovery in life sciences through AI and quantum computing. Using data from the Cleveland Clinic Sleep Signals, Testing, and Reports Linked to Patient Traits (STARLIT) registry, the researchers grouped patients into five risk categories. The model also predicted outcomes well for men and women, while the apnea hypopnea index has historically performed better in men. 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 Rogers, Ph.D., the corresponding author and professor adjunct, neurosurgery, Yale School of Medicine. "These findings demonstrate that routine medical tests can contain substantially more physiologic information than current clinical practice extracts from them." The model could also help researchers better understand how sleep impacts health outcomes. By looking beyond traditional measures, 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 disease, and survival, opening the door to earlier and more personalized care. "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." Carl Saab, Ph.D., a professor of biomedical engineering and Chief Scientist of Cleveland Clinic's Discovery Accelerator, said, "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." "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., 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. Our work shows how foundation models can begin to unlock the richness of these complex signals. And this is only the beginning." The research team included Erhan Bilal, Ph.D.; Matheus Lima Diniz Araujo, Ph.D.; Kristen Beck, Ph.D.; Catherine Heinzinger, D.O.; Samer Ghosn, B.S.; Nancy Foldvary-Schaefer, D.O.; Carl Saab, Ph.D.; Jeffrey Rogers, Ph.D.; and Reena Mehra, M.D.