AI Model Predicts Health Risks Based on Sleep Study Data

The study indicates that routine medical tests might provide more significant insights than are currently utilized in clinical settings. The AI system detected signals in standard overnight sleep study data that aren't captured by conventional metrics.

Researchers found distinct patient subtypes with varying long-term health risks. Specifically, individuals in the highest-risk category had double the mortality risk over five years compared to those in the lowest-risk group, a distinction not evident from the usual apnea-hypopnea index used to assess sleep apnea severity.

In the U.S., approximately 1 to 4 million sleep studies are conducted annually, mainly for evaluating sleep apnea. These studies collect extensive data related to patients' brain, lung, muscle, and heart functions, but clinicians typically focus only on a limited amount of this information to assess sleep apnea severity.

Dr. Reena Mehra, a sleep medicine specialist and senior author of the study, emphasized, "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."

The AI model was created by sleep physicians, AI experts, data scientists, and neuroscientists through the Discovery Accelerator, a collaborative research initiative between Cleveland Clinic and IBM aimed at advancing life sciences through AI and quantum computing. By using data from the Cleveland Clinic's STARLIT registry, the team categorized patients into five distinct risk groups. The model effectively predicted outcomes for both genders, while the apnea-hypopnea index has historically favored men in its predictions. These findings were also verified in a broader national patient cohort.

According to Jeffrey L. Rogers, a global research leader at IBM and co-author of the study, "Modern AI lets us recover much more information contained in a night’s worth of sleep physiology, revealing clinically meaningful patient groups with very different long-term health risks."

This model may enhance our understanding of how sleep impacts health by identifying subtle physiological patterns that predict risks for heart disease and neurological conditions. It proposes a more individualized approach to sleep medicine, emphasizing the importance of sleep in chronic disease management.

Matheus Lima Diniz Araujo, a sleep researcher at Cleveland Clinic, noted that nearly 70 million Americans suffer from chronic sleep and wakefulness disorders, highlighting the potential benefits of this discovery in personalizing sleep medicine.

Erhan Bilal, a lead author of the study, remarked on the limited use of physiological information from sleep studies, despite their relevance to overall health. Dr. Mehra concluded that as these methods undergo further validation, they might transform sleep studies from merely diagnostic tools into valuable resources for predicting an individual's future health and understanding the links between sleep physiology and chronic diseases.

The research was supported by the Cleveland Clinic-IBM Discovery Accelerator Program and a grant from the National Heart Lung and Blood Institute.