A new AI model analyzes routine overnight sleep studies to identify latent physiological patterns linked to future health outcomes, including heart disease, cognitive decline, and mortality.

Researchers categorized patients into five risk groups using the STARLIT data, with the highest-risk group showing about double the five-year mortality risk of the lowest-risk group.

The approach suggests routine medical tests hold rich physiological information not currently captured, enabling earlier and more personalized care.

Study data came from the Cleveland Clinic STARLIT registry and findings were independently validated in a nationwide cohort.

The work was developed through the Cleveland Clinic–IBM Discovery Accelerator and was validated across a nationwide cohort.

Researchers used STARLIT registry data and confirmed findings in a broader national sample.

The method achieved meaningful risk stratification in both men and women and was independently validated nationwide, though broader validation in diverse populations is needed before clinical deployment.

Key contributors include Reena Mehra, M.D.; Jeffrey L. Rogers, Ph.D.; Erhan Bilal, Ph.D.; Matheus Lima Diniz Araujo, Ph.D.; Carl Saab, Ph.D.; and others, reflecting a multidisciplinary team.

Notable authors include Reena Mehra (senior author) and Jeffrey L. Rogers, with collaboration from IBM and Cleveland Clinic, supported by the Discovery Accelerator and an NHLBI grant.

Study authors also include Erhan Bilal, Matheus Lima Diniz Araujo, Kristen Beck, Catherine Heinzinger, Samer Ghosn, Nancy Foldvary-Schaefer, and others.

Next steps involve validating findings in diverse international populations, expanding collaborations, and integrating the model into sleep-lab software for automated risk stratification alongside standard metrics.

Future plans include broader validation and deeper partnerships with industry and professional societies.