AI Model Uses Routine Sleep Study Data to Predict Long-Term Health Risks

A novel AI model developed by Cleveland Clinic and IBM analyzes routine sleep study data to predict long-term health risks, including heart disease, cognitive decline, and mortality. By grouping patients into five distinct risk categories, the model identified a high-risk group with double the five-year mortality risk of the lowest-risk group. This approach outperforms the traditional apnea-hypopnea index, which discards rich physiological data and historically exhibits sex bias, by delivering equal predictive accuracy for both men and women.

AI sleep study model

• ▪The artificial intelligence model was developed by a collaborative team of sleep physicians, AI researchers, data scientists, and neuroscientists through the Discovery Accelerator, a partnership between Cleveland Clinic and IBM
• ▪A novel artificial intelligence model can analyze routine sleep study data to identify patients' long-term health risks, including heart disease, cognitive decline, and death

Mortality risk stratification

• ▪Patients categorized in the highest-risk group by the artificial intelligence model had twice the mortality risk over the next five years compared to those in the lowest-risk group
• ▪The artificial intelligence model grouped patients into five distinct risk categories using data from the Cleveland Clinic Sleep Signals, Testing, and Reports Linked to Patient Traits registry

Traditional AHI limitations

• ▪The apnea-hypopnea index compresses complex physiological recordings into a single summary number of breathing pauses per hour, discarding continuous data regarding heart rate variability, brainwave architecture, muscle tone, and oxygen desaturation
• ▪The standard clinical measure used to assess sleep apnea severity, the apnea-hypopnea index, failed to capture the five-year mortality risk distinctions identified by the artificial intelligence model

Sex-balanced diagnostic accuracy

• ▪The artificial intelligence model predicted health outcomes with equal accuracy for both men and women, whereas the traditional apnea-hypopnea index has historically performed better in men
• ▪By analyzing full-spectrum physiological signals rather than relying solely on upper-airway obstruction counts, the artificial intelligence model achieves equal predictive power for cardiovascular disease, cognitive decline, and mortality across both female and male patients

Underutilized polysomnography data

• ▪An estimated 1 million to 4 million polysomnograms, or in-lab sleep studies, are performed annually in the United States, typically to evaluate sleep apnea
• ▪While routine polysomnograms collect rich data on a patient's brain, lungs, muscles, and heart, clinicians historically have focused on only a small subset of that information to grade sleep apnea severity