Analyzing routine sleep study data with artificial intelligence reveals a wealth of biomarkers predictive of a much broader range of health risks than doctors usually identify, researchers have found.
Polysomnograms, or in-lab sleep studies, are typically performed to look for obstructive sleep apnea, a common disorder in which throat muscles relax and repeatedly block the airway during sleep.
The studies collect data on breathing, brain and muscle activity and other factors, but doctors generally focus on the apnea-hypopnea index, or the average number of breathing pauses and shallow breathing events per hour.
For the new study, investigators tasked AI with uncovering hidden physiologic patterns in more than 10,000 full-night sleep studies and linking the patterns with long-term clinical outcomes in patients’ medical records.
AI, using a wealth of data typically ignored during such studies, identified five clinically meaningful patient subtypes, each with sharply different long-term health risks, they reported in Nature.
For example, patients whose sleep data patterns put them in the highest-risk group for serious health issues had twice the odds of death over the next five years compared to those in the lowest-risk group, a distinction not captured by the apnea-hypopnea index.
In this most at-risk group, odds were 65% higher for heart failure, 84% higher for heart attack, 93% higher for cognitive impairment, and more than 200% higher for atrial fibrillation and for epilepsy, compared to odds in the lowest-risk group.
AI predicted outcomes well for men and women, while the apnea-hypopnea index has historically performed better in men, the researchers noted.
Their model was equally accurate when they tested it in a separate nationwide cohort of more than 6,000 patients, they said.
“For decades we have distilled an overnight sleep study into a handful of summary measures,” study leader Dr. Reena Mehra of the University of Washington said in a statement.
“AI gives us the opportunity to move beyond those summaries and learn from the full richness of sleep physiology.”
