AI Program Shows Promise In Predicting Health Risks From Sleep Apnea

Recently, researchers at the Mt. Sinai Health System were awarded a five-year, $4.1 million grant to develop and test the artificial intelligence-powered model they built to predict health complications from OSA. The AI model they created has the potential to predict a range of high- and low-risk OSA outcomes, such as daytime sleepiness, depression, heart disease, and hypertension, more accurately than current methods. 

It’s estimated that obstructive sleep apnea (OSA) affects almost 1 billion people across the world. Trusted Source National Library of Medicine, Biotech Information The National Center for Biotechnology Information advances science and health by providing access to biomedical and genomic information. View Source As one of the most common sleep disorders in the U.S., Trusted Source National Library of Medicine, Biotech Information The National Center for Biotechnology Information advances science and health by providing access to biomedical and genomic information. View Source there is significant motivation for healthcare providers to better understand sleep apnea and its health consequences. 

Currently, physicians use the apnea-hypopnea index (AHI) from a patient’s sleep study to determine their OSA’s severity and the effects it may have on their health. Trusted Source Wiley Online Library Wiley Online Library is one of the largest and most authoritative collections of electronic journals published by Wiley, as well as a vast and growing collection of reference works and other books. View Source It indicates how many times, on average, they stop breathing or have shallow breathing during an hour of sleep. In recent years, experts have begun examining and criticizing AHI readings because they can’t forecast OSA’s effects on patients’ health particularly well. Trusted Source Oxford Academic Journals (OUP) OUP publishes the highest quality journals and delivers this research to the widest possible audience. View Source  

The Mt. Sinai team’s AI program evaluates data from several sleep study parameters commonly impacted by OSA, such as breathing, oxygen levels, and sleep stages, to calculate a probability score indicating a patient’s level of risk for complications. According to the research team, one AI model using data from 4,700 participants could predict death due to heart disease with up to 81% accuracy, while predictions based on a current model are only 58% accurate. 

The researchers also used a model for lower-risk issues like daytime sleepiness. Early data from close to 11,000 study participants across three groups suggests the model can predict sleepiness with up to 87% accuracy, compared to the AHI index’s 54%. 

The grant from the National Heart, Lung, and Blood Institute will allow the team to test their AI models on Mt. Sinai Integrative Sleep Center patients over three months. They will measure and record patients’ brain waves, breathing, heart rates, and oxygen levels, then analyze and validate the findings against existing data. 

Anit Parekh, Ph.D., principal investigator and assistant professor of medicine says, “Our study will assess the real-world performance of an AI approach and offer crucial evidence needed to translate metrics… into clinical practice.”

The researchers believe that honing their existing model and gathering more patient data will give physicians higher-quality tools to diagnose and treat patients with OSA. The grant reflects an innovative step forward in better understanding how OSA affects people’s health.