Project Summary
Obstructive sleep apnea (OSA) affects over one billion adults and is an independent risk factor for
cardiovascular disease (CVD). Yet, in randomized clinical trials (RCT), treatment of OSA has failed to
demonstrate a beneficial impact of continuous positive airway pressure (CPAP) on cardiovascular (CVD) event
rates in this population. In this proposal, we hypothesize that the non-significant RCTs are not due to lack of
continuous positive airway pressure (CPAP) effectiveness but instead due to suboptimal CPAP adherence,
variability in the clinical presentation of OSA as well as the heterogeneity of treatment effect with CPAP.
Notably, no study has applied machine learning (ML) to multimodal data that extends beyond
polysomnography to identify individuals at enhanced risk for atherosclerosis progression or experiencing CVD
events. The 2021 NIH Sleep Research Plan identified critical and high-priority areas for further research,
including leveraging ML analytic approaches for big data to advance our understanding of sleep disorders and
assist in the personalization of treatment. The overall goal of this proposal is to apply ML to well-characterized
datasets with multimodal data to develop separate prediction tools for predicting incident CVD events (Aim 1),
and heterogeneity of treatment effect with CPAP in OSA patients (Aim 2). We will then validate the models
using real-world electronic health records to ensure their generalizability and clinical relevance (Aim 3). This
groundbreaking proposal aims to revolutionize the management of OSA patients by developing prediction tools
using multimodal data and cutting-edge ML techniques, resulting in a more personalized approach to care that
can improve patient outcomes and reduce the burden of OSA-related CVD events. These decision tools will be
readily integrated into the clinical environment, guiding treatment decisions and assisting sleep physicians in
determining which patients should avoid CPAP use and which OSA patients should be prioritized for CPAP
treatment, optimizing treatment plans and reducing healthcare costs.