Development of a Machine Learning Model for Screening Sleep Apnea in Heart Failure Patients Using Sleep Sensor Data.

Abstract

Sleep apnea (SA) is a prevalent disorder among individuals with heart failure (HF), often leading to complications. Early identification is essential for timely interventions and better outcomes. This study explores the feasibility of developing a screening tool for SA in patients with HF using data from the Future Patient Telerehabilitation program. A random forest classifier was used to develop a predictive model, achieving a promising receiver operating characteristic area under the curve (ROC-AUC) of 0.85, suggesting that the random forest classifier has the potential as a SA screening tool for HF patients. However, the study lacked key variables, such as oxygen saturation, that are strong predictors for SA assessment according to current literature; this limits the model's generalizability. Despite this, the findings indicate that the ML model shows promise for screening SA in HF patients, highlighting the need for high-quality, standardized data from future clinical trials to enhance its accuracy and clinical utility.