Identification of Apnea Events Using a Chest‐Worn Physical Activity Monitor

Abstract

Obstructive sleep apnea (OSA) is a condition characterized by upper airway obstruction during sleep causing intermittent hypoxia and nighttime awakening. It is a common condition in the United States that is often undiagnosed. It is a significant risk factor for decreased daytime productivity, quality of life, cardiovascular disease, and death. The current gold standard for diagnosis of OSA is laboratory‐based polysomnography (PSG). While PSG is necessary for the diagnosis and monitoring of OSA, many patients have limited access to PSG due to wait times at PSG laboratories or economic or geographic limitations. Portable sleep monitoring has been studied as a possible solution for patients who do not have access to timely PSG. This study aimed to use the Zephyr BioHarness 3, a chest‐worn physical activity monitor that records movement and physiologic data in real‐time, to detect apnea events in patients with suspected OSA undergoing single‐night laboratory PSG. Twenty patients underwent single‐night laboratory‐based PSG while simultaneously wearing the Zephyr BioHarness 3. The Zephyr BioHarness 3 data was analyzed using three methods. First, apnea events were identified in 10‐second windows of Zephyr data via support vector machine, logistic regression, and neural network (sensitivity = 76.0 ± 0.3%, specificity = 62.7 ± 0.2%, accuracy = 63.7 ± 0.1%). Second, apnea events were identified using the mean, median, and variance of the 10‐second windows (sensitivity = 72.3 ± 0.3%, specificity = 69.4 ± 0.1%), accuracy 69.6 ± 0.1%). Third, apnea events were identified using phase‐space transformation of the Zephyr BioHarness 3 data (sensitivity = 76.9 ± 0.3%, specificity = 77.9 ± 0.1 %, accuracy = 77.9 ± 0.1%). The Zephyr BioHarness shows initial promise as a possible OSA screening tool for patients suspected of OSA but who lack access to timely laboratory‐based PSG.