A wearable-based pathway for characterizing COMISA, Insomnia, and OSA: a feasibility study in a community hospital
TL;DR
Multi-night wearable monitoring may complement group-level phenotype characterization and early triage in real-world sleep services and supportive robustness model replacing AHI metrics with BR-derived hypoxic burden improved model fit.
Abstract
Comorbid insomnia and obstructive sleep apnea (COMISA) – the co-occurrence of both disorders – is associated with greater morbidity and is underdiagnosed in fragmented care outside sleep centers. This study examined the feasibility of using an AI-powered wearable device, the Belun Ring (BR), to characterize COMISA, insomnia, and obstructive sleep apnea (OSA) in a community hospital setting. Consecutive new patients undergoing baseline diagnostic evaluation were classified as having OSA if AHI ≥ 15, as insomnia when met ICSD-3 clinical criteria or ISI ≥ 15 pragmatically when interview data were inconclusive. Diagnosed COMISA if both conditions were present. Participants completed at least 3 nights of BR assessment with sleep architecture, respiratory metrics, and pulse rate variability compared using nonparametric tests, regression analyses, and principal component analysis. Among 96 participants (13 COMISA, 47 insomnia, 36 OSA), COMISA and insomnia showed longer sleep onset latency, greater wake after sleep onset (WASO), and more awakenings than OSA, while AHI was higher in OSA and COMISA. In the prespecified primary non-AHI multinomial model, sleep efficiency, REM, total sleep time, and WASO retained discriminatory information. A supportive robustness model replacing AHI metrics with BR-derived hypoxic burden improved model fit. Multi-night wearable monitoring may complement group-level phenotype characterization and early triage in real-world sleep services.