Dynamic instability of breathing during sleep predicts cognition, disease, and mortality.
STUDY OBJECTIVES To test whether an interpretable effort-derived Breathing Stability Index (BSI) captures physiology and clinical risk beyond conventional sleep-disordered breathing metrics. METHODS We analyzed 43,611 overnight polysomnography records from two clinical cohorts and one community cohort. BSI was derived from abdominal and thoracic effort envelopes as a continuous measure of respiratory effort instability. We compared BSI with AHI, hypoxic burden, arousal index, periodic breathing/self-similarity, desaturation frequency, sleep stage, and event phenotype; evaluated cognition and nine matched cross-sectional EHR disease labels; and tested mortality in Cox models. RESULTS BSI increased across AHI severity and tracked comparator metrics while remaining partly nonredundant. In mutually adjusted models, BSI remained associated with periodic breathing, AHI, arousal index, and desaturation index. Within AHI strata, instability was lowest in deeper NREM sleep and higher in lighter sleep and REM sleep. Cognitive associations were cohort dependent and strongest in MrOS, where less stable breathing was associated with worse fluid and total cognition after age/sex adjustment. In matched EHR disease analyses, BSI features provided modest discrimination (AUROC 0.509-0.642) and age/sex-adjusted associations for 8 of 9 labels. BSI was associated with all-cause mortality after adjustment for age, sex, and comparator sleep metrics in Massachusetts General Hospital (HR 1.38 per SD, 95% CI 1.23-1.54) and Beth Israel Deaconess Medical Center (HR 1.10, 95% CI 1.02-1.19), but not MrOS. CONCLUSIONS Nocturnal breathing stability captures dynamic sleep-disordered breathing physiology that complements event counts and desaturation metrics and relates to cognition, cross-sectional disease labels, and cohort-dependent survival risk.