Abstract Introduction Respiratory instability during sleep reflects irregular ventilatory control and may contribute to sleep fragmentation. Traditional sleep apnea metrics, like the apnea-hypopnea index, count discrete events but miss continuous breathing physiology. We developed two novel respiratory-instability markers. Self-Similarity (SS) represents the percentage of respiratory epochs with self-similar patterns, where 10% indicates acceptable stability. The Breathing Stability Index (BSI) represents the median breath-to-breath stability across the night, where 0.5 indicates stable breathing and 1.5 indicates marked instability. We characterized their demographic patterns across a large clinical dataset. Methods We analyzed 24,596 polysomnograms from Beth Israel Deaconess Medical Center (BIDMC) and Massachusetts General Hospital (MGH) (age 55.7±16.7 years, 55.7% male), including diagnostic (n=13,579), split-night (n=5,185), and titration (n=5,832) studies. SS and BSI were computed from respiratory envelopes derived from effort belts, allowing for breath-to-breath amplitude comparison. Sleep measures included N1/N2/N3/REM percentages, sleep efficiency, and sleep fragmentation index (SFI). We examined distributions by study type, demographic associations using correlations and t-tests, and sex and site differences using Cohen's d. Results The mean SS was 5.2±6.9% (range 0-86%) and the mean BSI was 1.06±0.48 (range 0.2-4.8). Instability above clinical thresholds (SS≥10%) was evident in all study types, with the highest seen in split night studies: 11% of diagnostic studies (mean BSI=1.00), 11% of titration studies (mean BSI=1.02), and 34% of split-night studies (with 11% exceeding 20%; mean BSI=1.22). Both metrics increased with age (r=0.17, p 0.001) in steady, monotonic increments. Men exhibited higher instability (SS d=0.40; BSI d=0.19) than women. MGH showed higher SS (d=0.42) but slightly lower BSI (d=-0.10) than BIDMC. The two metrics showed moderate intercorrelation (r=0.43), and moderate correlation with N1% (r=0.24) and SFI (SS: r=0.16; BSI: r=0.23), indicating related but non-redundant sleep-instability information. Conclusion This large dataset confirmed a high prevalence of SS and warrants further clinical attention. The moderate correlation between SS and BSI (r=0.43) confirms they provide complementary insights into breathing control that are otherwise missed by discrete event-based metrics. These measures are now integrated into a larger multicomponent analysis suite to support clinical decision making at BIDMC by enabling more comprehensive assessment of sleep-disordered breathing phenotypes. Support (if any) 1R01HL161253-01A1 R01AG073410
Han et al. (Fri,) studied this question.