0846 Do Wearables Agree? Concordance of Commercial and Research-Grade Sleep Measures in Older Adults Across Healthy and Clinical Cohorts

Abstract Introduction Sleep in aging is characterized by decreased sleep quantity and quality, further amplified in clinical populations, such as Parkinson’s disease (PD) and people living with HIV (PLWH). Wearables such as actigraphy and commercial devices are increasingly deployed in remote sleep monitoring, yet validation of these technologies primarily reflects young healthy adults under laboratory conditions. Whether these devices align to daily sleep diaries – a standard measurement in the field – in older adults and clinical populations has not been adequately examined. Methods Adults ≥ 50 years were recruited into three groups: healthy controls (n=33), PD (n=21), PLWH (n=21). Participants completed 7-14 consecutive nights of at-home sleep monitoring, as part of an ongoing observational study, wearing an Axivity AX3 wrist actigraphy, Oura ring (generation 3 model), and completing daily electronic sleep diaries. Total sleep time (TST), sleep efficiency (SE), and wake after sleep onset (WASO) were derived for each night using the GGIR R package for actigraphy and default software for the ring. These metrics were compared across devices and diaries using night-by-night Spearman correlations and Bland-Altman analyses to calculate limits of agreement (LOA). Results Correlations between the two devices and sleep diaries were high for TST (ρ=0.64-0.74) and lower for SE (ρ=0.29-0.47) and WASO (ρ=0.27-0.45). Bland-Altman analyses revealed the ring reported lower TST compared to actigraphy with a mean difference of −12.6 minutes (95% LOAs -134.0, 108.8). The ring detected lower SE than actigraphy (mean -9.1%, 95% LOAs -26.3, 8.0). Both devices reported lower TST (actigraphy: -28.0 min 95% LOAs -179.9, 123.8; ring: -35.8 min 95% LOAs -183.0, 111.4) and SE (actigraphy: -4.1% 95% LOAs -28.6, 20.3; ring: -12.8% 95% LOAs -33.9, 8.3) compared to self-report. PD participants were biased towards longer actigraphy TST relative to ring TST. Conclusion Devices agreed moderately for TST and differed most in SE estimates across cohorts. Both devices can be used for scalable, remote sleep monitoring in real world cohorts. Future studies should compare devices with polysomnography in aging clinical populations. Support (if any) R01AG081144