Multi-night evaluation of the Oura Ring in free-living older adults: App-reported sleep efficiency as an indicator of measurement stability

OBJECTIVES: Consumer sleep trackers are increasingly used for long-term sleep monitoring, but evidence regarding their performance remains limited in older adults. Here, we evaluated agreement and night-to-night stability of sleep metrics derived from the Oura Ring against the Dreem 3 EEG headband in free-living older adults and examined whether Oura-determined sleep efficiency (SE) could indicate measurement stability. METHODS: Forty-five older adults (mean age±SD: 74.4±2.4 years; 23 females) concurrently wore the Oura Ring Gen3 and Dreem 3 headband during habitual sleep at home, contributing 210 nights of paired recordings. Total sleep time (TST), SE, sleep onset latency (SOL), wake after sleep onset (WASO), and sleep stages were compared. Participants were stratified by average Oura-determined sleep efficiency across nights (≥85% versus <85%). Device agreement, epoch-by-epoch performance, and night-to-night variability in Oura-Dreem bias were assessed. RESULTS: Compared with Dreem 3, Oura underestimated TST (-14.6±35.7 min), SE (-2.9±6.7%), and REM sleep duration (-23.8±29.3 min), while overestimating SOL (11.4±16.8 min). Overall sleep/wake agreement was moderate (accuracy: 85.3±4.8%; sensitivity: 88.4±6.0%; kappa: 0.58±0.10). Critically, participants with lower average Oura-determined SE (<85%) demonstrated larger TST bias and lower sleep/wake agreement, accuracy, sensitivity, and kappa compared those with higher SE. Lower mean Oura-determined SE was also associated with greater night-to-night variability in Oura-Dreem bias, most notably for SOL (ρ=-0.30), with weaker associations for WASO (ρ=-0.19) and TST (ρ=-0.10). CONCLUSIONS: Oura-determined sleep efficiency may provide useful contextual information regarding measurement stability and help identify circumstances in which wearable-derived sleep metrics should be interpreted with greater caution.