O007 Validation of an under-mattress sleep tracker to estimate sleep and wake during day and night sleep opportunities

INTRODUCTION: Consumer sleep trackers are useful for tracking daily sleep, particularly in shift-workers, yet large, multi-night validations against direct sleep measures are lacking. Conducted as one of the largest sleep tracker validation studies to date, this study evaluated performance of an under-mattress sensor (Withings Sleep Analyzer [WSA]) to classify sleep and wake versus polysomnography (PSG) in a multi-night laboratory protocol with both day and night sleep opportunities. METHODS: 27 healthy sleepers attended the laboratory twice, for 8 consecutive days (48% male, mean[SD] age=27[9]years). Participants underwent simultaneous PSG and WSA recordings for the initial night-time sleep (22:00-07:00) and 6 subsequent daytime sleeps (10:00-19:00) during simulated night-shift work. Epoch-by-epoch analysis of the 252 nights tested accuracy, sensitivity, and specificity for the first night-time sleep opportunity versus subsequent daytime sleeps. Sleep duration estimates were compared between WSA and PSG. RESULTS: Across all recordings, the WSA showed 83% sleep-wake classification accuracy, 87% sensitivity to sleep and 25% specificity to wake. The WSA significantly overestimated sleep duration versus PSG (25[65]minutes, p<0.05). Accuracy and specificity were higher for night versus day sleeps (88% and 38% versus 81% and 18% respectively, all p<0.05), while sensitivity (90% versus 88%) did not significantly differ. DISCUSSION: The WSA was moderately accurate compared to PSG, and comparable to other movement-based sleep trackers. The WSA was more accurate at classifying sleep and wake during night sleep opportunities compared to daytime sleeps. This differential performance is likely due to poorer sleep quality commonly observed during daytime sleep that is harder to accurately classify.