1002 Examination of Wearable and Non-Wearable Consumer Sleep-Tracking Devices Versus Polysomnography

The gold-standard sleep assessment technique, polysomnography (PSG), is impractical for use in many settings and populations (e.g., military) where objective sleep tracking is needed. Consumer-available sleep-tracking devices are being rapidly developed and, if validated, may be used to effectively track sleep in diverse settings and populations outside the laboratory or clinic. Therefore, we performed validation testing and comparison of wearable and non-wearable consumer sleep-tracking devices against PSG. In an ongoing study, we tested 18 healthy participants (6 men and 12 women, 27.1±4.0 y; mean±SD) who slept 8 hours with PSG on 3 consecutive nights in a controlled sleep laboratory along with a set of consumer-available sleep-tracking devices. This included two wrist-worn devices (LifeTrak Brite R440, Fitbit Alta HR), an under-mattress device (EarlySense Live), and a bedside device (ResMed S+). Participants also wore a standard wrist actigraph for comparison (Philips Respironics Actiwatch-2). They maintained consistent 8-hour sleep schedules at home for 4 nights prior to the study and abstained from naps, caffeine, and alcohol. Epoch-by-epoch validation statistics of sensitivity (for sleep), specificity (for wakefulness), and Cohen’s kappa were calculated for each device and compared with PSG-scored sleep and wakefulness. Preliminary results for the device epoch-by-epoch measures of sensitivity, specificity, and Cohen’s kappa, respectively, were as follows: Actiwatch-2 (97.1%, 36.8%, 0.42), Brite R440 (98.5%, 28.3%, 0.36), Alta HR (94.3%, 46.2%, 0.42), Live (96.1%, 45.4%, 0.47), and S+ (93.3%, 50.0%, 0.44). Overall, the consumer sleep-tracking devices we tested showed relatively high sensitivity but lower specificity, indicating a tendency for sleep devices to accurately detect sleep but to underdetect wakefulness. In this preliminary analysis, one wrist-worn device (Alta HR) and two non-wearable devices (Live and S+) showed the highest specificities, indicating these devices are promising for detection of wakefulness and comparable to standard wrist actigraphy. Further testing is needed to determine the validity and practicality for use of these consumer sleep-tracking devices within different settings and populations. Office of Naval Research (ONR), Code 30