0100 Daytime Sleep-Tracking Performance of Four Wearable Devices During Unrestricted Home Sleep

Abstract

Abstract Introduction Many previous studies, from our group and others, have tested the sleep-tracking performance of commercial wearable devices and generally found that many can track sleep-wake patterns on most nights in laboratory or home settings with equal or better performance as actigraphy. However, nearly all previous studies tested devices under fixed time in bed (TIB) and only during nighttime sleep. Despite the relevance for night shift workers, device algorithms are programmed/optimized for tracking nighttime sleep, and daytime sleep-tracking performance largely remains unexplored. We therefore tested the sleep-tracking performance of devices during unrestricted home daytime sleep. Methods Participants were 16 healthy young adults (6 men, 10 women; 26.6±4.6 years, mean±SD) with habitual daytime sleep schedules (i.e., slept between 06:00 and 22:00 for ≥1 hour at least twice weekly). Participants slept at home for 1 week under unrestricted conditions (i.e., self-selecting TIB) using a set of four commercial wearable sleep-tracking devices and completed sleep diaries. Wearables included the Fatigue Science Readiband, Fitbit Inspire HR, Oura Ring, and Polar Vantage V Titan. TIB biases and missed daytime sleep episodes were assessed against sleep diaries. Results In total, 86 episodes met criteria for “daytime sleep,” ranging from 2-10 episodes per participant. Percentage of daytime sleep episodes with TIB biases ≤15 and ≤60 minutes, and percentage of missed episodes in total and for short TIB (i.e., <4 hours), respectively, were as follows: Readiband (33.8%, 90.8%, 11.0%, 85.7%), Inspire HR (60.4%, 87.7%, 2.4%, 6.3%), Ring (39.5%, 90.7%, 35.8%, 85.7%), and Vantage V Titan (49.0%, 92.2%, 38.6%, 100%). Conclusion The commercial wearable devices generally had similar performance for tracking daytime sleep episode TIB. Like our previous findings when the devices were tested during nighttime sleep, TIB biases were also low for most daytime sleep episodes. However, the devices missed detecting several daytime episodes, which occurred more often when TIB was <4 hours. Preliminary findings suggest that daytime sleep TIB tracking is largely achievable with different commercial wearable devices; however, device sleep algorithms are not as reliable as when tracking nighttime sleep. Daytime sleep-tracking performance should be explored further. Support (If Any) Office of Naval Research, Code 34