Trait-like nocturnal sleep behavior identified by combining wearable, phone-use, and self-report data

Stijn A A Massar,Alyssa S C Ng,Xin Yu Chua,Tih Shih Lee,Ju Lynn Ong,Nicholas I Y N Chee,Michael W L Chee,Chun Siong Soon,Arko Ghosh

doi:10.1038/s41746-021-00466-9

Abstract

Using polysomnography over multiple weeks to characterize an individual’s habitual sleep behavior while accurate, is difficult to upscale. As an alternative, we integrated sleep measurements from a consumer sleep-tracker, smartphone-based ecological momentary assessment, and user-phone interactions in 198 participants for 2 months. User retention averaged >80% for all three modalities. Agreement in bed and wake time estimates across modalities was high (rho = 0.81–0.92) and were adrift of one another for an average of 4 min, providing redundant sleep measurement. On the ~23% of nights where discrepancies between modalities exceeded 1 h, k-means clustering revealed three patterns, each consistently expressed within a given individual. The three corresponding groups that emerged differed systematically in age, sleep timing, time in bed, and peri-sleep phone usage. Hence, contrary to being problematic, discrepant data across measurement modalities facilitated the identification of stable interindividual differences in sleep behavior, underscoring its utility to characterizing population sleep and peri-sleep behavior.

Highlights

Sleep is increasingly recognized as a major modifiable lifestyle risk factor and this has contributed to a boom in sales of consumer wearables that track it
There was good agreement in bed and wake times determined by EMA and Oura, while tappigraphy-based sleep duration was longer (Fig. 3b, c)
We combined sleep measurement using wearable sleep tracking with smartphone based tappigraphy and EMA to provide redundant as well as complementary information about sleep behavior

Summary

INTRODUCTION

Sleep is increasingly recognized as a major modifiable lifestyle risk factor and this has contributed to a boom in sales of consumer wearables that track it. Wearable sleep trackers measure body motion, heart rate, and skin temperature or some combination of these and provide periodic feedback about user behavior through appealing graphs and charts These inexpensive devices hold promise for collecting large-scale longitudinal sleep data that could dramatically improve the characterization of longitudinal sleep patterns compared to cross-sectional estimates of sleep duration or quality obtained from questionnaire data. A common approach for garnering acceptance of the use of wearable and smartphone-based technologies for research is to compare their sleep measurement performance to PSG and research actigraphy Such comparisons are important for benchmarking the accuracy of each device, but they do not highlight the merits of using less accurate but more convenient and.

RESULTS

Summary variable

DISCUSSION

Participants and procedures

CODE AVAILABILITY