Reliability of commercially available sleep and activity trackers with manual switch-to-sleep mode activation in free-living healthy individuals.

INTRODUCTION

Wearable health devices have become trendy among consumers, but it is not known whether they accurately measure sleep and physical activity parameters. To address this question, we have studied the measured data of two consumer-level activity monitors (Up Move Jawbone (U) and Withings Pulse 02 (W)) and compared it with reference methods for sleep and activity recordings, namely the Bodymedia SenseWear Pro Armband actigraph (SWA) and home-polysomnography (H-PSG).

METHODS

Twenty healthy patients were assessed at home, during sleep, with the four devices. An additional 24-h period of recording was then planned during which they wore the 2 trackers and the SWA. Physical activity and sleep parameters obtained with the 4 devices were analyzed.

RESULTS

Significant correlations with H-PSG were obtained for total sleep time (TST) for all the devices: r=0.48 for W (p=0.04), r=0.63 for U (p=0.002), r=0.7 for SWA (p=0.0003). The best coefficient was obtained with SWA. Significant correlations were also obtained for time in bed (TIB) for U and SWA vs PSG (r=0.79 and r=0.76, p<0.0001 for both) but not for W (r=0.45, p=0.07). No significant correlations were obtained for deep sleep, light sleep, and sleep efficiency (SE) measurements with W, U and SWA. Sleep latency (SL) correlated with H-PSG only when measured against SWA (r=0.5, p=0.02). Physical activity assessment revealed significant correlations for U and W with SWA for step count (both r=0.95 and p<0.0001) and active energy expenditure (EE) (r=0.65 and 0.54; p=0.0006 and p<0.0001). Total EE was also correctly estimated (r=0.75 and 0.52; p<0.0001 and p=0.001).

CONCLUSION

Sleep and activity monitors are only able to produce a limited set of reliable measurements, such as TST, step count, and active EE, with a preference for U which performs globally better. Despite the manual activation to sleep mode, U and W were not suitable for giving correct data such as sleep architecture, SE, and SL. In the future, to enhance accuracy of such monitors, researchers and providers have to collaborate to write algorithms based reliably on sleep physiology. It could avoid misleading the consumer.