ABSTRACT
The purpose of this study was to compare activPAL algorithm-estimated values for time in bed (TIB), wake time (WT) and bedtime (BT) against self-report and an algorithm developed by van der Berg and colleagues. Secondary analyses of baseline data from the Community Activity for Prevention Study (CAPs) were used in which adults ≥ 18 years wore the activPAL for seven days. Mixed-effects models compared differences between TIB, WT, and BT for all three methods. Bland-Altman plots examined agreement and the two-one-sided test examined equivalence. activPAL was not equivalent to self-report or van der Berg in estimating TIB, but was equivalent to self-report for estimating BT, and was equivalent to van der Berg for estimating WT. The activPAL algorithm requires adjustments before researchers can use it to estimate TIB. However, researchers can use activPAL’s option to manually enter self-reported BT and WT to estimate TIB and better understand 24-hour movement patterns.
Acknowledgments
Thank you to the CAPs study participants. Thank you to Dr. Zach Weller in the CSU Statistics Department for consulting on the analyses and R code. Thank you to van der Berg and colleagues for making their algorithm freely available and providing trouble-shooting tips.
Declaration of conflicts of interest
Kate Lyden is the Chief Science Officer at VivoSense, Inc. She previously worked as an independent consultant for several wearable device manufacturers, including PAL Technologies.