Article Title



Emma Cate Jones, Melissa Gagnon, Emma Sullins, Joseph Pederson, Rebecca Rogers, Mallory Marshall. Samford University, Birmingham, AL.

Background: Little is known about the effect on the validity of wrist-worn step counting devices during activities limiting arm swing (i.e. holding on to an object). The purpose of this study was to determine the difference in step counts as measured by various wrist-worn step counters (Apple Watch 3, Apple Watch 6, FitBit Inspire, AmazFit Bip, ActiGraph GT9X Link, and Garmin Vivofit 4) when worn during activities with typical arm swing versus with limited arm swing. Methods: N=12 male and female college-age participants completed a single lab visit with 15 trials (free walking at 2 and 3.5 mph, jogging at 6 mph, walking at 2 and 3.5 mph while carrying a coffee mug (right and left hand), a textbook (right and left hand), and pushing a shopping cart, walking at 3.5 mph while pushing a stroller, and jogging at 6 mph while pushing a stroller). The six devices were randomly assigned to a wrist placement with three devices on the right wrist and three on the left. Device step counts were then compared to hand-counted steps from GoPro camera footage; the camera was mounted to a chest harness and pointed towards the participants’ feet during all trials. For each trial, absolute percent error (APE) was calculated as ([device measured steps - actual steps]/actual steps) × 100. One-sample t-tests were used to compare APE for each device to the gold standard of hand-counted steps, and one way ANOVA with Tukey correction for multiple comparisons was used to compare the six devices to one another. In all comparisons, alpha of 0.05 was utilized. Results: Most devices undercounted steps for the trials with limited arm swing. For the control trials with typical arm swing, devices undercounted by as much as 23.9% (ActiGraph at 2mph) and overcounted by as much as 7.6% (AppleWatch 3 at 3.5mph). When holding a coffee mug or textbook, devices had an APE ranging from -15.5% to 13.7% at 2mph and -19.9% to 0.7% at 3.5mph. When holding an object fixed to the floor (i.e. a stroller or shopping cart), device APE was significantly greater than for holding the book or mug (p<0.05 for all devices). APE for the shopping cart trial ranged from -23.8 to -93.7% while APE for the stroller ranged from -7.3 to -94.8% at 3.5mph and -12.3 to -96.6% at 6mph. For all stroller and cart trials, the ActiGraph Link had significantly greater APE than the other devices (p<0.05). Conclusions: Consumer-grade wrist-worn step counters may undercount steps during activities with limited arm swing (i.e. hand on an object); this is particularly true when the object is fixed to the floor, such as a shopping cart or stroller.

This document is currently not available here.