Sources of Error with Wearable Step Counters

Abstract

PURPOSE: To investigate sources of error with 11 wearable step counting devices, during common types of physical activities. METHODS: 20 participants performed 15 activities for 2 min each, while wearing 11 step counters on the waist, ankle, or non-dominant wrist. Arm activities included: snacking, brushing hair, folding laundry, sweeping, brushing teeth, and meal preparation. Overground activities included: walking holding onto backpack strap, walking with umbrella, walking with hands in pockets, and pushing stroller. Treadmill activities included: walking at 1 mph, walking at 2 mph, walking at 3 mph, walking at 3 mph holding onto bars, and jogging at 6 mph. Wrist-worn devices included: Garmin Vivofit 2, Fitbit Charge, Polar A360, Withings Pulse Ox, and ActiGraph GT3X. Waist-worn devices included: Yamax Digi-Walker SW-200, Fitbit Zip, Omron HJ-322U, and ActiGraph GT3X (without low-frequency extension). Ankle-worn devices included: two StepWatch 3 devices, one with preprogrammed, default settings and one with modified cadence and sensitivity settings. A researcher hand-counted steps during each activity; this served as the criterion. The step counts reported from each device compared to the hand count using a 1-way (1x12) repeated measured ANOVA. If the overall effect for an activity was significant, the outputs from individual devices and the criterion were analyzed using planned contrasts. Devices with significant contrasts (p ≤ 0.05) and observed power greater than 0.8 were considered to be significantly different than the criterion. RESULTS: During arm activities, the wrist-worn devices overcounted steps while hip-worn devices slightly undercounted steps. The ActiGraph GT3X on the wrist greatly overcounted steps during arm activities, while other devices had smaller errors. During treadmill walking at 1 mph, all wrist and hip-worn devices undercounted steps. The ankle-worn device (StepWatch 3) had the smallest error across all activities, especially when programmed with the modified setting. CONCLUSIONS: Individuals using step counting devices should be aware of sources of error in step counts. Contributing factors to error are the wear location, the algorithms used to count steps, and the activities performed.