Accuracy of Activity Trackers during Treadmill Walking Versus Outdoor Walking

Abstract

Establishing the accuracy of devices that measure daily activity is important in controlled lab settings and in real-life settings. PURPOSE: To assess the accuracy of a pedometer and 6 popular activity trackers at measuring steps while walking on a treadmill and walking outside. METHODS: Twenty-three college students (Mean±SD; 22.2±3.9yrs; 24.9±4.1kg/m2, 11 male) walked 500 steps at 3mph on a treadmill while wearing 7 different activity trackers (Pedometer (PED), Blaze (BLA), Charge HR (CHR), Alta (ALT), Flex (FLE), Zip (ZIP), One (ONE)). During a second visit, participants wore the devices while walking 400 meters at 3mph outside. Steps were counted by a trained researcher using a hand tally counter. Mean Absolute Percent Error (MAPE) values were calculated for each device relative to the tally counter and were correlated between each of the devices and the tally counter using Pearson correlations. Significance was set at p<0.05. Mean bias scores were calculated between the step counts for each device and the tally counter. RESULTS: MAPE values were significantly correlated between the treadmill and outdoor protocol for the PED (r=0.634, p<0.001). The remaining devices were not correlated between protocols (p>0.05). The treadmill protocol produced underestimations in step counts for 5 devices (mean bias±SD: PED = -1.4±41.5steps; BLA = -35.3±70.8; CHR = -3.9±51.9; ALT = -86.5±74.4; FLE = -16.9±71.6) and slight overestimations for 2 devices (mean bias±SD: ZIP = 2.1±3.5; ONE = 0.3±2.2). The outdoor protocol produced step count overestimations for all devices. MAPE values were approximately twice as large or greater for all devices in the outdoor protocol compared to the treadmill protocol except for the PED (MAPE±SD: 4.0±7.2 treadmill vs. 3.5±5.2 outdoor). Besides the PED, the ONE was the most accurate during the treadmill protocol (MAPE±SD: 0.3±0.3) and the ZIP was the most accurate during the outdoor protocol (MAPE±SD: 14.7±6.6). CONCLUSION: The step counting devices in this study performed better in the controlled laboratory setting compared to the outdoor setting with a device worn on the waist producing the best results in each trial. These findings indicate that step counts in real-life settings from commercial activity devices may produce significant error.