Low-Cost Consumer-Based Trackers to Measure Physical Activity and Sleep Duration Among Adults in Free-Living Conditions: Validation Study

Abstract

Background: Wearable trackers for monitoring physical activity (PA) and total sleep time (TST) are increasingly popular. These devices are used not only by consumers to monitor their behavior but also by researchers to track the behavior of large samples and by health professionals to implement interventions aimed at health promotion and to remotely monitor patients. However, high costs and accuracy concerns may be barriers to widespread adoption.

Objective: This study aimed to investigate the concurrent validity of 6 low-cost activity trackers for measuring steps, moderate-to-vigorous physical activity (MVPA), and TST: Geonaut On Coach, iWown i5 Plus, MyKronoz ZeFit4, Nokia GO, VeryFit 2.0, and Xiaomi MiBand 2.

Methods: A free-living protocol was used in which 20 adults engaged in their usual daily activities and sleep. For 3 days and 3 nights, they simultaneously wore a low-cost tracker and a high-cost tracker (Fitbit Charge HR) on the nondominant wrist. Participants wore an ActiGraph GT3X+ accelerometer on the hip at daytime and a BodyMedia SenseWear device on the nondominant upper arm at nighttime. Validity was assessed by comparing each tracker with the ActiGraph GT3X+ and BodyMedia SenseWear using mean absolute percentage error scores, correlations, and Bland-Altman plots in IBM SPSS 24.0.

Results: Large variations were shown between trackers. Low-cost trackers showed moderate-to-strong correlations (Spearman r=0.53-0.91) and low-to-good agreement (intraclass correlation coefficient [ICC]=0.51-0.90) for measuring steps. Weak-to-moderate correlations (Spearman r=0.24-0.56) and low agreement (ICC=0.18-0.56) were shown for measuring MVPA. For measuring TST, the low-cost trackers showed weak-to-strong correlations (Spearman r=0.04-0.73) and low agreement (ICC=0.05-0.52). The Bland-Altman plot revealed a variation between overcounting and undercounting for measuring steps, MVPA, and TST, depending on the used low-cost tracker. None of the trackers, including Fitbit (a high-cost tracker), showed high validity to measure MVPA.

Conclusions: This study was the first to examine the concurrent validity of low-cost trackers. Validity was strongest for the measurement of steps; there was evidence of validity for measurement of sleep in some trackers, and validity for measurement of MVPA time was weak throughout all devices. Validity ranged between devices, with Xiaomi having the highest validity for measurement of steps and VeryFit performing relatively strong across both sleep and steps domains. Low-cost trackers hold promise for monitoring and measurement of movement and sleep behaviors, both for consumers and researchers.

Keywords: accelerometry; fitness trackers; mobile phone; physical activity; sleep.

Conflict of interest statement

Conflicts of Interest: None declared.

©Laurent Degroote, Gilles Hamerlinck, Karolien Poels, Carol Maher, Geert Crombez, Ilse De Bourdeaudhuij, Ann Vandendriessche, Rachel G Curtis, Ann DeSmet. Originally published in JMIR mHealth and uHealth (http://mhealth.jmir.org), 19.05.2020.

Figures

Figure 1

Tracker characteristics.

Figure 2

An example of the measurement protocol for one period.

Figure 3

Correlations between steps estimates per day from the trackers and the ActiGraph.

Figure 4

Correlations between moderate-to-vigorous physical activity estimates per day from the trackers and the ActiGraph.

Figure 5

Bland-Altman plots of the trackers for measuring steps. The middle line shows the mean difference (positive values indicate an underestimation of the wearable and negative values indicate an overestimation) between the measurements of steps of the wearables and the ActiGraph and the dashed lines indicate the limits of agreement (1.96×SD of the difference scores).

Figure 6

Bland-Altman plots of the trackers for measuring moderate-to-vigorous physical activity. The middle line shows the mean difference (positive values indicate an underestimation of the wearable, and negative values indicate an overestimation) between the measurements of moderate-to-vigorous physical activity of the wearables and the ActiGraph, and the dashed lines indicate the limits of agreement (1.96×SD of the difference scores).

Figure 7

Correlations between total sleep time estimates from the trackers and the SenseWear.

Figure 8

Bland-Altman plots of the trackers for measuring total sleep time. The middle line shows the mean difference (positive values indicate an underestimation of the wearable and negative values indicate an overestimation) between the measurements of total sleep time of the wearables and the SenseWear, and the dashed lines indicate the limits of agreement (1.96×SD of the difference scores).