Identifying an accurate and feasible measurement tool for assessing physical activity (PA) is a priority in epidemiologic research. Although wearable activity monitors yields high potentials for being utilized in research, little information is available on the accuracy of those monitors in young children.
PURPOSE: To evaluate the accuracy of the Fitbit Flex (FF) activity monitor for assessing preschooler’ PA and sedentary behavior (SED) in free-living conditions, using the previously validated accelerometer-based monitor, ActiGraph GT3X+ (AG), as a criterion measure.
METHODS: 27 preschool age children (Girl: 41%, Age: 4.9 ± 1.1 yrs, BMI: 16.5 ± 1.6 kg·m2) wore AG (on dominant hip) and FF (on the non-dominant wrist) simultaneously for 24 hours. Using age-appropriate cutpoints for AG (Pate’s 2006) and manufacturer-specific algorithms for FF, data from AG (15-sec epoch) and FF (60-sec epoch) were reduced in terms of time spent (min/day) in SED, moderate-to-vigorous PA (MVPA), and Total PA (TPA). Pearson correlations was used examine agreement between the estimates from AG and FF. Mean absolute percent errors (MAPEs) were computed as measurement errors. Equivalence test using SAS PROC MIXED procedure was used to compare the 85% confidence intervals (CI) of the estimates from the FF with the respective equivalence zone (EZ; ± 15% of the mean estimates) from the AG.
RESULTS: The FF yielded significantly equivalent estimates of SED (FF: Mean (M) = 673 min, 85% CI: 632-714 min vs. AG: M = 631, EZ: 537-726 min) and TPA (FF: M = 337 min, 85% CI: 309-365 min vs. AG: M = 379 min, EZ: 303-435 min) as the AG. However, the estimate of MVPA from the FF was not equivalent to that from the AG. Correlations between FF and AG were consistently high for SED (r = 0.86, P <.01) and TPA (r = 0.70, P <.01), but moderate for MVPA (r = 0.59, P <.01). MAPEs were 9.2, 70.1, and 14.5% for SED, MVPA, and TPA, respectively.
CONCLUSIONS: The accuracy of the FF for estimating SED and TPA was supported by the high correlations and significant equivalence to AG. However, relatively large MAPEs and results from equivalence test suggest that the MVPA estimates from the FF were not equivalent to those from AG. Future studies utilizing FF in preschool age children should be aware of these findings. Efforts to replicate our findings with longer monitoring in larger samples are warranted.