Validation of Apple Watch for Estimating Moderate-to-Vigorous Physical Activity in Children

Abstract

Identifying accurate instruments for assessing physical activity (PA) is crucial for surveillance and promotions of PA in children. Apple Watch has been appeared as one of the most popular wearable devices that are designed to monitor individual’s PA. However, little knowledge is available whether it provides an accurate estimate of time spent in moderate-to-vigorous PA (MVPA) in children. PURPOSE: To examine the validity of Apple Watch 3 in measuring MVPA in children using a portable indirect calorimetry system (Cosmed K5) as a criterion measure. We hypothesized the estimate of MVPA from Apple Watch would be comparable with that from Cosmed K5 for simulated free-living activities. METHODS: 20 school-age children (girls: 45%; age: 9.7 ± 2.0 yrs, BMI: 16.3 ± 3.2 kg/m2) were fitted with an Apple Watch 3 on their dominant wrist and Cosmed K5 portable indirect calorimetry. All participants performed sedentary (i.e., sitting, watching TV), light PA (i.e., slow walking, playing with toys), and MVPA (i.e., brisk walking) over a 45-minute period. We calculated METRMR by dividing obtained VO2 values from Cosmed K5 by child’s predicted resting metabolic rates using the age- and sex-specific Schofield equations. We used ≥ 3 METRMR as a cut-off value to define a criterion measure of MVPA time (i.e. Cosmed K5) againt which Exercise Time from Apple Watch was compared. Pearson’s correlation coefficient and mean absolute percent error (MAPE) were calculated, and equivalence test using SAS Proc Mixed procedure was performed to examine whether the Apple Watch’s Exercise Time is comparable to MVPA time from Cosmed K5. RESULTS: The Exercise Time from Apple Watch showed a moderate correlation (r = 0.35, p = 0.13) and relatively high MAPE (30.2 ± 4.9%) in relation to MVPA time from Cosmed K5. The 90% confidence interval of the Apple Watch’s Exercise Time was from 16.7 to 22.9 min, which was not completely included within the zone of equivalence defined as ± 10% of the values of Cosmed K5 (22.8 to 27.9 min); this indicates non-equivalence between Apple Watch and Cosmed K5. CONCLUSION: Apple Watch showed poor agreement with Cosmed K5 for estimating MVPA time in this convenient sample of children. Subsequent research is needed to further examine the validity of Apple Watch monitor for quantifying various intensities of PA under free-living conditions. Copyright © 2019 by the American College of Sports Medicine