Accelerometer-determined Step Counts In Simulated Free-living Activities In Children And Adolescents

Abstract

PURPOSE: To evaluate step counting accuracy of the ActiGraph GT3X+ accelerometer in waist- and wrist-worn locations during simulated free-living activities among children and adolescents. METHODS: 123 children and adolescents (6-20 years) performed simulated free-living activities (i.e., seated watching a movie and basketball skill development activity) while concurrently wearing accelerometers on their waist and non-dominant wrist. Each activity was performed for 5 minutes and visually tallied steps served as the criterion (steps/min). Accelerometer data were processed using two filters: 1) normal (NF), and 2) low-frequency extension (LFE). Confidence intervals (95%) for the difference (criterion-test method) were generated using the non-parametric bootstrap (n = 5000 replicates) and used within an equivalence testing framework to evaluate accuracy relative to the pre-defined equivalence margins of ± 3 and ± 10 steps/min for movie watching and basketball, respectively. RESULTS: Compared to visually tallied steps (steps/min), both waist and wrist-worn accelerometers over-counted steps taken during movie watching (waist: NF = 0.02 [0, 0.05], LFE = 0.37 [0.23, 0.54]; wrist: NF = 0.50 [0.37, 0.64], LFE = 2.2 [1.64, 2.81]); however, all confidence intervals were completely contained within the pre-defined equivalence margin. Conversely, waist and wrist-worn accelerometers under-counted steps while performing a simulated basketball skill development activity (waist: NF = -36.51 [-40.31, -33.23], LFE = -11.87 [-15.37, -8.95]; wrist: NF = -28.85 [-32.64, -25.51], LFE = -17.43 [-21.19, -14.11]) with none of the associated confidence intervals falling completely within the pre-defined equivalence margin. CONCLUSION: In agreement with direct observation, the ActiGraph GT3X+ detected few false steps during movie watching with the waist-worn location and NF combination performing best. All combinations of wear location and filter choice underestimated step counts during a simulated basketball activity. Given these limited data, applying the LFE filter to a waist-worn device may provide the best overall free-living step count estimates for children and adolescents. Supported by NIH NCHID 1R21HD073807-01A1