Examining accelerometer validity for estimating physical activity in pre-schoolers during free-living activity

Abstract

The present study examined the efficacy of accelerometers for the assessment of free play physical activity (PA) in pre-school-aged children with consideration of epoch length and wear location. Following ethics approval, parental informed consent, and child assent, 66 pre-schoolers aged 3-4 years (30 females and 36 males) wore an accelerometer (ActiGraph GT3X; sampling at 100 Hz) on their non-dominant wrist and their right hip during 1 hour of free play. Concurrently, direct observation, using the OSRAC-P, was used to determine sedentary behavior (SB), light (LPA), or moderate-to-vigorous (MVPA) intensity PA. For the ActiGraph, vertical axis counts and summed vector magnitude (VM) for hip, and VM for wrist, were downloaded using 5-, 10-, 15-, and 30-second epoch lengths. Accelerometer counts were averaged over each 30 seconds to match the observation periods. Receiver operating curve analysis was used to evaluate the ability of the ActiGraph to predict SB, LPA, and MVPA. SB and MVPA obtained from wrist- and hip-worn accelerometers demonstrated fair agreement with direct observation (AUC => 0.7). LPA determined by accelerometer had poor agreement with observed LPA, for both the hip and wrist placement (AUC = 0.53-0.56), with weak levels of specificity (0.34-0.43), although sensitivity was fair (0.74-0.84). This study is the first to examine accelerometer validity, considering wear location and epoch in pre-schoolers during free play, and suggests that the ActiGraph is a fair measure for SB and MVPA in pre-school children. Neither placement performed predominantly better irrespective of epochs or used count data (vertical axis, VM).

Keywords: OSRAC; actigraph; activity counts; observation; validation.