The effect of reintegrating Actigraph accelerometer counts in preschool children: comparison using different epoch lengths.

OBJECTIVES: The purpose of this study was to determine whether ActiGraph accelerometer activity counts and estimates of moderate-to-vigorous physical activity collected at a single larger epoch are comparable to those collected at smaller epochs reintegrated into a larger epoch.
DESIGN: A cross-sectional study design.
METHODS: Thirty-one preschoolers (3-5years) concurrently wore four accelerometers that were each initialized at four different epoch lengths (1s, 15s, 30s, and 60s) during a full preschool day. Counts collected at 1s, 15s, and 30s epoch were each reintegrated and compared to those collected at a larger epoch (e.g., counts from one 15s epoch vs. consecutive sum of counts from fifteen 1s epochs). Six sets of cut-points (Pate, Freedson, Sirard, Van Cauwenberghe, Evenson and Puyau) were applied to estimate moderate-to-vigorous physical activity minutes. Paired t-test and Cohen's d were used to compare group mean differences. Absolute percent errors Bland-Altman plots with limits of agreement were used to compare individual differences.
RESULTS: Minimal group mean differences were found for counts and moderate-to-vigorous physical activity estimates between larger and reintegrated epochs. Relatively smaller absolute percent errors (6.2-9.2%) and limits of agreements (-15.52%, 18.00% to -28.27%, 28.02%) were observed for counts than absolute percent errors (10.1-50.3%) and limits of agreements (-27.3%, 33.3% to -156.9%, 137.9%) for moderate-to-vigorous physical activity estimates.
CONCLUSIONS: Smaller individual differences in activity counts tended to yield larger individual variations in moderate-to-vigorous physical activity estimates, despite minimal group mean differences. Therefore, researchers reintegrating smaller epochs into a larger epoch should be conscious of possible differences in moderate-to-vigorous physical activity estimates obtained from a single larger epoch. Crown