Impact of Accelerometer Epochs on Classification of Children Activity Patterns

Introduction Children have intermittent activity patterns and this must be considered when assessing activity with accelerometers. This study examines the impact of

accelerometer epoch duration on the recorded time spent in physical activity during recess. A second goal was to determine whether variability in activity is associated with overall time spent in activity.

Methods Data were obtained from one school in the Ready for Recess project. Participants (60, 3rd-5th grade children) wore an Actigraph accelerometer for four days. Data during recess were processed using two different methods to determine the impact of epoch length on group activity levels. One method used the recorded (5 s) epochs and the other used aggregated 1 min epochs (similar to ActiLife5 software). Both methods used the same equation (Freedson-Trost) to classify epochs into intensities (Sedentary, Light, Moderate and Vigorous). The relation between variability and time spent in activity were examined with an indicator of count variability (SD of absolute difference between 5 s counts).

Results Boys spent 25.6% time in Sedentary, 16.6% in Light, 35.6% in Moderate and 22.3% in Vigorous. Girls spent more time in Sedentary (33.9%) and Light (21.4%) but less time in Moderate (30.3%) and Vigorous (14.4%). When the same data were aggregated into 1 min epochs the allocations were drastically different. In boys, the respective allocations shifted to 10.7%, 18.5%, 54.0%, and 16.7% and in girls, the values shifted to 12.5%, 29.5%, 50.5%, and 7.5%. Time in Sedentary and Vigorous were considerably lower (61% and 35%, respectively) while time spent in Light and Moderate were considerably higher (27% and 59%, respectively). MVPA levels were reduced by 25.9%. The bias was consistent in boys and girls, however the underestimation of girl’s Vigorous activity was 2 times greater than boys (48% vs 25%). The indicator of epoch variability was positively correlated with percent time in Vigorous (r=0.72), Moderate (r=0.20), and Light (r=0.24), but negatively associated with Sedentary activity (r=-0.28). The associations for Vigorous were considerably higher in boys (r=0.83) than girls (r=0.51).

Discussion and Conclusion The results reveal large differences in group activity profiles depending on how the data were processed. An indicator of activity variability was shown to be highly correlated with total time spent in MVPA confirming that activity in children is inherently intermittent (even within a minute).

Using shorter epochs (e.g. 5 seconds) captures children’s intermittent activity patterns and provides a more accurate indicator of time spent in MVPA.