Epoch Length and Accelerometer Outputs in Children: Comparison to Direct Observation

Purpose: To determine the effects of epoch length and activity count cutpoints on ActiGraph (AG; ActiGraph Health Services, Pensacola, FL) accelerometer estimates of time in moderate-to-vigorous physical activity (MVPA) in fifth-grade children monitored during physical education (PE) compared with a direct observation (DO) criterion standard.

Methods: A sample of 32 fifth-grade males and females (mean age = 10.3 +/- 0.5 yr) wore an AG attached at the waist for a 30-min PE class. Participants were concurrently videotaped, and the Computerized System for Observing Fitness Instruction Time (C-SOFIT) was used to create a DO measure of MVPA time (DO_MVPA). AG data were collected in 5-s epochs then integrated into 10-, 15-, 20-, 30-, and 60-s epochs. AG activity counts were converted into time (in s) in MVPA using validated (and epoch-adjusted) children's activity count cutpoints established by Treuth et al. (AG_T), Mattocks et al. (AG_M), and Freedson et al. (AG_F).

Results: All AG_T and AG_M epoch detected significantly lower time in MVPA than DO_MVPA. The percentage of DO_MVPA detected by AG_T and AG_M epochs ranged from 46% to 61% and from 26% to 47%, respectively. All AG_F epochs yielded similar (i.e., nonsignificant) mean estimates of MVPA versus DO_MVPA, with modest increases in root mean squared error (RMSE) with increasing epoch length. The percentage of DO_MVPA detected by AG_F epochs ranged from 93% to 100%.

Conclusions: All AG_F epoch lengths provide comparable mean estimates to DO-detected MVPA time in fifth-grade children during PE. To minimize error among individual estimates, shorter epoch lengths should be used, with 5-s epochs yielding the lowest RMSE in the current study. Considerations of both epoch length and activity count cutpoint are important to improved detection of intermittent bouts of MVPA among fifth-grade children.