Comparing physical activity estimates in children from hip-worn Actigraph GT3X+ accelerometers using raw and counts based processing methods

This study examined differences in physical activity (PA) estimates provided from raw and counts processing methods. One hundred and sixty-five children (87 girls) wore a hip-mounted ActiGraph GT3X+ accelerometer for 7 days. Data were available for 129 participants. Time in moderate PA (MPA), vigorous PA (VPA) and moderate-vigorous PA (MVPA) were calculated using R-package GGIR and ActiLife. Participants meeting the wear time criteria for both processing methods were included in the analysis. Time spent in MPA (-21.4 min.d-1, 95%CI -21 to -20) and VPA (-36 min.d-1, 95%CI -40 to -33) from count data were higher (P<0.001) than raw data. Time spent in MVPA between the two processing methods revealed significant differences (All P<0.001). Bland-Altman plots suggest that the mean bias for time spent in MPA, VPA and MVPA were large when comparing raw and count methods. Equivalence tests showed that estimates from raw and count processing methods across all activity intensities lacked equivalence. Lack of equivalence and poor agreement between raw and count processing methods suggest the two approaches to estimate PA are not comparable. Further work to facilitate the comparison of findings between studies that process and report raw and count physical activity data may be necessary.