Calibration And Cross-validation Of Wrist Worn Actigraph GT3X+ In Young Preschoolers
- Presented on June 17. 2013
Introduction In order to obtain accurate estimates of sedentary time and physical activity in young children objective methods are needed. Wrist worn accelerometers have shown good feasibility among participants in previous studies. The acceleration signal collected by many commercially available activity monitors is usually summarized in an arbitrary unit (counts). Calibration studies are needed to translate activity counts into intensity categories. The placement of the activity monitor may affect the output, thus affecting the validity of the measurement.
Purpose The aim was to calibrate the Actigraph GT3X+ accelerometer for wrist worn placement in ambulatory young preschoolers by developing intensity thresholds for sedentary, low and high physical activity. Furthermore, to cross-validate the developed thresholds in a sample of young preschoolers and apply the developed thresholds on a sample of free living two year old children.
Methods Wrist worn Actigraph GT3X+ was used to measure physical activity during 30 minutes of structured activities and free play in 38 children (15-36 months). Activity was video recorded and scored based on Children´s Activity Rating Scale (CARS) and combined with accelerometer data on a five second level. To develop intensity thresholds for sedentary, low and high physical activity ROC analysis was performed on data from 26 randomly selected children. The remaining 12 children were used for cross-validation. We applied the developed thresholds on data from 20 free-living children who wore an Actigraph on the wrist for 7 consecutive days.
Results Optimal thresholds for sedentary were ≤89 Y-counts and ≤221 VM-counts/5 sec. Optimal thresholds for high physical activity were ≥440 Y-counts and ≥730 VM-counts/5 sec. Sensitivity and specificity was the same for the Y-axis and VM. Sensitivity for the sedentary threshold was 100% and specificity 60%. For the high physical activity threshold sensitivity was 60% and specificity 92,3%. Strong correlations were found between the developed thresholds for the accelerometer and CARS scoring time in sedentary, low and high intensity physical activity. Free-living children were categorized as sedentary for 384 (SD 70,3); low physical activity for 307 (SD 45,7) and engaged in high physical activity for 89 (SD 33,7) minutes per day.
Conclusions A wrist mounted Actigraph GT3X+ activity monitor can accurately assess sedentary behavior and different levels of physical activity in young preschoolers.