Analysis Of Accelerometer Counts During Sedentary Activities In 5-11 Year Old Children Using a Wrist-Worn Accelerometer

PURPOSE: A calibration study was conducted to derive cut-point thresholds for distinguishing wrist movement during sedentary activities from wrist movement during light intensity walking using a popular wrist-mounted accelerometer in children aged 5-11 years. METHODS: 176 children, performed up to 8 sedentary activities for 5 minutes while wearing accelerometers on both wrists. All activities were performed in a summer day camp setting and represented free-living, seated sedentary activities that require arm/wrist movement normally performed by children. Activities included reading books, sorting cards, cutting & pasting, playing board games, eating snack, playing games on tablets, watching T.V., and writing. Participants walked at a self-selected pace to elicit light physical activity. Direct observation was used as the criterion measure for verification of sedentary behavior versus light activity. RESULTS: Participants were stratified by gender and age, then assigned to a calibration group (n=100) or a validation group (n=67). A 10-fold Receiver Operator Characteristic (ROC) analysis resulted in an optimal cut-point threshold of 203 counts per 5-second epoch with sensitivity, specificity and an area under the curve (AUC) of 71.56, 70.83 and 0.72, respectively for the non-dominant wrist. For the dominant wrist, an optimal cut-point threshold of 229 counts per 5-second epoch with sensitivity, specificity and an AUC of 67.10, 66.77 and 0.67, respectively. A 10-fold cross-validation resulted in an AUC of 0.70 for the non-dominant wrist, demonstrating good classification accuracy. A 10-fold cross-validation resulted in an AUC of 0.67 for the dominant wrist. CONCLUSION: The cut-point thresholds derived in the current study are 41-123 counts/5 second epoch higher compared to previously published wrist-based cut-points for this population with similar classification accuracy. The newly developed cut-point thresholds may more accurately classify sedentary behavior versus light intensity activity compared to previously published thresholds due to the wide variety of sedentary activities included.