Validating the Sedentary Sphere method in children: does wrist or accelerometer brand matter?

This study aimed to validate the Sedentary Sphere posture classification method from wrist-worn accelerometers in children. Twenty-seven 9-10-year-old children wore ActiGraph GT9X (AG) and GENEActiv (GA) accelerometers on both wrists, and activPAL on the thigh while completing prescribed activities: five sedentary activities, standing with phone, walking (criterion for all 7: observation) and ten minutes free-living play (criterion: activPAL). In an independent sample, 21 children wore AG and GA accelerometers on the non-dominant wrist and activPAL for two days of free-living. Percent accuracy, pairwise 95% equivalence tests (±10% equivalence zone) and intra-class correlation coefficients (ICC) analyses were completed. Accuracy was similar, for prescribed activities irrespective of brand (non-dominant wrist: 77%-78%; dominant wrist: 79%). Posture estimates were equivalent between wrists within brand (±6%, ICC>0.81, lower 95% CI>0.75), between brands worn on the same wrist (±5%, ICC>0.84, lower 95% CI>0.80) and between brands worn on opposing wrists (±6%, ICC>0.78, lower 95% CI>0.72). Agreement with activPAL during free-living was 77%, but sedentary time was underestimated by 7% (GA) and 10% (AG). The Sedentary Sphere can be used to classify posture from wrist-worn AG and GA accelerometers for group-level estimates in children, but future work is needed to improve the algorithm for better individual-level results