An algorithm for identifying physical activity patterns from motion data.

An algorithm was developed to describe how physical activity (PA) patterns relate to overall motion counts. Thirty-five children wore an accelerometer (7-days). Each motion count was compared with the mean of surrounding counts within 21 min. Counts per minute similar to the mean were grouped into bouts. Counts that differed by more than 20% of the coefficient of variations (based on the mean and SD of the 21 min period) indicated transitions between bouts. Children with more daily motion had more and longer moderate (MPA) and vigorous (VPA) bouts, higher counts during MPA bouts, and more transitions from VPA to VPA bouts. In addition to differences in PA levels, highly active and less active children perform PA differently.