Measurement of physical activity with accelerometers in children
The aim of this dissertation was to examine the use of accelerometers to measure physical activity (PA) in children with and without disabilities and address some of the current methodological issues involved with their use. Study I compared the activity level of children with Down syndrome (DS), children with intellectual disabilities without DS, and children without disability. Subjects wore ActiGraph accelerometers for seven days. There were no significant differences in PA level among the three groups of children as measured by mean ActiGraph counts per day. No difference was found between groups in the amount of time spent daily at different activity intensities. This study was the first to objectively assess PA in children with disabilities using accelerometers and demonstrated that accelerometers are feasible for PA measurement in children with cognitive impairments.
Study II and Study III examined methodology issues with accelerometer data interpretation. Study II compared the accuracy of several count cut-point values used to define activity intensity ranges. Fifty-one children wore an ActiGraph at the hip while performing locomotor or free-play activities. Actual MET level, measured via indirect calorimetry, was compared to the predicted MET level from the different ActiGraph cut-points. Percent agreement between actual and estimated intensity level ranged from five to 100 percent between the different cut-point thresholds. These results quantified the differences in intensity classification seen with the different cut-points and demonstrated ActiGraph results extrapolated from different cut-points cannot be compared.
Study III assessed the use of the count coefficient of variation (CV) to distinguish locomotor from non-locomotor activities in children. The accuracy of energy expenditure (EE) prediction using a dual-equation method based on count CV was compared to the accuracy of using a single prediction equation. The count CV distinguished between locomotor and non-locomotor activities quite well. Despite this, using the activity specific dual-equation method to predict EE did not improve accuracy compared to single prediction equations. No prediction equation tested adequately estimated EE from count values in this study. Due to the variable nature of movement and play in children, and metabolic and size differences, predicting EE from hip-mounted accelerometers is not ideal.