Estimation of Physical Activity Energy Expenditure in Children during Stimulated Free-living Activities: Precision of the ActiHeart Sensor

Introduction
The validity of the Actiheart – a combined heart rate and movement sensor – for assessing physical activity energy expenditure has been demonstrated in older children during treadmill walking and running and in a limited range of ‘lifestyle’ physical activities (Corder et al., 2005, 2007). Our aim was to evaluate the precision of the Actiheart in younger children performing a wider range of child-relevant simulated free-living activities.
Methods
Thirty-two children (13 girls and 19 boys) aged 9-11 years volunteered to participate. Using regional data on the most commonly performed physical activities in this age group, we devised two routines of six activities ranging from sedentary (e.g. card playing) to vigorous intensity (e.g. soccer). Participants were randomly assigned to one routine, performing each activity for 5 min, with 5 min rest. Indirect calorimetry (Cosmed, K4 b2) was used to estimate resting energy expenditure (short protocol) and criterion physical activity energy expenditure above rest in J/min/kg (Weir, 1949). We derived Actiheart-based estimates from the built-in child ‘group’ equation, and from individual heart rate: energy expenditure relationships observed in a modified 8-min ramped step test. Valid data were available for 21 children (9 girls, 12 boys). Accounting for the hierarchical structure of the data (repeated measures within subjects) we applied a linear mixed model (random slopes and intercepts, unstructured covariance) to derive the within-subjects standard error of the estimate. This statistic provides the typical error in predicting criterion physical activity energy expenditure from either of the Actiheart estimates, at the level of the individual participant.
Results
There was no substantial difference between methods in the mean physical activity energy expenditure. The within-subjects standard error of the estimate from the Actiheart group equation was 53 J/kg/min (90% confidence interval 44 to 63 J/min/kg), versus 49 J/min/kg (42 to 58 J/kg/min) from the Actiheart individual calibration equation.
Discussion
The increase in precision afforded by the simple individual calibration was trivial. The typical within-subject prediction error from both Actiheart equations was around 15% of the mean across all activities (336 J/kg/min). We conclude that this level of precision is adequate for the assessment of free-living physical activity energy expenditure in children.