Prediction of Energy Expenditure by Combined Movement and Heart Rate Sensing During Varying Activities in Children

Abstract

Physical activity is difficult to assess accurately, especially in children. Various equations have been derived to estimate physical activity energy expenditure (PAEE) from body movement measured by accelerometry or heart rate (HR) data. However, few studies have utilised combined HR and movement sensing (HR+M). PURPOSE: The primary purpose of this study was to compare the accuracy of uniaxial accelerometry and HR+M to predict PAEE during six common activities in children. As a secondary aim we assess the validity of three sets of treadmill-derived equations (Corder et al, MSSE 2005) to predict PAEE in this sample. METHODS: PAEE was measured by indirect calorimetry during six activities (lying, sitting, slow walking, walking, jogging and hopscotch) in 181 children (12.4 ± 0.2y). Associations between measured and predicted PAEE (accelerometry output and HR+M) were assessed by linear regression analysis. The validity of these equations was cross-validated in a sub sample of participants. The validity of previously derived PAEE equations from treadmill walking and running was assessed. RESULTS: Data from the Actigraph and the HR+M were significantly associated with measured PAEE values (r2 = 0.91 and 0.90, P < 0.01). In cross-validation analyses, significant correlations were observed between the estimation errors of both predictions (Actigraph r=0.46, P < 0.01; Actiheart r=0.27, P < 0.01), both manifesting as under estimations at high-energy expenditures, increasing with PAEE. Systematic errors (i.e. significant correlations between estimation errors) were observed for all treadmill-derived equations. Uniaxial accelerometry over estimated PAEE significantly (r = −0.74, P < 0.01). The branched equation model over estimated PAEE at low intensities (r = 0.23, P < 0.01), whereas the HR+M prediction equation showed less systematic error (r = −0.09, P < 0.01). CONCLUSIONS: Both accelerometry and HR+M are valid to predict PAEE during selected physical activities in children. However, both models seem to underestimate PAEE at high intensity physical activity. Accelerometry derived PAEE during a progressive treadmill test was not suitable for predicting PAEE during the six activities. Both the HR+M model and the branched equation model derived from combined HR and movement sensing during treadmill locomotion showed less systematic error and were valid for PAEE prediction. Our results suggest that it may be possible to derive accurate PAEE prediction models using HR+M data that would not be possible using movement data alone due to the mechanical limitations of accelerometers.