Comparison of PAEE from combined and separate heart rate and movement models in children

PURPOSE:Accurate measurement of physical activity in children is a challenge. Combining physiological (e.g., heart rate (HR)) and body movement registration (e.g., accelerometry) may overcome limitations with either method used alone. This study aimed to compare the estimated physical activity energy expenditure (PAEE) from hip- and ankle-mounted MTI Actigraphs, a hip-mounted Actical, and a new combined HR and movement sensor, the Actiheart (Cambridge Neurotechnology, Papworth, UK). METHODS:Resting EE and submaximal EE (treadmill walking and running) were measured in 39 children (13.2 +/- 0.3 yr) by indirect calorimetry during a progressive treadmill exercise bout. Associations between monitor outputs (activity counts, HR, and activity counts + HR) and the criterion were examined by linear regression models. The agreement between measured and predicted PAEE was examined by modified Bland-Altman plots in a subsample of participants. RESULTS:The combined Actiheart model (activity counts + HR) had the strongest relationship with PAEE (R2 = 0.86), compared with those from the single-measure models (R2 = 0.69 and 0.82 for the activity model and HR model). The explained variances from the other activity monitors were lower (R2 = 0.50, 0.37, and 0.67) for the hip MTI, ankle MTI, and Actical, respectively. In cross-validation analyses, significant correlations were observed between estimation errors of the methods with the criterion (r = 0.49 to 0.90) in all models using only activity counts indicating a large systematic error. The HR and combined models indicated less systematic error (r = 0.41 and 0.33, respectively). CONCLUSIONS:Of the techniques considered, combined HR and movement sensing is the most valid for estimating PAEE in children during treadmill walking and running, compared with movement or HR alone. It also has the lowest level of systematic error.