Prediction of Physical Activity Intensity with Accelerometry in Young Children.

Abstract

Background: An algorithm for the classification of ambulatory and non-ambulatory activities using the ratio of unfiltered to filtered synthetic acceleration measured with a triaxial accelerometer and predictive models for physical activity intensity (METs) in adults and in elementary school children has been developed. The purpose of the present study was to derive predictive equations for METs with a similar algorithm in young children. Methods: Thirty-seven healthy Japanese children (four- to six-years old) participated in this study. The five non-ambulatory activities including low-intensity activities, and five ambulatory activities were selected. The raw accelerations using a triaxial accelerometer and energy expenditure by indirect calorimetry using the Douglas bag method during each activity were collected. Results: For non-ambulatory activities, especially light-intensity non-ambulatory activities, linear regression equations with a predetermined intercept (0.9) or quadratic equations were a better fit than the linear regression. The equations were different from those for adults and elementary school children. On the other hand, the ratios of unfiltered to filtered synthetic acceleration in non-ambulatory activities were different from those in ambulatory activities, as in adults and elementary school children. Conclusions: Our calibration model for young children could accurately predict intensity of physical activity including low-intensity non-ambulatory activities.