Abstract Overview
Background: There is a scarcity of openly-accessible algorithms for classifying children’s physical activity types and postures. The ActiPASS software – based on the open-source Acti4 algorithm – can be used to classify physical activity types in adults using thigh-worn accelerometry. However, the original algorithm has not been validated in children, nor has it been compared to a child-specific set of algorithm thresholds.
Purpose: The objective of this study was to evaluate the accuracy of two ActiPASS thresholds for classifying physical activity types and postures in children.
Methods: The laboratory condition comprised 41 children (11.0 ± 4.8 years; 46.5% male), while the free-living condition comprised 15 children (10.0 ± 2.6 years; 66.6% male). Children were asked to complete a standardized activity protocol in the laboratory condition and up to 2 hours of self-selected activities during the free-living condition. A single accelerometer (Axivity AX3, 100 Hz and ±8 g) was worn by the participants on the dominant thigh. Annotated video recordings were used as reference across both conditions.
Results: Using the original adult thresholds, the mean balanced accuracies (95% CI) for the laboratory condition ranged from 0.62 (0.56 – 0.67) for lying down to 0.97 (0.94 – 0.99) for running. For the free-living condition, the balanced accuracies ranged from 0.65 (0.51 – 0.79) for lying down to 0.96 (0.92 – 0.99) for cycling. No differences between the adult and child-specific algorithm thresholds were observed, except for walking in laboratory conditions.
Conclusions: The validation results indicate that ActiPASS can accurately classify the types of physical activity performed by children using a single accelerometer worn on the thigh.
Practical implications: Researchers may choose to use ActiPASS with either of the two existing thresholds to identify basic activity types in children.
Funding: CL is supported by a fellowship of the German Academic Exchange Service (DAAD).
Additional Authors