Improving precision in physical activity measurement in developmental coordination disorder research

Abstract

Developmental coordination disorder (DCD) is characterized by significant problems with skilled physical movements which cannot be attributed to intellectual disability, visual impairment, or neurological conditions.¹ Consequently, children with DCD have lower levels of physical activity compared to typically developing children.^{2, 3} Evidence provided by Tran et al.⁴ suggested that children with DCD spend less time in moderate-to-vigorous physical activity than typically developing children. Despite the benefits of objective physical activity measures, we know very little about specific activities or contexts in which children with DCD experience significant deficits. The article by Letts et al. does a remarkable job by characterizing physical activity types that pose enormous challenges to preschoolers with DCD.⁵

In this paper, the authors use a robust DCD identification approach based on multiple motor assessments and demonstrate an innovative application of machine learning to characterize physical activity in preschool age. The authors reported no group differences in daily time spent in light-intensity activity, moderate-to-vigorous intensity activities, or sedentary behavior. However, they showed significant differences in the types of physical activity performed, with children with DCD spending less time in continuous motor skills (e.g. walking and running) than typically developing children. The results suggest that traditional processing pipelines are likely obscuring relevant physical activity information. These findings recognize the relevance of advanced computational methods in DCD research and suggest opportunities for translational science.

A major challenge is the lack of diversity in the sample involved in this study. The majority of device-measured physical activity studies among children with DCD has been conducted in societies that are often described as WEIRD (Western, Educated, Industrialized, Rich and Democratic).⁶ This does not represent the full spectrum of DCD or global pediatric population. The lack of sample diversity limits generalizability and further contributes to disparities or inequalities in our field. A related issue is the physical activity heterogeneity problem. We will not be able to elucidate variability in physical activity behavior by studying a homogenous group of children. It is necessary to diversify our samples and/or deploy robust approaches to advance precision characterization of activity profiles.

Finally, emerging machine learning models are yet to be validated in developmental disabilities, including children with DCD. Validating advanced computational methods among preschoolers with DCD is essential as it will lead to optimized models that can enable analysis of big data needed to create generalizable knowledge to improve physical activity for all children.

These findings have important implications for practice and future research. From a clinical practice perspective, targeting the types of physical activity in which children with DCD spend less time performing could improve activity participation and mitigate the adverse effects associated with physical inactivity across the lifespan. Physical activity measurement in DCD has grown considerably and will continue to evolve in innovation and discovery. If we genuinely apply novel methodologies, we will be able to create better metrics to address physical inactivity in children with DCD in early development.

As DCD research expands, it is critically important to keep pace with rapidly advancing computational methods without forsaking the need to improve generalizability, reduce disparities, and maximize developmental potential for all children, particularly those from marginalized groups or communities. The application of reliable motor assessments and sophisticated computational approaches in DCD research is laudable, but future studies should do better to precisely characterize activity and sources of variation in highly representative or diverse samples. I applaud Letts et al. for this incredible effort and hope this commentary will stimulate further discussions and nudge us all toward precision characterization of physical activity in our field.