Non-linear modelling of motor development in typically developing children and youth using robotic assessments

Background Clinical tasks are most often used to differentiate motor performance of individuals who have impaired function. However, these are not as accurate and repeatable as robotic tasks. Additionally, motor development occurs rapidly at early ages and slows as they reach adulthood, resulting in a non-linear model of performance. There is also evidence that variability in performance changes as children and youth age. Accurate normative models of performance are necessary to identify differentiate deficiencies in motor performance and to track the efficacy of therapies. Methods Two-hundred and eighty-eight participants who are typically developing (ages 5-18) completed a robotic point-to-point reaching task and an object hitting task using the Kinarm Exoskeleton. Exponential or quadratic curves were fit to performance parameters generated by Kinarm to model typical performance. These models included a linear term to account for changing variability with age. Results Most performance parameters showed improvement with age, and none showed deterioration. Some parameters showed large changes in variability in performance with age, up to a 74% decrease in the range of typical performance. Conclusions Reduced variability occurs with age, indicating the need to account for differences in variability when developing models of typical motor performance in children and youth. Models used to identify deficits in motor performance should account for changing variability in data and changing repeatability with age to increase accuracy of identification of deficits.