Integrated linear and nonlinear trunk dynamics identify residual concussion deficits

•Mild traumatic brain injury, or concussion, affects postural control.•Recurrence quantification analysis revealed differences in the dynamics of postural sway between the concussion and control group regardless of single and dual-task conditions.•Single and dual-task condition did not differentially affect concussion and control groups.•Recurrence quantification analysis may play an important part in quantifying postural control following a concussion.•Findings complement past investigations of a concussion’s effect on postural control using non-linear approaches and extend past observations that have used traditional measures. Postural sway is significantly affected by a mild traumatic brain injury, or concussion, and myriad methods have been developed to quantify the severity of concussion symptoms. The current manuscript quantifies postural sway—as measured by an inertial sensor—in youth athletes with concussion (n = 43, age = 14.4 ± 2.3 years, 56% female, tested median 7 days post-concussion) and healthy controls (n = 38, age = 14.9 ± 2.0 years, 55% female) during single-task and dual-task postural sway. A nonlinear analysis (i.e., recurrence quantification analysis [RQA]) and several common linear measures were used to quantify postural sway. Respectively, the two complementary types of analyses describe the structure and magnitude of postural sway. We hypothesized that participants who recently experienced a concussion would display differing postural sway dynamics (i.e., different in structure and magnitude) than control participants who had not experienced a concussion. Additionally, a logistic regression was performed to determine which combination of variables (nonlinear and linear) and task (single and dual) would best differentiate concussion and control participants. Significant differences between concussion and control participants were found in percent determinism, laminarity, and standard deviation of postural sway acceleration in both the single and the dual task. In the single task alone, mean diagonal line length and trapping time were additionally significantly different between groups. Moreover, the logistic regression model revealed that a mixture of linear and nonlinear measures across both single and dual tasks best classified concussed and non-concussed participants. Additionally, history of concussion was found to be a significant covariate in the model. These results extend past observations by demonstrating that a combination