Relationships of Linear and Non-linear Measurements of Post-stroke Walking Activity and Their Relationship to Weather

Background: Stroke survivors are more sedentary than the general public. Previous research on stroke activity focuses on linear quantities. Non-linear measures, such as Jensen-Shannon Divergence and Lempel-Ziv Complexity, may help explain when and how stroke survivors move so that interventions to increase activity may be designed more effectively. Objectives: Our objective was to understand what factors affect a stroke survivor's physical activity, including weather, by characterizing activity by step counts, structure, and complexity. Methods: A custom MATLAB code was used to analyze clinical trial (NCT02835313, https://clinicaltrials.gov/ct2/show/NCT02835313) data presented as minute by minute step counts. Six days of data were analyzed for 142 participants to determine the regularity of activity structure across days and complexity patterns of varied cadences. The effect of steps on structure and complexity, the season's effect on steps, structure, and complexity, and the presence of precipitation's effect on steps and complexity were all analyzed. Results: Step counts and regularity were linearly related (p < 0.001). Steps and complexity were quadratically related (r(2) = 0.70 for mean values, 0.64 for daily values). Season affected complexity between spring and winter (p = 0. 019). Season had no effect on steps or structure. Precipitation had no effect on steps or complexity. Conclusions: Stroke survivors with high step counts are active at similar times each day and have higher activity complexities as measured through patterns of movement at different intensity levels. Non-linear measures, such as Jensen-Shannon Divergence and Lempel-Ziv Complexity, are valuable in describing a person's activity. Weather affects our activity parameters in terms of complexity between spring and winter.