Trajectories of physical activity after ischemic stroke: exploring prediction of change

Abstract Background and aims Physical activity (PA) is associated with lower risk of stroke and better functional outcome. However, sedentary behavior after stroke is prevalent. We aimed to identify predictors for PA after first-time ischemic stroke and to develop prediction models to assess change in PA after stroke. Methods We used the Physical Activity Scale for the Elderly (PASE) to quantify PA prior to stroke and six months after stroke. Demographic and clinical data were enriched with registry data on socioeconomic status (SES). Associations with post-stroke PA were analyzed using linear regression. Elastic net regression models were used to predict decrease from higher PASE quartile to the lowest and increase from lowest to higher. Results A total of 523 patients, from The Efficacy of Citalopram Treatment in Acute Ischemic Stroke (TALOS) trial, had complete PASE data. Median IQR age was 69 years 59, 77, 181 (35%) were female and median IQR National Institute of Health Stroke Scale score was 3 2, 5. Overall, median PASE score did not change, but 20 % of patients decreased to the lowest PASE quartile whereas 48% from the lowest PASE quartile increased to a higher level. The prediction model performance, as represented by the area under receiver operating curve, was 0.679 for predicting a decrease and 0.619 for predicting an increase in PA. SES was the most consistent predictor for both decrease and for a lack of increase in PA. Conclusions Almost 1/2 of the least active patients increased their PA after stroke whereas 1/5 decreased to the least active level with socioeconomic factors being the most consistent predictors for change. Despite including comprehensive data, the PA prediction models had a modest predictive performance. Efforts to optimize PA should include all stroke survivors to improve PA for previously inactive patients and prevent PA decrease.