Predicting long-term movement behavior patterns after stroke: Development of a Clinical Prediction Rule

Background: Within the first years post-discharge, movement behavior of people with a first-ever stroke often deteriorates, with inactive movement behavior increasing the risk of recurrent cardiovascular events. Early identification of patients at risk of inactive movement behavior is essential for referring the right patients and tailoring movement-behavior change interventions, which could support secondary prevention of recurrent cardiovascular events. Aims: This study aimed to develop and internally validate a clinical prediction rule to identify at hospital discharge people at risk of an inactive movement behavior pattern within the first 2 years following a stroke. Methods: A prospective cohort study was conducted using data from 200 participants with a first-ever stroke (age 67.8 ± 11.5 years; 64% male; median NIHSS = 3), who were discharged to their home environment. Eligible participants were ⩾18 years, pre-stroke independent, ambulatory, and able to communicate. Movement behavior was objectively assessed within 3 weeks, and at 6 months (n = 184, 92%), 1 year (n = 175, 88%), and 2 years (n = 146, 74%) post-discharge. Movement behavior patterns were based on the amount of light and moderate-to-vigorous physical activity (PA) and prolonged sedentary bouts: “sedentary exercisers” (active), “sedentary movers” (inactive), and “sedentary prolongers” (inactive and prolonged sedentary bouts). Baseline characteristics, including demographic, stroke-related, and health-related factors, were used to identify “sedentary movers and prolongers” (step 1) and “sedentary prolongers” (step 2) by multinominal logistic regression. Results: Female sex (B = −1.03, p < 0.001), older age (B = 0.05, p < 0.001), and increased fatigue (B = 0.04, p = 0.003) predicted inactive movement behavior in the first 2 years after discharge. Inactive movement behavior with prolonged sedentary bouts was predicted by “prolonger” pattern directly after discharge (B = −3.35, p < 0.001), slower walking speed (B = 0.10, p = 0.003), and lower anxiety levels (B = −0.07, p = 0.057). The final model showed good fit (Quasi-likelihood under Independence Model Criterion (QICC) = 737.02) and acceptable discrimination (area under the curve (AUC) = 0.74). Internal validation confirmed the model’s robustness, with a shrinkage factor of 0.96. Conclusion: A clinical prediction rule to identify patients at risk of inactive movement behavior post-stroke was developed and internally validated. Early identification based on age, sex, and patient-reported fatigue can facilitate stratification for tailored behavior change interventions aimed at secondary prevention of recurrent cardiovascular events. External validation is required before clinical implementation. Data access statement: The datasets used and/or analyzed in this study are accessible from the corresponding author on reasonable request.