External Validation of the TWIST Algorithm for Predicting Walking Independence Post-Acute Stroke: An Irish Prospective Cohort Study

Abstract

Abstract Background Early prediction of walking dependency post-stroke is important for clinician and patient decision-making. National stroke guidelines recommend using the ‘Time to Walking Independently after Stroke’ (TWIST) model to predict walking independence post-stroke. External validation of the TWIST model is required prior to its clinical integration. This study aimed to evaluate whether the TWIST model is a valid predictor of walking independence post-stroke in an Irish population. Methods This prospective cohort study recruited patients admitted to a tertiary hospital with an acute stroke from July-November 2023. The Trunk Control Test (TCT) and paretic-limb hip extensor strength (Medical Research Council (MRC) scale) were assessed within 3 days post-stroke. Walking independence (Functional Ambulatory Category) was measured at 6 or 12 weeks post-stroke based on TWIST algorithm predictions. Predictive performance was measured using the area under the curve (AUC), accuracy, sensitivity, specificity, positive and negative predictive value. Results Sixty-one participants (38 male, median (interquartile range) age 74 years (68-86) and National Institutes of Health Stroke Scale 7 (4-11.5)) were included. The AUC (95% CI) was 0.88 (0.81-0.96) and 0.7 (0.21-1) at 6 and 12 weeks post-stroke respectively. Predictive accuracy (95% CI) was higher at 6 weeks (88.52% (77.78-95.26)) than 12 weeks post-stroke (86.96% (66.41-97.22)). Overall, the TWIST model made accurate predictions for 83.61% (71.91-91.85) of participants. Conclusion The TWIST model demonstrated fair-good predictive performance when externally validated. However, it performed poorly at 12 weeks post-stroke for those with poor post-stroke trunk control (TCT < 40) but moderate hip extensor strength (≥ 3 MRC scale). Further research is needed to explore its utility in this subgroup. Implication of Findings The TWIST model may benefit clinicians and patients by accurately predicting walking independence post-stroke. Further research is needed to explore its utility in specific subgroups, particularly those with poor post-stroke trunk control but moderate hip extensor strength.