External Validation of the Walking Independence Prognostic Model TWIST Score After Stroke: A Multicenter,Prospective Cohort Study

Background

The Time to Walking Independently after Stroke (TWIST) prediction tool is designed to predict the number of post-stroke weeks at which patients are expected to achieve walking independence. The external validation of the TWIST tool’s clinical applicability and generalizability has been desired.

Objective

We performed a geographic external validation of TWIST prediction in a multicenter prospective cohort.

Patients and Methods

Adult post-stroke patients with lower-limb weakness and inability to walk independently were enrolled. Each patient’s TWIST score (max. score: 4 points) was calculated using age, knee-extension strength, and the Berg Balance Scale score at 1 week post-stroke. The TWIST score predicts independent walking by 4, 6, 9, 16, or 26 weeks post-stroke, and we calculated the overall fit, calibration, and discrimination at each of these timepoints to assess the model’s performance.

Results

The validation cohort consisted of 145 patients (median age 73 years, 44% women, 36% moderate-severe stroke). At 9 and 26 post-stroke weeks, 60.7% and 72.4% of the patients achieved walking independence. The overall fit explained a moderate proportion of the outcome’s variance (Nagelkerke R² = 0.28-0.49), and the discrimination performance was good (c-statistic >0.75). Calibration performance showed over-prediction at all timepoints (calibration-in-the-large = −1.62 to −0.34). Higher TWIST scores (3 or 4 points) over-predicted early post-stroke; lower TWIST scores (1 or 2 points) became increasingly over-predictive over time.

Conclusions

The TWIST prediction tool optimistically predicted walking independence in Japanese patients with stroke. Further validation studies are necessary to assess this tool’s precise clinical impact and generalizability.

Clinical Trial Registration URL:

https://www.umin.ac.jp/

UMIN000050588