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

Area of Science:

• Neurology
• Rehabilitation Medicine
• Clinical Prediction Models

Background:

• The Time to Walking Independently after Stroke (TWIST) tool predicts post-stroke walking independence.
• External validation is crucial for assessing the TWIST tool's clinical applicability and generalizability.

Purpose of the Study:

• To perform a geographic external validation of the TWIST prediction tool in a multicenter prospective cohort.
• To assess the TWIST tool's performance in predicting walking independence in a Japanese stroke patient population.

Main Methods:

• Adult stroke patients with lower-limb weakness and inability to walk independently were enrolled.
• TWIST scores were calculated at 1 week post-stroke using age, knee-extension strength, and Berg Balance Scale.
• Model fit, calibration, and discrimination were assessed at predicted timepoints (4, 6, 9, 16, 26 weeks).

Main Results:

• The validation cohort included 145 patients; 60.7% and 72.4% achieved walking independence by 9 and 26 weeks, respectively.
• The TWIST tool showed good discrimination (c-statistic >0.75) and moderate fit (Nagelkerke R² 0.28-0.49).
• Calibration analysis revealed consistent over-prediction across all timepoints, particularly with higher TWIST scores early and lower scores later.

Conclusions:

• The TWIST prediction tool demonstrated optimistic predictions of walking independence in Japanese stroke patients.
• The tool's over-prediction suggests potential limitations in its current calibration for this population.
• Further validation studies are essential to refine the TWIST tool's clinical utility and generalizability in diverse stroke populations.