External Validation of the EPOS Prediction Model for Independent Gait After Stroke

Accurate gait recovery prognosis is crucial for guiding stroke rehabilitation. The Early Prediction of functional Outcome after Stroke (EPOS) model predicting independent gait 6 months poststroke, was previously validated in 2 small, homogenous samples of first-ever stroke patients without prestroke disability. This study aimed to externally validate the EPOS model in a heterogeneous stroke sample.

A prospective longitudinal cohort study conducted in 2 Swiss acute stroke units. Eligible patients could not walk independently within 72 hours poststroke, regardless of stroke characteristics or prestroke disability. Sitting balance and paretic leg strength were assessed within 72 hours. Independent gait was evaluated at 3 months. The model's predictive performance was assessed using multiple metrics.

Two hundred eighty patients were analyzed (median age 76 years; 42% female). The model showed acceptable fit (Brier score 0.14, 95% confidence interval [CI] 0.11-0.18) and acceptable discrimination (area under the receiver operating characteristic curve 0.74, 95% CI 0.67-0.81). At the 0.5 threshold, sensitivity was 0.98 (95% CI 0.95-0.99) and specificity 0.32 (95% CI 0.21-0.45). Decision curve analysis indicated clinical usefulness across thresholds from 0.28 to 0.90.

The EPOS model demonstrated moderate predictive performance for regaining independent gait at 3 months when applied within 72 hours poststroke in a heterogeneous cohort. A high-sensitivity threshold is recommended, classifying only patients lacking sitting balance and hemiplegic leg strength as unlikely to regain gait independence, preventing missing recovery potential. As a next step, the model's impact on decision-making, patient outcomes, and healthcare resources should be evaluated in comparable settings.

NCT06438770, NCT05039047.