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

Area of Science:

• Neurology
• Rehabilitation Medicine
• Clinical Prediction Models

Background:

• Accurate prognosis of gait recovery is vital for tailoring stroke rehabilitation strategies.
• The Early Prediction of functional Outcome after Stroke (EPOS) model was previously validated in limited homogenous stroke populations.
• External validation in a heterogeneous stroke cohort is necessary to assess the EPOS model's generalizability.

Purpose of the Study:

• To externally validate the EPOS model's predictive performance for independent gait recovery in a heterogeneous stroke population.
• To assess the EPOS model's accuracy and clinical utility in a real-world, diverse patient sample.

Main Methods:

• A prospective longitudinal cohort study involving 280 acute stroke patients unable to walk independently within 72 hours.
• Assessment of sitting balance and paretic leg strength within 72 hours post-stroke.
• Evaluation of independent gait at 3 months, with predictive performance assessed using Brier score, AUC, sensitivity, specificity, and decision curve analysis.

Main Results:

• The EPOS model demonstrated acceptable fit (Brier score 0.14) and discrimination (AUC 0.74) in the heterogeneous cohort.
• High sensitivity (0.98) was observed at a 0.5 threshold, indicating few false negatives for gait recovery.
• Decision curve analysis suggested clinical usefulness across a range of probability thresholds (0.28-0.90).

Conclusions:

• The EPOS model offers moderate predictive performance for 3-month independent gait recovery in diverse stroke patients treated within 72 hours.
• A high-sensitivity threshold is recommended to maximize the identification of patients likely to regain gait independence.
• Further research should evaluate the model's impact on clinical decision-making and patient outcomes.