A patient-specific predicting tool for functional recovery after stroke.


Objectives: Existing prognostic models of stroke recovery are rare. This study aimed to develop and validate a patient-specific model for predicting functional recovery up to 1 year post-stroke. Methods: Data were from 495 patients recruited from the population-based South London Stroke Register. Functional assessments were performed using the Barthel Index (BI) at 1, 2, 3, 4, 6, 8, 12, 26 and 52 weeks after stroke. Multilevel growth models were used to predict BI trajectories, recovery curves, allowing for day-to-day and between-patient variation. Cross-validation procedures were used for selecting strong predictors of BI model parameters. The predictive performance of the recovery curves was validated using 10-fold internal cross-validation. The model was also validated for classification accuracy of poor BI (<8) outcomes at 3 and 12 months using 2 external samples totaling 1830 stroke patients. Results: Mean age was 71 years, 51% were females and 24% died within the first year. Age, gender, NIH Stroke Scale, Glasgow Coma Scale and stroke subtype were identified as independent predictors of BI. Accuracy of the recovery curves model were satisfactory, with a root mean square deviation less than 2.3 BI points at 3 and 12 months. The model was highly effective at classifying patients as likely to have poor outcome or not at 3 and 12 months, which is a clinically useful distinction. Conclusions: The model predicts functional recovery over time after stroke and could potentially aid clinicians in early identification of and intervention with patients at risk of poorer than expected functional outcome.

Proc. European Stroke Conference