External validation and updating of the Early Prediction of Functional Outcome after Stroke model

This study investigates whether the external validation and updating for the EPOS model enhances the predictive accuracy of ambulatory independence post-stroke.

This study employs a prospective cohort design. The study population consisted of 487 persons diagnosed with stroke and hospitalized at Nerima-Hikarigaoka Hospital between October 2021 and November 2023. The predictors were collected at the time seven days post-stroke. The assessment of physical function included the sitting balance item of the Trunk Control Test, the lower extremity subscale of the Motricity Index, and Functional Ambulation Category (FAC). Additionally, brain imaging findings were assessed, including the hemisphere of onset, site of onset, and Fazekas scale. The clinical outcome was the FAC at six-months post-stroke, which was collected via a questionnaire-based survey. Statistical analysis was performed to evaluate the external validation and update of brain imaging findings for the EPOS model. The updating of brain imaging findings factor was tested with a univariable logistic regression model. The discrimination was evaluated using the area under the curve (AUC) in receiver operating characteristic (ROC) curves. Calibration was evaluated by examining calibration plots.

The final cohort comprised 41 persons (77.1±12.2 years). The Fazekas scale was extracted as a factor influencing ambulatory independence at the six-months post-stroke [odds ratio (OR) = 0.51, 95% confidence intervals (CI) = -1.30 - -0.14, p value = 0.02].　In the AUC for the ROC curve at six-months post-stroke, the discrimination for external validation of the EPOS model was 0.75 (95% CI = 0.60 - 0.90), and updating of the EPOS model yielded a discrimination value of 0.85 (95% CI = 0.71 - 0.98) (p  0.05). In the calibration plot, external validation of the EPOS model and updating of the EPOS model indicated a tendency for the predicted probability to align with the observed proportion at higher and lower to moderate predicted probabilities, respectively.

Updating for the EPOS model enhanced the accuracy of predicting ambulatory independence at the six-months post-stroke. It is therefore reasonable to posit that updating the EPOS model has the potential to serve as an effective tool for predicting long-term ambulatory functional independence in individuals who have experienced a decline in physical function post-stroke.

The use of prognostic models based on the assessment of physical function and brain imaging findings in the acute stroke phase is important to capture the expected improvement in patient physical function.