Development of a model to predict gait independence in individuals with very severe gait disorder due to subacute hemiparetic stroke

Objectives: We aimed to develop and validate a model to predict gait independence at discharge from inpatient rehabilitation in individuals with subacute hemiparetic stroke who have very severe gait disorder. Methods: Overall, 298 individuals with subacute hemiparetic stroke and completely dependent gait were selected in one hospital as the training cohort. Seventy-seven individuals were selected in another hospital as the validation cohort. The prediction model was developed using multivariable logistic regression analysis, with individual characteristics selected based on a p-value threshold (<0.10) in the training cohort. Sensitivity, specificity, and area under the curve of the receiver operating characteristic curve were calculated in the training cohort, and external validation was conducted using the validation cohort. Results: In total, 102 (34.2%) and 40 (52.0%) individuals in the training and validation cohorts achieved independent gait while hospitalized, respectively. The prediction model factors were age, days from onset to admission, stroke type, affected side, severity of paresis, unaffected side function, and cognitive function. The sensitivity, specificity, and area under the curve in the training cohort were 0.81, 0.80, and 0.88, respectively. Corresponding values in the validation cohort were 0.82, 0.70, and 0.83, respectively. Conclusions: A model combining age, days from onset to admission, stroke type, affected side, severity of paresis, unaffected side muscle strength, and cognitive function effectively predicted gait independence at discharge in individuals with very severe gait disorder due to subacute hemiparetic stroke.