Prediction model for timing of independent walking on varied surface conditions in patients with early subacute stroke: A multicenter prospective study

PURPOSE: This study aimed to develop and internally validate a model predicting when patients with early subacute stroke achieve independent walking on multiple types of surfaces. METHODS: In this prospective multicenter observational study, adults admitted to rehabilitation wards for cerebral infarction or hemorrhage (April 2023-October 2025) were assessed. Walking independence was evaluated from 6 to 24 weeks after stroke onset using the Functional Ambulation Categories (FAC); FAC 5 indicated independent walking on multiple types of surfaces. A Cox proportional hazards model was developed using prespecified admission predictors: age, knee extension, Berg Balance Scale, Mini-Mental State Examination, visuospatial function, and continence. Missing data were multiply imputed. Internal validation used 200 bootstrap resamples. RESULTS: Of 367 screened patients, 168 were analyzed (mean age 74.9 years). The cumulative achievement rate for FAC 5 plateaued at 51% by week 20. The model satisfied proportional hazards assumptions. Time-dependent area under the curve values ranged from 0.86 to 0.92. Calibration was favorable, with a slope of 0.92, expected/observed ratios of 1.00-1.03, and integrated calibration indices of 0.04-0.06. CONCLUSIONS: A model based on routine assessments in the early subacute phase showed favorable internal performance in predicting the timing of independent walking across multiple surface conditions after stroke; however, external validation is required before clinical implementation.