Clinical variables provide the main predictive signal for early in-hospital rehabilitation triage after small-artery occlusion stroke: development and internal validation of an interpretable model.

Background: Small artery occlusion stroke is often mild, but some patients need rehabilitation during acute hospitalization. We developed and internally validated an interpretable model to assist early in-hospital rehabilitation screening and triage and tested whether white matter hyperintensity burden and inflammatory-metabolic biomarkers improved prediction beyond routine clinical variables. Methods: Consecutive adults with acute ischemic stroke classified as small artery occlusion were identified from a single-centre stroke registry and electronic health records. Early rehabilitation need was assessed within 48 h after admission by trained rehabilitation physicians using a structured protocol. The prespecified primary model used clinical variables. Extended models added modified Fazekas score, inflammatory metabolic biomarkers, or both. Eight regression and machine learning algorithms were compared. Internal validation used stratified tenfold cross validation with nested fivefold tuning. Performance was assessed using discrimination, precision recall, calibration, and decision curve analysis. Results: Among 411 patients, 155 had early rehabilitation need (37.7%). The clinical-variable model performed best, with an area under the receiver operating characteristic curve of 0.682 and an area under the precision recall curve of 0.591. Adding modified Fazekas score, inflammatory metabolic biomarkers, or both did not improve performance. Logistic regression performed comparably to complex algorithms. In the admission fixed-landmark sensitivity analysis, the clinical-variable model yielded an AUROC of 0.623, which decreased to 0.507 after variables directly reflecting acute neurological deficits were excluded. Leading predictors reflected neurological severity, early functional impairment, and care pathway characteristics. Conclusion: Early rehabilitation need was common among patients with small-artery occlusion stroke during acute hospitalization. Routine clinical variables provided the dominant predictive information, whereas white matter hyperintensity burden and inflammatory-metabolic biomarkers added no stable incremental value.