Answers are grounded in this paper and cite their sources.
Objectives: This study aimed to develop an interpretable decision tree rule for predicting early attainment of supervised overground gait within 4 weeks after initiating robot-assisted gait training (Welwalk [WW-1000 and WW-2000; Toyota Motor, Toyota, Aichi, Japan]) in patients with post-stroke hemiparesis and to compare its performance and value with logistic regression models. Methods: This retrospective single-center study analyzed patients with first-ever supratentorial stroke and hemiparesis who underwent Welwalk training (n = 89). The outcome was attainment of supervised 10-m gait (Gait Ability Assessment [GAA] = 5) within 4 weeks from training initiation. A classification and regression tree (CART) was developed from initiation assessments and evaluated using 10-fold cross-validated out-of-fold (OOF) predictions. Two prespecified logistic regression models were evaluated similarly. Decision curve analysis (DCA) assessed net benefit across threshold probabilities of 0.20–0.60. Results: Thirty patients (33.7%) attained supervised gait within 4 weeks. CART yielded a 2-level rule: baseline GAA (cutoff 2) and, among patients with GAA ≥2, Stroke Impairment Assessment Set (SIAS) lower-limb total (cutoff 2). Early-attainment rates were 86.2% (25/29) for GAA ≥2 and SIAS ≥2, 0.0% (0/9) for GAA ≥2 and SIAS <2, and 9.8% (5/51) for GAA <2. The cross-validated CART performance was: area under the receiver operating characteristic curve (AUC), 0.804; accuracy, 0.888; sensitivity, 0.833; and specificity, 0.915. Logistic regression AUCs were 0.842 and 0.837. CART showed a net benefit comparable to and partly higher than that of logistic models across 0.20–0.60 probabilities. Conclusions: A minimal rule combining baseline gait ability and paretic lower-limb motor function stratified early attainment of supervised gait within 4 weeks after Welwalk initiation and may support timely ward-level decision making.
No takes yet. Share an insight, caveat, or question.
IKEUE et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: