Gait Posture. 2026 Jun 18;130:110264. doi: 10.1016/j.gaitpost.2026.110264. Online ahead of print.
ABSTRACT
BACKGROUND: Independent outdoor ambulation is a key rehabilitation goal after stroke, but it is unclear whether instrumented gait analysis adds prognostic information beyond conventional clinical measures.
RESEARCH QUESTION: Do spatiotemporal gait parameters provide incremental predictive value beyond clinical assessments-independent of admission Functional Ambulation Category (FAC)-for outdoor ambulation in subacute stroke patients?
METHODS: We retrospectively analysed 137 subacute stroke inpatients with admission FAC 2-3 (89 outdoor, 48 indoor-only ambulators at discharge); one patient was excluded for an implausible step time. Admission FAC was excluded from candidate predictors to avoid overlap with the outcome. Hierarchical logistic regression compared a clinical model (Motricity Index [MI], time since onset) with one adding GAITRite-derived spatiotemporal parameters, using likelihood-ratio and DeLong tests with Firth-penalised sensitivity analysis.
RESULTS: Model 1 achieved near-perfect discrimination (AUC = 0.995, 95% CI 0.984-0.999). Adding affected-side single- and double-support percentages (Model 2) significantly improved fit (likelihood-ratio χ² = 9.39, p = 0.009; AUC = 0.998); the AUC difference was not significant by DeLong’s test (p = 0.262). Firth-penalised analyses produced concordant, stable coefficients. MI and gait velocity were the strongest single predictors (both AUC = 0.981); bootstrap optimism was ≤ 0.002.
SIGNIFICANCE: Beyond a near-perfect clinical model, affected-side support-phase parameters add statistically detectable model information (improved fit and calibration) rather than a clinically decisive gain in discrimination, which is constrained by a ceiling effect. Their value is best understood as quantifying paretic-limb weight-bearing and balance-related gait quality not captured by bedside scales, in subacute stroke inpatients with admission FAC 2-3. The Youden cut-offs (MI ≥ 40, velocity ≥ 32.1 cm/s, single support ≥ 29.7%) are hypothesis-generating and require external validation before clinical use.
PMID:42335499 | DOI:10.1016/j.gaitpost.2026.110264
