Step width as a robust predictor of fall risk in patients with moderate-to-severe knee osteoarthritis: A gait-based prediction model

<title>Abstract</title> <p>Objective Patients with knee osteoarthritis (KOA) have a significantly increased risk of falls, yet the relationship between gait parameters and fall risk in moderate-to-severe KOA remains unclear. This study aimed to identify independent gait predictors of falls and develop a simple clinical screening tool. Methods A total of 120 patients with Kellgren–Lawrence grade III–IV KOA (60 fallers, 60 non-fallers) were enrolled. Spatiotemporal, kinetic, and kinematic parameters were collected using the Footscan® plantar pressure system. Lasso regression with 10-fold cross-validation was used for variable selection, and Firth penalized-likelihood logistic regression was applied to address complete separation. Model stability was evaluated by bootstrap internal validation (1000 resamples) and leave-one-out cross-validation (LOOCV), with clinical utility assessed via decision curve analysis (DCA). Results Mean step width emerged as the sole independent predictor of falls (univariate Firth regression: OR = 2.24, 95% CI: 1.58–3.91, P &lt; 0.001). The model demonstrated excellent discriminative ability (LOOCV AUC = 0.956, accuracy = 95.8%, sensitivity = 100%, specificity = 91.7%), with 100% sign consistency in bootstrap validation. A threshold of mean step width &gt; 18 cm effectively identified high-risk individuals. DCA confirmed a positive net benefit across clinically relevant threshold probabilities. Conclusion Mean step width is a robust, low-cost gait-based predictor of fall risk in moderate-to-severe KOA. The proposed threshold of &gt; 18 cm enables rapid clinical screening and community-based risk stratification.</p>