Association Between Average Step Width and History of Falls in Patients with Moderate-to-Severe Knee Osteoarthritis: A Cross-Sectional Gait Analysis

Purpose To investigate the association between average step width and a history of falls in patients with moderate-to-severe knee osteoarthritis (KOA), and to determine whether this association is independent of Kellgren-Lawrence (KL) grade and other potential confounders. Methods A cross-sectional study was conducted. A total of 120 patients with KL grade III–IV KOA were recruited from the orthopaedic outpatient and inpatient departments of a tertiary hospital between June and December 2026, including 60 patients with a history of falls in the past year and 60 without. Spatiotemporal, kinetic, and kinematic parameters were collected using the Footscan® plantar pressure gait analysis system. The primary exposure was average step width, and the primary outcome was a history of falls within the past year. Two-way ANOVA was used to examine the main and interaction effects of KL grade and fall history on gait parameters. Firth's penalized-likelihood logistic regression was used to evaluate the independent association between average step width and fall history, adjusting for age, sex, BMI, KL grade, and walker use. Internal validation was performed using bootstrap and leave-one-out cross-validation (LOOCV), and decision curve analysis (DCA) was used to assess exploratory clinical net benefit. Results Average step width was significantly larger in the fall history group than in the no-fall group (approximately 22.9 cm vs. 8.8 cm, P < 0.001), accompanied by slower gait speed, shorter step length, and prolonged double support time. Two-way ANOVA revealed significant interaction effects between KL grade and fall history for several gait parameters (P < 0.05). After adjusting for confounders, average step width remained independently associated with a history of falls (OR = 2.97, 95% CI: 1.71–8.49, P < 0.001). Internal validation showed high discrimination, with an LOOCV AUC of 0.956, a median bootstrap AUC of 0.960, a sensitivity of 100%, and a specificity of 91.7%; however, due to quasi-complete separation of step width between groups, these performance metrics may be overestimated. DCA suggested a certain net benefit, but this was an exploratory analysis within the same cohort. Conclusions In patients with moderate-to-severe KOA, average step width is independently associated with a history of falls and may serve as a low-cost, easily accessible candidate associative indicator. Given the cross-sectional design, retrospective collection of fall history, and quasi-complete separation of step width, it cannot be inferred that average step width predicts future falls or has a causal relationship. The reference threshold of approximately 18 cm in this sample should only be considered for clinical risk stratification after multicentre, prospective external validation.