Gait Speed as a Predictor of Fall Risk: An AI-Based, Risk-Stratified Study in Specialist Falls-Clinic and Community-Dwelling Older Adults

Background Gait speed is an established biomarker of health and function in older adults, yet its clinical utility in falls assessment remains under-used. We hypothesised that slow gait speed would be associated with falls across both specialist falls clinic attendees and healthy controls. Methods This was a prospective cohort study that used GaitKeeper, an artificial intelligence smartphone application that analyses spatiotemporal gait metrics. All statistical analysis was performed in SPSS Statistics (version 29.0.20). Results Fifty participants (mean age 78.7 ± 6.0; 46% males) were stratified into three subgroups: high-risk fallers (≥1 fall in twelve months; n=26), low-risk fallers (falls-clinic attendees with no recent falls; n=6) and healthy controls (no falls; n=18) that were similar in age (p=0.631) and gender (p=0.070). Charlson Comorbidity Index (CCI median [IQR]: 4 [3–5] vs 6 [4–7] vs 3 [3–4]; H₂=16.25,p< 0.001) and clinical frailty score (CFS median [IQR]: 4 [2–5] vs 3.5 [2–5] vs 2 [2–3]; H₂=6.61,p=0.037) were higher in both falls-risk groups than in healthy controls. Mean gait speed differed (1.02 ± 0.36m/s vs 1.18 ± 0.43m/s vs 1.29 ± 0.24m/s; p< .01), with high-risk fallers walking slower than healthy controls (Δ –0.268m/s; p=0.031). Adding gait speed to a model of CCI and CFS increased Nagelkerke R2 from 0.115 to 0.169 (Δ=0.054) and AUC from 0.663 to 0.709, showing gait speed adds predictive power for risk of falls. Slow-gait prevalence (<0.8m/s) was 23.1% in high-risk fallers, 16.7% in low-risk fallers, and absent in controls (p=0.093). Participants with gait speeds <0.8 m/s had higher CCI (median 6 [4–9] vs 4 [2–8]; U=84.5,p=0.057) and higher CFS (median 5 [3–6] vs 2 [1–4]; U=31.5,p< 0.001) than those walking ≥0.8 m/s. Conclusion Gait speed is a pragmatic clinical filter for identifying the highest-risk subgroup in complex older adult cohorts.