Predictive accuracy of gait speed for falls: an individual participant data meta-analysis

Background Gait speed is included in the World Falls Guidelines (WFG) fall risk algorithm, yet its ability to discriminate fallers from non-fallers remains unclear. This individual participant data meta-analysis examined the discriminative ability of the WFG-recommended cut point (<0.8 m/s) and the performance of a higher cut point (<1.0 m/s) for predicting falls in community-dwelling older adults and clinical populations at elevated risk of falls. Methods Individual data from 28 studies with a quantitative measure of gait speed and at least three months of prospectively reported falls were analysed using modified Poisson regression and Negative Binomial regression, followed by random-effects meta-analyses. Results Eight studies involving community-dwelling older adults (n = 3627) and twenty studies involving clinical populations (n = 3981) were included. Walking at < 0.8 m/s was associated with increased risk of falling (Relative Risk: 1.27 (95%CI 1.17 – 1.38)) and fall rate (Incidence Rate Ratio: 1.54 (95%CI 1.34 – 1.77)). Diagnostic accuracy was modest (58%, specificity 77%, sensitivity 35%), with consistent findings across planned subgroup analyses for population, fall history, and sex. Analyses using the < 1.0 m/s cut point produced similar effect sizes and accuracy metrics but identified a larger proportion of fallers in both community-dwelling (30.1% vs. 8.7%) and clinical populations (66.9% vs. 46.0%) compared to the < 0.8 m/s cut point. Conclusion Slower gait speed is associated with an increased risk and rate of falling across both population groups, but discriminative accuracy is low. While the < 0.8 m/s threshold shows consistent associations, a < 1.0 m/s cut point may be more clinically useful in community-dwelling and clinical settings because it identifies more fallers.