The role of cognitive function in age-related changes in spatiotemporal gait: Findings from a community-based...

Gait alterations are common in cognitive impairment, but how cognitive decline uniquely influences gait independent of chronological aging remains unclear. This study aimed to characterize spatiotemporal gait profiles across the dementia spectrum and to identify the mediating role of cognitive function.

This cross-sectional study included 289 community-dwelling older adults categorized into cognitively normal (CN, n = 58), subjective cognitive decline (SCD, n = 152), and cognitive impairment (CI, n = 79) groups. Eighteen gait features captured via smartphone-based computer vision were classified into spatial, temporal, and combined spatiotemporal components. Permutation entropy (PE) and autocorrelation quantified gait complexity. Generalized additive models constructed age-related gait profiles, and structural equation modeling (SEM) evaluated global cognitive mediation effects.

Stride speed differed significantly across groups under single- and dual-task conditions. Spatial and combined spatiotemporal gait measures showed age-related reductions across all groups. Under dual-task conditions, the CI group exhibited less favorable temporal gait profiles than the CN and SCD groups. Lower cognitive status was associated with reduced spatial complexity, whereas temporal gait measures showed greater irregularity and weaker sequential dependence. SEM showed significant direct and cognition-specific indirect associations of age with the spatial (stride length) and combined spatiotemporal (stride speed) domains. For temporal components (stride time and single-support phase), age showed significant cognition-specific indirect associations but no detectable residual direct associations.

Temporal gait components were associated with age primarily through cognition-related pathways and may serve as candidate digital markers of cognitive impairment, although longitudinal validation is required. Interpretation is limited by the cross-sectional design and the restricted representation of individuals with moderate-to-severe dementia.