An accelerometry-based marker of gait automaticity is lower in older adults reporting recent falls

Falls affect one-third of older adults annually, yet traditional gait metrics poorly reflect underlying motor control. The Attractor Complexity Index (ACI), computed as long-term divergence (Lyapunov) exponent from trunk acceleration dynamics, has been validated as a marker of gait automaticity sensitive to attentional demands and age-related changes. This study investigated whether ACI discriminates between older adults with and without recent fall history. Sixty community-dwelling older adults (22 recent fallers, 38 non-fallers; mean age 76.1±6.4 years) performed 410 m of indoor walking at preferred speed wearing triaxial accelerometers at the lumbar region and right foot. Four gait complexity markers were compared between groups: detrended fluctuation analysis (DFA) scaling exponent (α), and three ACI measures (norm, anteroposterior, vertical). One-sided Welch's t-tests and standardized mean difference (SMD) with 90% bootstrap confidence intervals were used. All complexity markers showed significantly reduced values in recent fallers compared to non-fallers: ACI-AP (SMD=−0.51, p=0.026), DFA-α (SMD=−0.48, p=0.021), ACI-V (SMD=−0.44, p=0.041), and ACI-N (SMD=−0.42, p=0.049). These effect sizes were comparable to established clinical benchmarks for fall risk discrimination. Gait complexity markers derived from accelerometry effectively discriminate older adults with recent fall history, supporting their interpretation as measures of gait automaticity and their potential relevance for fall risk assessment. This study demonstrates that ACI, computationally derived from continuous trunk acceleration without stride detection, provides discrimination of fall history comparable to established clinical benchmarks, offering a promising approach for unsupervised gait monitoring. • Gait complexity markers are reduced in older adults with recent fall history • Trunk acceleration-derived indices discriminate fallers with medium effects • Effect sizes match Timed Up and Go test benchmarks for fall discrimination • Reduced complexity reflects shift from automatic to executive gait control • Results support accelerometry for unsupervised fall risk assessment