Gait & Posture · Oct 27, 2026

Smartphone assessment of postural sway and prospective falls in older adults: A large community-based study

By Silsupadol P, O'Brien SR, Bautista S et al.

BACKGROUND: Impaired postural control is a key risk factor for falls in older adults. However, laboratory-based postural sway assessments are costly and difficult to implement in real-world environments. Smartphones may offer a portable, low-cost, and easy-to-administer alternative for quantifying postural sway in large-scale community settings. RESEARCH QUESTION: Can a smartphone-embedded accelerometer quantify postural sway, distinguish retrospective fallers from non-fallers, and predict future falls over 12 months? METHODS: A total of 580 adults aged ≥ 55 years were assessed between 2023 and 2026. Fall history was collected, with prospective falls monitored over 12 months. Sway area and path length were measured using the Physiological Profile Assessment swaymeter during four standing conditions (eyes open/closed, firm/foam surfaces). Concurrently, smartphone-derived accelerometer data were used to compute acceleration trajectory area and path length. Pearson correlations, two-way ANOVA, and multivariable logistic regression examined relationships between instruments, group differences, and fall prediction. RESULTS: Smartphone-derived measures were strongly correlated with swaymeter outcomes for area (r = 0.713, p < 0.001) and path length (r = 0.803; p < 0.001). Both systems demonstrated significant group effects for sway area (p < 0.05), and smartphone-derived acceleration trajectory path length differed significantly between fallers and non-fallers (p = 0.036). Adding smartphone-derived acceleration trajectory path length to established risk factors increased Nagelkerke R² from 0.142 to 0.150, but the acceleration measure was not a significant predictor of future falls. SIGNIFICANCE: Smartphone-based accelerometry quantified postural sway and distinguished retrospective fallers from non-fallers but did not improve future fall prediction beyond established risk factors.

doi.org/10.1016/j.gaitpost.2026.110680PubMed 42956683Biomarkers