One-Year Prospective Study on Smartphone-Based Coefficient of Variation Analysis of Seated Stepping Movements for Fall Risk Prediction in Older Adults

Older adults with a recent fall history tend to have larger variability in stepping than those without a fall history. In this study, we examined whether variability in step timing-defined as the coefficient of variation (CV) of step timings during seated stepping exercises-can identify individuals at higher risk of falling. The CV was measured at baseline (initial assessment), and fall occurrences were tracked over one year among 58 older adults in 11 senior housing facilities participating in online exercise programs. Participants who experienced falls exhibited marginally higher CV values at baseline than those who did not fall, and those who fell specifically while walking showed significantly higher CVs compared to non-fallers (<i>p</i> = 0.035). Logistic regression analysis indicated that the CV significantly predicted walking-related falls (odds ratio = 1.24, <i>p</i> = 0.032), and receiver operating characteristic curve analysis yielded an area under the curve of 0.747, suggesting moderate discriminative ability. Including prior fall history in the model further improved predictive performance (AUC = 0.807 for overall falls and 0.925 for walking-related falls), suggesting that combining CV with prior fall history enhances predictive performance. These findings suggest that evaluating timing variability during seated stepping exercises, especially when combined with prior fall history, may be a useful indicator for predicting fall risk over the following year without exposing individuals to fall hazards during assessment.