Gait variability predicts real-life falls in high-functioning stroke survivors

Background: While over 60 % of adults with stroke fall each year, the risk is greatest in high-functioning individuals with mild motor impairments and greater physical mobility. We lack sensitive predictors of falls in this population. Therefore, our study aimed to determine the relative contribution of gait variability and widely used tests of balance and mobility in predicting real-life falls in high-functioning adults with stroke. Methods: Twenty-four adults with stroke who had the ability to walk independently, Fugl-Meyer lower-extremity score of >= 19/34, and Frenchay Activities Index >= 16/45 performed overground walking, Timed-up and go, and Berg balance scale. We quantified the gait speed, and gait variability for stride length and stride time. We recorded the history of falls in the past one year. Findings: The incidence rate of past falls was 50 %. Stride length variability and Berg balance scale score were associated with previous falls in univariate analyses and were subsequently included in the multivariate model. Multivariate analyses showed that only stride length variability significantly predicted past falls (OR = 2.73, 95 % CI 1.05-7.08, p = 0.03). A cut-off of 3.98 % for stride length variability had 75 % sensitivity and 91.7 % specificity in predicting previous falls (AUC = 0.83, 95 % CI 0.64-1.00, p < 0.001). Interpretation: In high-functioning adults with stroke, stride length variability during overground walking is a strong predictor of the past incidence of falls compared with traditional balance and mobility tests. Our findings highlight the importance of gait variability in accurately determining fall risk among high-functioning poststroke individuals.