Gait variability and stability measures: Minimum number of strides and within-session reliability

Background: Several methods are proposed in the literature for the quantification of gait variability/stability from trunk accelerations. Since outputs can be influenced by implementation differences, reliability assessment and standardization of implementation parameters are still an issue. The aim of this study is to assess the minimum number of required strides and the within-session reliability of 11 variability/stability measures. Method: Ten healthy participants walked in a straight line at self-selected speed wearing two synchronized tri-axial Inertial Measurement Units. Five variability measures were calculated based on stride times namely Standard deviation, Coefficient of variation, Inconsistency of variance, Nonstationary index and Poincaré plot. Six stability measures were calculated based on trunk accelerations namely Maximum Floquet multipliers, Short term/long term Lyapunov exponents, Recurrence quantification analysis, Multiscale entropy, Harmonic ratio and Index of harmonicity. The required minimum number of strides and the within-session reliability for each measure were obtained based on the interquartile range/mean ratio. Measures were classified in five categories (namely excellent, good, average, poor, and very poor) based on their reliability. Results: The number of strides required to obtain a reliable measure was generally larger than those conventionally used. Variability measures showed average to poor reliability, while stability measures ranged from excellent to very poor reliability. Conclusion: Recurrence quantification analysis and multiscale entropy of trunk accelerations showed excellent reliability and a reasonable number of required strides. Based on these results, these measures should be taken into consideration in the assessment of fall risk. © 2014 Elsevier Ltd