Assessment of the metrological properties of the spatiotemporal gait parameters is a key point because they are used as a monitoring criterion of post-stroke recovery and as outcome in interventional studies. The objective of the study was to explore the relevance of gait assessment at maximal gait speed (MGS) rather than spontaneous gait speed (SGS) for post-stroke subjects through (1) the study of between and within-day reliability data of spatiotemporal gait parameters at SGS and MGS; and (2) a comparison of the direct measurement of velocity at MGS versus its estimate. It was also designed to (3) highlight the importance of considering the distribution of the random error to use an adequate expression of the measurement error (MErr).

Spatiotemporal gait parameters – velocity, walk ratio, span acceleration, and spatiotemporal asymmetries indexes – from 30 post-stroke subjects (mean±SD age, 51.6±16.2 years; median [range] time since stroke, 4.4 [1 to 101] months; 60% of male; 70% using a walking assistive device) were collected using the Locometer (Satel, Blagnac, France) on three times, about 7 days (between-day) and 30minutes (within-day) apart, at SGS and MGS. The minimal metrically detectable change with a 95% confidence interval was computed in original unit, or in percentage (MDCp) if the distribution of random error was heteroscedastic.

Reliability was better at MGS than at SGS, and for within than for between-day test-retest. MDCp for velocity were equal to 36.2%/16.2% at SGS for between/within-day test–retest, and 21.2% at MGS for between-day test-retest. Gait velocity at MGS should be measured and not estimated because of unacceptable error of prediction in 70% of cases. An inappropriate expression of MErr without taking into account the random error distribution leads to over-/underestimate of the significant modifications of gait parameters.

Data from the present study allow for a suitable interpretation of spatiotemporal gait parameters modification for post-stroke subjects. We highlighted the necessary gait assessment at MGS, being the most likely to show differences related to therapeutic intervention for post-stroke subjects and showed that the velocity at MGS should be measured and not predicted.