**1. Introduction**

Walking is a key activity to improve the quality of life and social participation in individuals with stroke sequelae [1,2]. However, studies showed that 30% of patients do not recover the ability to walk independently six months after the stroke, and nearly 65% consider that their social participation is limited [3]. The most often reported parameters in the literature to estimate an individual's capacity to walk in a social environment are the speed—commonly evaluated with a 10-m test—and the distance (d), assessed with a 6-min walk test [4]. Perry et al. demonstrated that a speed over 0.4 ms−<sup>1</sup> was the threshold defining the ability to walk in a social environment [5]. However, there may be a gap between a subject's performance in these tests and the actual capacity to walk freely due to obstacles in the environment [6,7]. This is especially true since the distances walked in social situations are significantly longer than in tests [8]. For example, Salbach et al. showed in a study including 24 cities around the world that an individual has to travel a mean distance between 57 and 98 m to go to a post office, between 77 m and 114 m to a place of worship, 260 m for a doctor appointment (SD: 78 m), and 129–381 m to go to a shopping mall [4].

To assess the individual's real walking ability outside of institutions, health professionals need reliable and practical tools that can be used in daily life situations [7,9]. The most commonly used tools to measure the walked capacity in terms of distance or number of steps are global positioning systems (GPSs), pedometers, and accelerometers [10]. On the one hand, the number of steps tends to be an unreliable parameter for medical practitioners as the step length is variable depending on the type and level of motor deficiency, which causes a lower comparability between individuals [11,12]. On the other hand, distance is more reliable, with better comparability between individuals and readily accessible data pertaining to social participation in the literature [2,4,8,13]. It can be measured accurately using a GPS, but its use is limited to outdoor situations due to the necessity of satellite synchronization [14].

Accelerometers and pedometers can be used indoors and were validated for estimations of the walked distance in healthy populations [15,16]. These systems generally use the product of the step count and the step length. Step length is evaluated by measuring the distance traveled over 20 steps [15]. However, this method remains debatable in post-stroke subjects because of the variability of a given individual's step length [17]. Furthermore, Crouter et al. reported that the accuracy of these devices sharply drops when individuals walk slower than 0.8 ms−<sup>1</sup> [15], and post-stroke patients typically tend to walk slower than healthy individuals [18]. Caroll et al. indeed demonstrated in a population of 51 post-stroke individuals that pedometers placed on the hip or around the neck were unable to detect steps when walking slower than 0.5 ms−<sup>1</sup> [19]. Other types of sensors are, however, able to make accurate estimations of the number of steps even in slow-walking individuals. For instance, Fulk et al. demonstrated that the accelerometer-based step activity monitor had an estimation error of five steps in a 2-min walk test with an intraclass coefficient correlation (ICC) of 0.97, even in subjects walking slower than 0.5 ms−<sup>1</sup> [20]. Additionally, Klassen et al. demonstrated that the location of the sensor had a direct impact on the results; a triaxial accelerometer reported a standard error lower than 15% when placed on the ankle, while its standard error was greater than 80% when placed on the hip in a study including 43 post-stroke subjects. The impact of the sensor's position was even greater in individuals who walked slowly (<0.5 ms<sup>−</sup>1) [21].

Naturally, when confronted with a large number of possible devices and positions recommended by the manufacturer, medical practitioners and users are at a loss when it comes to choosing the right sensor and position to obtain the best estimation of walked distance. Therefore, it seems essential to check the validity of each option to point toward a global recommendation in post-stroke populations.

The objective of this study was to evaluate the validity of the walked distance estimations by pedometers and accelerometers and to evaluate the impact of their position (ankle, hip, wrist/affected, unaffected side) in a post-stroke population walking at a comfortable speed.
