Walking test procedures influence speed measurements in individuals with chronic stroke

Abstract

Background:

Walking speed measurements are clinically important, but varying test procedures may influence measurements and impair clinical utility. This study assessed the concurrent validity of walking speed in individuals with chronic stroke measured during the 10-meter walk test with variations in 1) the presence of an electronic mat, 2) the speed measurement device, and 3) the measurement distance relative to the total test distance.

Methods:

Twenty-five individuals with chronic stroke performed walking tests at comfortable and maximal walking speeds under three conditions: 1) 10-meter walk test (without electronic mat) measured by stopwatch, 2) 10-meter walk test (partially over an electronic mat) measured by software, and 3) 10-meter walk test (partially over an electronic mat) measured by stopwatch. Analyses of systematic bias, proportional bias, and absolute agreement were performed to determine concurrent validity between conditions.

Findings:

Walking speeds were not different between measurements (P≥0.11), except maximal walking speed was faster when speed was measured with software vs. stopwatch (P=0.002). Absolute agreement between measurements was excellent (ICC≥0.97, P<0.001). There was proportional bias between software vs. stopwatch (R²≥0.19, P≤0.03) and between tests with vs. without the electronic mat (R²=0.27, P=0.008). Comparisons between conditions revealed that walking speed and concurrent validity may be influenced by walking test distance, presence of an electronic mat, speed measurement device, and relative measurement distance.

Interpretation:

Walking test procedures influence walking speed and concurrent validity between measurements. Waking test procedures should be as similar as possible with normative data or between repeated measurements to optimize validity.

1). Introduction

Walking speed is an important measure for a variety of purposes after stroke, including assessing community ambulation status, tracking functional progress, and prescribing exercise intensity (1, 2). To facilitate these uses, measurements of walking speed are made repeatedly throughout different phases post-stroke (e.g. acute care, rehabilitation, and community dwelling). Measuring walking speed with a stopwatch during the 10-meter walk test (10MWT) is popular and recommended as a standard method because this test can be performed quickly and easily without special equipment (1, 3, 4). In the post-stroke population, the 10MWT has concurrent validity with several functional measures, including endurance (R²=0.64–0.88), mobility (R²=0.74), functional ambulation (R²=0.45–0.77), and community walking speed (R²=0.77) (5–12). The 10MWT also has excellent interrater reliability (ICC: 0.91–0.99) and test-retest reliability (ICC: 0.63–0.99) in this population (7, 8, 10–23). One limitation to the measurement of walking speed with the 10MWT is the risk of stopwatch measurement error.

An alternative method for measuring walking speed which removes the risk of stopwatch error is to use an electronic walkway, such as the GAITRite mat (CIR Systems Inc., Franklin, NJ, USA). In addition to walking speed, electronic walkways also output valuable information about the spatiotemporal characteristics of walking, which provide an enhanced description of deficits and functional progress and can be useful for fall prediction (24). In the post-stroke population, GAITRite walking speed has concurrent validity with walking speed from a motion capture system (mean difference=0.005 m/s), functional ambulation (R²=0.77), and mobility (R²=0.81) (9, 25, 26). GAITRite also has excellent interrater (ICC: 0.997–0.999) and test-retest reliability (ICC=0.76–0.99) in this population (26–28). Despite the benefits of minimizing stopwatch error, the validity of GAITRite measurements of walking speed may be affected by other test procedures including the distance walked, the vertical clearance required to step onto/off the mat, and the distance over which walking speed is measured relative to the overall test distance.

Recently, we found limited concurrent validity for walking speeds measured during the 10MWT by stopwatch and during a ~4-meter walk test by the GAITRite mat software in individuals with chronic stroke (29). Similarly, walking speeds measured by stopwatch during the 3-meter walk test (3MWT) and by the GAITRite mat software have limited concurrent validity in the post-stroke population (30). Walking speeds measured by stopwatch and GAITRite software have been reported to differ by as much as 0.09 m/s (31). These studies suggest that comparisons of walking speed measurements over time or with normative data may be limited by differences in the walking test procedures. However, in these studies (including our recent study), procedures have differed in multiple ways (e.g. presence/absence of an electronic mat and speed measurement device), making it unclear how each procedural difference uniquely contributes to the measurement of walking speed.

Understanding how variations in measurement procedures affect walking speed is important because measurement errors could mislead conclusions about ambulation status, functional progress, and exercise prescription. For example, an individual with a measured walking speed of ~0.49 m/s would be considered a “most-limited community ambulator” based on classifications from Perry et al. (32). However, if a measurement error of ~0.09 m/s had occurred (31), this individual may actually be classified as a “least-limited community ambulator,” a “most-limited community ambulator,” or a “unlimited household ambulator.” As another example, if a treatment induced the minimal clinically important difference for walking speed of 0.16 m/s (33) but measurement procedures differed from before to after treatment, there actually may have been minimal change or a substantial improvement. Clearly, errors in walking speed can have a substantial impact for clinicians. To address how walking test procedures influence walking speed measurements, the purpose of this study was to assess the concurrent validity of walking speed in individuals with chronic stroke measured during the 10MWT with variations in 1) the presence of the GAITRite electronic mat, 2) the speed measurement device, and 3) the measurement distance relative to the total test distance. We hypothesized that differences in each of these walking test procedures would influence walking speed and affect concurrent validity.

2). Methods

Participants were included in the study if they were 40–80 years of age and had sustained a single, mono-hemispheric stroke at least 6 months prior to enrollment. Data for this study were collected in conjunction with another study which used these additional exclusion criteria: 1) other neurological disorders, 2) use of anti-spasticity medications, 3) <5° of volitional movement in both ankles, and 4) contraindications to brain stimulation, such as metal implants, skull fractures or abnormalities, or history of seizures. The target sample size for this study was determined from our previous study (29) which found a difference between walking speed measured by a stopwatch during the 10MWT and by the GAITRite software during walking over the GAITRite mat. Based on the effect size from this study (Cohen’s dz=0.70), we determined that a sample size of 24 was required to detect an effect between measurement methods (α=0.05; β=0.10). All participants provided written informed consent, and the study was approved by the institutional review board at the University of Illinois at Chicago in accordance with all ethical standards.

Participants performed the 10MWT with and without the presence of the GAITRite electronic mat (classic 14’ (4.27 m) model, CIR Systems Inc., Franklin, NJ, USA). For trials with GAITRite, the mat was placed in the middle of the 10-meter vinyl floor walkway with each end taped down. Under both conditions (with or without GAITRite), two trials were performed at self-selected comfortable speed and at maximal speed with standardized instructions to “walk at your normal comfortable speed” and “walk as fast as you safely can,” respectively. No practice trials were performed. Conditions (with or without GAITRite mat) were performed in a blocked, counterbalanced order. To minimize the effects of fatigue, at least 30 seconds of rest was given between trials, with more rest given as needed. Participants wore their typical footwear and were allowed to use their usual orthoses and assistive devices. The walking area was marked with cones, and participants were given a ~2 m acceleration/deceleration zone (in addition to the 10-meter walkway). For safety, all participants wore a gait belt, and the investigator spotted the participant while walking slightly behind them.

During all trials, the time from when the participant’s toes crossed the start cone to when they crossed the end cone was measured with a stopwatch. Mean walking speed across trials was calculated. During trials with the GAITRite mat, walking speed was also computed by the GAITRite software (GAITRite Plus v4.8.3) as the distance between the centroid of the first and last footfalls (always <4.27 m, and participant-specific) divided by the time between these footfalls. Mean walking speed across trials was calculated. All measurements were performed by one investigator. Thus, walking speed was measured under three different conditions (Fig. 1): 1) 10MWT (without GAITRite mat) measured by a stopwatch (labeled 10MWT_stopwatch), 2) 10MWT (partially over the GAITRite mat) measured by the GAITRite software over the middle ~4 meters of the 10-meter walkway (labeled 10MWT_mat_GAITRite), and 3) 10MWT (partially over the GAITRite mat) measured with a stopwatch (labeled 10MWT_mat_stopwatch).

Figure 1: Walking speed measurement conditions.

Walking speed was measured under three different conditions: A) 10MWT_stopwatch: 10MWT (without GAITRite mat) measured by a stopwatch, B) 10MWT_mat_GAITRite: 10MWT (partially over the GAITRite mat) measured by the GAITRite software over the middle ~4 meters of the 10-meter walkway, and C) 10MWT_mat_stopwatch: 10MWT (partially over the GAITRite mat) measured by a stopwatch. Dashed lines represent the 10MWT path with a 2 meter acceleration and deceleration zone. Gray boxes represent the GAITRite mat. Stopwatch and computer symbols indicate whether the measurement method was by a stopwatch or the GAITRite software.

Our statistical analyses were performed with the goal of investigating systematic bias, absolute agreement, and proportional bias. Differences between measurement conditions were assessed by comparing the difference score (condition 1 – condition 2) to 0 with a one-sample t-test (determination of systematic bias). Absolute agreement was tested with ICC [two-way mixed, single measures, i.e. ICC (3,1)], and interpreted as: poor (<0.50), moderate (0.5–0.75), good (0.75–0.90), and excellent (>0.90) (34). Bland-Altman plots were constructed with mean difference values and linear limits of agreement (mean difference value + or − 2.064*SD) based on the t-test critical value (df=24). The linear regression (determination of proportional bias) between difference and mean values was tested, and when significant, this linear regression and limits of agreement (linear regression line + or − SD of residuals) were plotted. All measurements were normally distributed per the Shapiro-Wilk test, so parametric statistics were used. All statistical tests were performed with SPSS Statistics 25 (IBM, NY, USA), with statistical significance of P<0.05.

Calculations of test-retest variability were also performed for all three measurement conditions. Standard error of measurement (SEM) was calculated as: $\text{SEM}=\sqrt{\frac{{\text{SD}}_{\text{1}}{}^2+{\text{SD}}_{\text{2}}{}^2}{2}}\times \sqrt{1-\text{ICC}}$, where SD₁ and SD₂ are the standard deviations from the first and second trials, and ICC is intraclass correlation [two-way mixed, absolute agreement, single measures, i.e. ICC (3,1)]. The minimal detectable change for a 95% confidence interval (MDC₉₅) was calculated as: ${\text{MDC}}_{\text{95}}=1.96\times \sqrt{2}\times \text{SEM}$, and the minimal detectable percentage change for a 95% confidence interval (MDC_95%) was calculated as: ${\text{MDC}}_{\text{95\%}}=\left(\frac{{\text{MDC}}_{\text{95}}}{\text{mean}}\right)*100$, where mean is the mean value between trials.

3). Results

Twenty-five individuals with chronic stroke (age: 61 (8) years old; 22 male, 3 female) participated in the study. Time since stroke was 5.8 (3.8) years (range: 0.9–15.7 years). Thirteen individuals had right-sided hemiparesis, and 12 individuals had left-sided hemiparesis. During walking trials, no assistive devices were used except for a cane (n=3). Comfortable and maximal walking speeds, ICC, SEM, MDC₉₅, and MDC_95% under each measurement method are shown in Table 1. Test-retest reliability was excellent for all methods (ICC≥0.90). MCD₉₅ ranged from 0.11–0.23 m/s, and MDC_95% ranged from 11–23%.

Table 1: Walking speed.

Comfortable and maximal walking speeds are shown from three different measurements: 1) 10MWT_stopwatch: stopwatch measurement of the 10MWT (without GAITRite mat), 2) 10MWT_mat_GAITRite: GAITRite software measurement over the middle ~3–4 meters of the 10MWT, and 3) 10MWT_mat_stopwatch: stopwatch measurement of the 10MWT (with GAITRite mat). For each measurement, speeds are shown for trial 1, trial 2, the average of the two trials, and the range of values across participants. Also shown are the intraclass correlations (ICC) between trials, the standard error of measurement (SEM), the minimal detectable change based on 95% confidence interval (MDC₉₅), and the minimal detectable change percent based on 95% confidence interval (MDC_95%).

Walking speed (m/s)
Measurement	Trial type	Trial 1	Trial 2	Average	Range	ICC	SEM (m/s)	MDC95 (m/s)	MDC95%
10MWT_stopwatch	Comfortable	0.80 (0.28)	0.81 (0.27)	0.81 (0.28)	0.25–1.35	0.99 (0.97–0.99)*	0.04	0.11	13%
	Maximal	1.09 (0.42)	1.10 (0.42)	1.09 (0.42)	0.31–1.79	0.98 (0.96–0.99)*	0.07	0.18	17%
10MWT_mat_GAITRite	Comfortable	0.80 (0.28)	0.82 (0.28)	0.81 (0.28)	0.33–1.39	0.97 (0.91–0.99)*	0.06	0.18	22%
	Maximal	1.10 (0.40)	1.13 (0.42)	1.11 (0.40)	0.43–1.69	0.97 (0.93–0.99)*	0.08	0.23	21%
10MWT_mat_stopwatch	Comfortable	0.79 (0.26)	0.81 (0.26)	0.80 (0.26)	0.33–1.24	0.96 (0.90–0.98)*	0.07	0.18	23%
	Maximal	1.06 (0.38)	1.07 (0.37)	1.07 (0.38)	0.43–1.65	0.99 (0.98–0.99)*	0.04	0.12	11%

Values are mean (SD).

^*P<0.05.

3.1). 10MWT_stopwatch vs. 10MWT_mat_GAITRite

Comfortable (mean difference: −0.01; 95% CI: −0.04, 0.02; t=−0.39, P=0.70) and maximal walking speeds (mean difference: −0.02; 95% CI: −0.06, 0.02; t=−1.10, P=0.28) were not different when measured during the 10MWT (without GAITRite mat) by stopwatch (10MWT_stopwatch) and during the 10MWT (partially over the GAITRite mat) by the GAITRite software (10MWT_mat_GAITRite). Absolute agreement between measurement conditions was excellent for comfortable [ICC=0.97 (95% CI: 0.93, 0.99; F=58, P<0.001)] and maximal walking speed [ICC=0.97 (95% CI: 0.93, 0.99; F=65, P<0.001)]. Linear limits of agreement were 0.29 for comfortable speed and 0.42 for maximal speed (Fig. 2A). Linear regression showed no evidence of proportional bias for comfortable (R²=0.001, F=0.03, P=0.87) or maximal walking speed (R²=0.007, F=0.16, P=0.70).

Figure 2: Bland-Altman plots.

Relationship between mean and the difference in speed between conditions for comfortable walking speed (left column) and maximal walking speed (right column). A) comparison between 10MWT_stopwatch and 10MWT_mat_GAITRite, B) comparison between 10MWT_mat_stopwatch and 10MWT_mat_GAITRite, and C) comparison between 10MWT_stopwatch and 10MWT_mat_stopwatch. Horizontal gray lines represent the mean difference value (middle line) and the linear limits of agreement (top and bottom line). When significant, angled gray lines represent the linear regression (middle line) and the regression-based limits of agreement (top and bottom line). Equations and R² values for regression lines are displayed. Each data point represents values for one participant.

3.2). 10MWT_mat_stopwatch vs. 10MWT_mat_GAITRite

Comfortable walking speeds (mean difference: −0.02; 95% CI: −0.03, 0.002; t=−1.80, P=0.09) were not different when measured during the 10MWT (partially over the GAITRite mat) by stopwatch (10MWT_mat_stopwatch) and during the 10MWT (partially over the GAITRite mat) by the GAITRite software (10MWT_mat_GAITRite). Maximal walking speeds (mean difference: −0.05; 95% CI: −0.08, −0.02; t=−3.45, P=0.002) were different between these measurements (systematic bias). Absolute agreement between measurement methods was excellent for comfortable [ICC=0.98 (95% CI: 0.96, 0.99; F=140, P<0.001)] and maximal walking speed [ICC=0.98 (95% CI: 0.91, 0.99; F=119, P<0.001)]. Linear limits of agreement were 0.18 for comfortable speed and 0.29 for maximal speed (Fig. 2B). There were weak, but significant, linear regressions (Fig. 2B) between difference and mean values (proportional bias) for comfortable (R²=0.19, F=5.5, P=0.03) and maximal walking speed (R²=0.21, F=6.1, P=0.02); individuals who walked faster tended to have faster walking speeds during 10MWT_mat_GAITRite than 10MWT_mat_stopwatch. Regression-based limits of agreement ranged from 0.05 to 0.17 for comfortable speed and 0.36 to 0.48 for maximal speed.

3.3). 10MWT_stopwatch vs. 10MWT_mat_stopwatch

Comfortable (mean difference: 0.01; 95% CI: −0.02, 0.04; t=0.78, P=0.44) and maximal walking speeds (mean difference: 0.03; 95% CI: −0.01, 0.06; t=1.69, P=0.10) were not different when measured during the 10MWT (without GAITRite mat) by stopwatch (10MWT_stopwatch) and during the 10MWT (partially over the GAITRite mat) by stopwatch (10MWT_mat_stopwatch). Absolute agreement between measurement methods was excellent for comfortable [ICC=0.98 (95% CI: 0.94, 0.99; F=79, P<0.001)] and maximal walking speed [ICC=0.98 (95% CI: 0.95, 0.99; F=100, P<0.001)]. Linear limits of agreement were 0.25 for comfortable speed and 0.32 for maximal speed (Fig. 2C). There was a weak, but significant, linear regression (Fig. 2C) between difference and mean values for maximal walking speed (R²=0.27, F=8.3, P=0.008); individuals who walked faster tended to have a faster walking speed as measured during 10MWT_stopwatch than 10MWT_mat_stopwatch. Regression-based limits of agreement ranged from 0.17 to 0.31 m/s for maximal walking speed. Linear regression revealed no evidence of proportional bias for comfortable walking speed (R²=0.08, F=1.9, P=0.18).

4). Discussion

In this study, we assessed the concurrent validity of walking speed in individuals with chronic stroke measured during the 10MWT with variations in the procedures of the walking tests. Our results suggest that the test distance, presence of the GAITRite mat, measurement method, and relative measurement distance all influence walking speed. Below, we discuss these results and the implications for clinicians and researchers.

In the current study, we examined the concurrent validity of walking speeds measured with a stopwatch and GAITRite software when both measurements were made during the 10MWT (10MWT_stopwatch vs. 10MWT_mat_GAITRite). We found that comfortable and maximal walking speed measurements were not statistically significantly different between the stopwatch and GAITRite software, absolute agreement was excellent, and limits of agreement were acceptable. In a previous study from our lab, we examined the concurrent validity of walking speed measured with a stopwatch during the 10MWT (without GAITRite mat, i.e. 10MWT_stopwatch) and measured with GAITRite software across the GAITRite mat during a ~4-meter walking test (29). In contrast to the current study, we found that walking speeds were faster when measured with a stopwatch during the 10MWT than when measured with GAITRite software during a ~4-meter walking test, and absolute agreement was lower. Differences between the previous and current study from our lab likely can be attributed to the effect of differences in the walking test distance; in the previous study the GAITRite software measurement was made during a shorter test than the stopwatch measurement, whereas in the current study GAITRite and stopwatch measurements were performed during tests of equal length. These results suggest that concurrent validity between walking speed measurements is better when the distance walked is similar, and walking speeds may be slower when the walking test distance is shorter. However, others have found that different walking test distances do not influence walking speeds in individuals with chronic stroke (35).

4.1). Presence of GAITRite mat

In faster walkers, maximal walking speeds were faster when the 10MWT was performed without the GAITRite (10MWT_stopwatch) than when performed with the GAITRite mat (10MWT_mat_stopwatch). These results suggest that walking speeds may be slower when tests are performed across the GAITRite mat, possibly from a fear of falling associated with the required foot clearance and narrower path when walking across the GAITRite mat (36–38). This may be particularly true for faster walkers because faster speeds allow less time to adjust foot clearance and path width. Thus, the presence or absence of the GAITRite mat may affect walking speed measurement and concurrent validity between measurements that differ by this test procedure.

4.2). Speed measurement device

In the current study, we found that walking speeds were faster when measured with GAITRite software than with stopwatch (10MWT_mat_stopwatch vs. 10MWT_mat_GAITRite), particularly at faster speeds and in faster individuals. Peters et al. (30) also found that walking speeds were faster when measured with the GAITRite software than with a stopwatch. In contrast, Youdas et al. (31) found that walking speeds are faster when measured by stopwatch than by GAITRite software. These disparate findings suggest that stopwatch error may have a range of effects on walking speed measurements and may affect concurrent validity between measurements that differ by this test procedure. In our study, we believe that the faster walking speeds measured with GAITRite software likely reflect differences in relative measurement distance (see discussion below), not speed measurement device.

4.3). Relative measurement distance

In the current study, we found that walking speeds were faster when measured by GAITRite software than stopwatch during the 10MWT (10MWT_mat_stopwatch vs. 10MWT_mat_GAITRite), particularly at faster speeds and in faster walkers. These discrepancies may reflect differences in the speed measurement device or the relative measurement distance. When considered in light of the findings from Youdas et al. (31) mentioned above, faster speeds during 10MWT_mat_GAITRite vs. 10MWT_mat_stopwatch likely reflect differences in the relative measurement distance. Namely, walking speeds are faster when measured over a portion of the walking test distance, not the total distance. Walking speeds may have been slower during 10MWT_mat_stopwatch because participants were accelerating/decelerating through portions of the test, whereas, over the ~4 meters of GAITRite software measurement, a steady-state, peak speed was likely achieved. In accordance with this explanation, Salbach et al. 2001 (39) found that the 10MWT was less responsive than the 5-meter walk test, likely because participants could not maintain speed for 10 meters. This effect may be particularly evident in faster walkers merely because faster speeds lead to greater range of speeds during the test. Overall, the relative measurement distance may affect walking speed measurement and concurrent validity between measurements that differ by this test procedure.

4.4). Test-retest reliability, SEM, and MDC

In the current study, we found excellent within session test-retest reliability for all three measurement methods. Previous studies have also found good to excellent within session test-retest reliability for the 10MWT (15, 16, 20, 21, 23) and GAITRite (28) in the post-stroke population. In most cases, the limits of agreement between conditions were larger than the MDC₉₅ values calculated within each condition (0.11–0.23 m/s). Thus, differences in walking speed between conditions may reflect measurement error within one or both conditions. Other studies have found between session MDC₉₅ values between 0.05–0.4 m/s (8, 11, 13, 18, 22, 30) and MDC_95% values between 16–34% (7, 22). Our values are on the lower end of these ranges, probably because of additional error for between session than within session measurements. Consequently, even greater differences between conditions may merely reflect measurement error.

4.5). Limitations

Participants only performed 2 trials of each condition, with no practice trials. We adopted this strategy to reduce testing time while still yielding reliable measurements (23). However, other studies have shown an increase in walking speed across the first couple of trials for the 10MWT and GAITRite (20, 40). Thus, not having practice trials may have increased the variability between trials and conditions, limiting insights from this study. We evaluated chronic stroke survivors who could walk independently (likely higher functioning), and only 3 participants used an assistive device. The results from this study should not be generalized to other populations. Although this study provides insight into how several procedures of walking tests influence walking speed in individuals with chronic stroke, it is important to note that other factors are also important, such as the walking surface and the instructions used (17, 41, 42). For example, the prompt/instructions may account for as much as 61% of the variance in walking speed in healthy young adults (42). Thus, as with the procedures identified in this study, care must be taken to standardize as many procedural aspects of walking tests as possible.

4.6). Conclusions

Procedures of walking tests, including the presence of the GAITRite mat and speed measurement device, influence walking speed and concurrent validity between different measurements of walking speed. When considered alongside previous research, our results also suggest that other procedures may influence walking speed and concurrent validity, including test distance and relative measurement distance. To best guide comparisons with normative data or between repeated measurements (optimize validity) these procedures should be as similar as possible between measurements. Walking speeds may be slower when the walking test distance is shorter, when tests are performed across the GAITRite mat, or when walking speed is measured over a greater distance relative to the total test distance. Measurement device (stopwatch vs. software) appears to have a variable effect. Considering the impact of varied test procedures, the 10MWT_stopwatch procedure may be optimal for measuring walking speed in this population. This procedure is the quickest, easiest, and most cost effective, and thus the most common. As a result, the 10MWT_stopwatch procedure is most likely to ensure similarity of measurement with normative data or measurements from other sources. If differences in these test procedures are unavoidable, then the effect of each characteristic on walking speed should be considered to assess the risk of over- or under-estimating walking speed compared to another measurement.

Table 2: Comparison of walking test procedures.

For the comparison of each pair of measurement conditions at each walking speed, we show the mean difference (Diff) with t-value, the linear limits of agreement (LOA), the intraclass correlation (ICC), and whether there was evidence of proportional bias (significant linear regression on Bland-Altman plot).

Comparison	Speed	Diff. (95% CI) m/s	t	LOA (m/s)	ICC (95% CI)	Proportional bias (R2)
10MWT_stopwatch vs.	Comfortable	−0.01 (−0.04, 0.02)	0.39	−0.15, 0.14	0.97 (0.93, 0.99)*	0.001
10MWT_mat_GAITRite	Maximal	−0.02 (−0.06, 0.02)	1.10	−0.23, 0.19	0.97 (0.93, 0.99)*	0.007
10MWT_mat_stopwatch vs.	Comfortable	−0.02 (−0.03, 0.002)	1.80	−0.11, 0.08	0.98 (0.96, 0.99)*	0.19*
10MWT_mat_GAITRite	Maximal	−0.05 (−0.08, −0.02)	3.45*	−0.20, 0.10	0.98 (0.91, 0.99)*	0.21*
10MWT_stopwatch vs.	Comfortable	0.01 (−0.02, 0.04)	0.78	−0.11, 0.14	0.98 (0.94, 0.99)*	0.08
10MWT_mat_stopwatch	Maximal	0.03 (−0.01, 0.06)	1.69	−0.13, 0.19	0.98 (0.95, 0.99)*	0.27*

^*P<0.05.

• Inclusion of an electronic mat influences measurement of walking speed.

• Measurement device (stopwatch or software) influences measurement of walking speed.

• Test distance may influence measurement of walking speed.

• Relative measurement distance may influence measurement of walking speed.

• These factors may influence comparisons within individuals and with normative data.