Volume 87, Issue 12 , Pages 1648-1652, December 2006
Agreement Between the GAITRite Walkway System and a Stopwatch–Footfall Count Method for Measurement of Temporal and Spatial Gait Parameters
Article Outline
Abstract
Youdas JW, Hollman JH, Aalbers MJ, Ahrenholz HN, Aten RA, Cremers JJ. Agreement between the GAITRite walkway system and a stopwatch–footfall count method for measurement of temporal and spatial gait parameters.
Objective
To determine the agreement for measurements of stride length, cadence, and walking speed obtained from the GAITRite system and the stopwatch–footfall count technique.
Design
Criterion standard.
Setting
Research laboratory in a physical therapy education program.
Participants
Forty healthy volunteers (13 men, 27 women) without lower-extremity injury.
Interventions
Participants walked across a GAITRite mat with embedded pressure sensors at their self-selected walking speed. Simultaneously, an examiner used a stopwatch to record the elapsed time necessary to cross the mat and counted the number of complete footfalls.
Main Outcome Measures
Walking speed, cadence, and stride-length measures were compared between the GAITRite system and the stopwatch–footfall count technique.
Results
Correlation coefficients comparing both systems were .97 for walking speed, .75 for cadence, and .85 for stride length. Ninety-five percent of the time we would expect the between-methods differences to range between .09 and −.05m/s for walking speed, between −1.5 and −24.3 steps/min for cadence, and between .01 and .37m for stride length.
Conclusions
This study shows that the GAITRite and stopwatch–footfall count methods lack clinically acceptable agreement for the measurements of cadence and stride length in a group of healthy volunteers walking at their self-selected speeds. Clinicians who require precise measurement of cadence and stride length should consider using the GAITRite system instead of the stopwatch–footfall count technique.
Key Words: Gait, Rehabilitation, Reproducibility of results, Walking
MEASUREMENTS OF TEMPORAL and spatial gait parameters are commonly used by rehabilitation professionals to identify gait deviations, to screen elderly people for risk of falling, to monitor patient progress, and to determine the effectiveness of therapy interventions.1, 2 The GAITRite analysis systema is a popular computerized walkway system for the quantification of temporal and spatial gait parameters. The standard system is a portable, carpeted electronic walkway embedded with pressure sensors that detect a series of footfalls. The walkway is connected to a personal computer with application software that calculates the temporal and spatial gait parameters.3, 4, 5, 6, 7, 8 Recently, numerous investigators have examined the validity and reliability of the GAITRite system. The GAITRite has shown high levels of concurrent validity for obtaining measurements of temporal and spatial parameters when compared with (1) a 6-camera motion analysis system (Vicon 512b)8, (2) a single-camera video-based system (Peak Performance 3.1b),5 (3) a Clinical Stride Analyzerc composed of a pair of insoles inserted within a subject’s shoes,4 (4) a paper-and-pencil method,6 and (5) a standard video camera.6 Furthermore, investigators have quantified the test–retest reliability of temporal and spatial parameters obtained with the GAITRite over a 2-week period,9 over a 1-week period,7 over a 24-hour period,3 and on the same day.6 These repeated measurements were highly reliable, indicating there is minimal error variation associated with the measurement of temporal and spatial parameters when using the GAITRite system.
Although the GAITRite system has well-documented capabilities, it is costly; many rehabilitation departments may lack the financial resources necessary to afford a device that may not be used routinely, or they may not have the physical space to store or set up the GAITRite’s carpeted walkway. Two traditional and inexpensive methods of gait analysis include the use of videotape and a standard stop watch and tape measure. Most clinics that examine gait have access to a video camera. Calculation of walking speed, cadence, and stride length from a videotape of a subject walking is straightforward, provided the videotape displays the subject traversing several sequential markings placed on the floor separated by a known distance (30cm [1ft]). When the videotape is played, an observer can record the elapsed time and count the number of footfalls necessary to cover the known distance. Walking speed, cadence, and stride length can then be easily calculated using standard formulas.10 Although the videotape technique is technically easy to perform, it requires a clinician to select a walkway free from obstacles. In addition, the clinician must measure a known distance, generally at least 5m, and apply a series of marks to the floor with tape. In an inpatient or outpatient physical therapy (PT) setting, this walkway can be used repeatedly by clinicians without the need to remark the walkway’s dimensions. However, clinicians who provide PT services to patients in a home health care practice setting would find the videotape method inconvenient when attempting to measure walking speed, cadence, and stride length. A client’s home may not lend itself to arranging a 5-m walkway that can be clearly viewed with a video camera. Furthermore, the physical therapist may find the process of applying numerous strips of tape to the floor time intensive. Therefore, despite its low-tech nature, a clinician who practices in a home health setting may prefer the stopwatch and tape measure method because of its application. Whittle10 described a basic method for calculating the average temporal and spatial parameters with a stopwatch and tape measure. Walking speed, cadence, and stride length when quantified provide a rudimentary assessment of walking performance. Speed is measured by recording the time in which a subject walks a known distance:
Cadence may be measured with a stopwatch by counting the number of footfalls a subject takes during a known time period:
This calculation uses an integer value for the number of footfalls taken and does not consider partial footfalls. Such a technique may underestimate a subject’s cadence. Stride length can be indirectly calculated from the distance walked, elapsed time, and number of footfalls:
However, if the steps counted over a known distance are integer values versus partial steps, then the stride length will be overestimated. This manual method described by Whittle10 is inexpensive and can be completed within minutes. However, measurements of temporal and spatial parameters obtained with the stopwatch and counting the number of footfalls over a known distance may not agree with the measurements obtained from the GAITRite system. The purpose of this study was to document the agreement between measurements of temporal and spatial parameters obtained from the GAITRite system and the stopwatch–footfall count technique. If the 2 procedures agreed, then a clinician could be confident that measurements of temporal and spatial parameters obtained with the stopwatch–footfall count technique as described by Whittle10 compare favorably with the more sophisticated and expensive GAITRite system.
Methods
Participants
After receiving Mayo Foundation institutional review board approval, 40 healthy volunteers (13 men, 27 women) were recruited via bulletin board announcements. For inclusion in this study, subjects needed to be between 18 and 60 years of age. Exclusion criteria included use of assistive devices or an antalgic or pathologic gait pattern as determined by visual observation. We chose to study healthy subjects because they were homogeneous, were readily accessible, and provided a good starting point for examining the agreement between the GAITRite system and the stopwatch–footfall count method. Moreover, we believed that healthy subjects walking at their self-selected speeds would challenge the potential limitations of the stopwatch–footfall count method due to the error associated with latencies in starting and stopping the watch and the problems encountered with counting complete footfalls. Written informed consent was obtained from each subject before data collection. The subjects’ mean age, body mass, and height were 25.9±5.8 years, 73.3±15.0kg, and 170.9±11.0cm, respectively.
Instrumentation
The GAITRite included a 61-cm–wide by 526-cm–long roll-up mat with 6 sensor pads encapsulated at intervals along the length of the mat. The active area contained 13,824 1-cm2 pressure sensors placed on 1.27-cm centers and arranged in a 48×288-cm grid pattern. A serial interface cable connected the instrumented walkway to a personal computer with GAITRite Gold, version 3.4 software.a The sampling frequency of the system was 80Hz.
A handheld stopwatchd was used to measure the elapsed time necessary to traverse the instrumented walkway. A tape measure and masking tape were used to create a start line and finish line, each 2m on either end of the instrumented mat.
Procedure
Each subject completed 3 successive passes across the instrumented walkway without shoes at his/her self-selected walking speed. Subjects started and finished the pass 2m in front of and beyond the edge of the instrumented mat. As each subject traversed the instrumented mat, the examiner simultaneously recorded the elapsed time it took him or her to cross the mat between 2 imaginary vertical planes at either end of the mat and counted the number of complete footfalls made on the mat during the pass. We did not include partial footfalls because of the inability to assign a numeric value to a fraction of a footfall. The examiner showed the stopwatch to the recorder, who recorded the elapsed time to ensure that the examiner was blinded. The examiner verbally reported to the recorder the number of complete footfalls that landed on the mat. This procedure was repeated 2 more times for a total of 3 trials. After the trials, the examiner calculated the temporal and spatial parameters via the stopwatch–footfall count method, whereas the temporal and spatial parameter data computed by the GAITRite system were available within seconds of completing the pass across the instrumented mat. Conventional data were assessed by using the length of the GAITRite walking mat as 5.26m. Data collected from the 3 trials were averaged. Temporal and spatial parameters assessed in this study were walking speed (in m/s), cadence (in footfalls/min), and stride length (in meters).
Intratester reliability for the stopwatch–footfall count method was established with a subgroup of 12 subjects who did not participate in the primary study. An intraclass correlation coefficient model 3,1 (ICC3,1) and standard error (SE) of measurement were used to estimate the measurement consistency for walking speed (ICC3,1=.88, SE of measurement=.03m/s), cadence (ICC3,1=.87, SE of measurement=3.1 steps/min), and stride length (ICC3,1=1.0, SE of measurement=0m).11 All 3 temporal and spatial parameters had good intratester reliability.
Data Analysis
Descriptive statistics were calculated for the temporal and spatial parameters. The Pearson product-moment correlation coefficient (r) was used to examine the relation between the temporal and spatial parameters obtained from the GAITRite device and those obtained by use of the stopwatch–footfall count technique (α=.05). Furthermore, the Bland and Altman method12 (95% limits of agreement) was used to express the agreement between the measurements obtained with the stopwatch–footfall technique and those calculated by the GAITRite system. This graphic technique allows one to visually assess the agreement between the 2 methods. The width of the interval ±1.96 standard deviations (SDs) reflects the variability of the differences. The Bland-Altman plot will also show systematic error in the measurements. Bias can be assessed by examining the mean between-methods difference (). Furthermore, paired sample t tests were performed to determine if there was a significant difference between the mean value computed by the GAITRite system versus that computed by the stopwatch–footfall count method. The level of significance was α equal to .05. In addition, to quantify a clinically significant difference between the methods for the 3 temporal and spatial parameters, we calculated the minimal detectable change (MDC), or the minimal amount of change not likely to be attributed to chance variation in a measurement. The MDC is calculated using the following formula:
where the z score represents the confidence interval, SD is the standard deviation of the variables measured with the stopwatch–footfall count method, and r is the form of the ICC3,1. The multiplier of √2 accounts for the uncertainty created when using difference scores from measurements at 2 points in time.13, 14
Results
The mean values for temporal and spatial parameters for each of the 2 methods are shown in table 1. Paired samples correlation showed a high correlation with statistical significance (P<.05) between the stopwatch–footfall count method and the GAITRite analysis system. Walking speed had a correlation of .97, followed by stride length and cadence with r equal to .85 and r equal to .75, respectively. Figure 1 shows the plots of the algebraic differences (y axis) versus the mean value (x axis) for temporal and spatial parameters when measured by the 2 methods. For walking speed (see fig 1A) the 95% level of agreement was .02±.07m/s. For cadence (see fig 1B) the 95% level of agreement was −12.9±11.4 steps/min. For stride length (see fig 1C) the 95% level of agreement was .19±.18m. We found a statistically significant difference between the stopwatch–footfall count method and GAITRite system for walking speed (t39=3.1, P=.004), cadence (t39=−14.0, P<.001), and stride length (t39=13.4, P<.001). The MDCs were .07m/s for walking speed, 12.6 steps/min for cadence, and .14m for stride length.
Table 1. Mean Values for Subjects Measured With the GAITRite and the Stopwatch–Footfall Count Method
| Gait Variable | GAITRite | Stopwatch–Footfall Count |
|---|---|---|
| Speed (m/s) | 1.40±0.13 | 1.42±0.14 |
| Cadence (steps/min) | 119.6±7.6 | 106.7±8.8 |
| Stride length (m) | 1.42±0.13 | 1.61±0.17 |
Fig 1.
The visual plot of the difference between the stopwatch–footfall count and GAITRite methods as a function of the mean value of each pair of readings for (A) walking speed, (B) cadence, and (C) stride length. The solid line marks the mean difference score between the stopwatch–footfall count and GAITRite measurements. The dashed line represents the upper and lower 95% limits of agreement between the 2 methods.
Discussion
Based on the Pearson correlation coefficients, the strength of the relation between the stopwatch–footfall count method and GAITRite methods was high (r=.97) for walking speed and good for stride length (r=.85).15 The correlation coefficient for cadence was below .80. However, these correlations do not signify agreement. Agreement is best represented by the magnitude and variability of difference scores between methods of measurement. Therefore, we also calculated the 95% limits of agreement in this study.
Clinically, self-selected walking speed is a good indicator of how people walk.5, 6, 16 Walking speed decreases with injury or pain and increases with recovery. Investigators have documented that walking speed is related to other gait factors in postoperative orthopedic patients and healthy controls.17 Furthermore, walking speed is sensitive in screening older people for dysfunction in both balance and mobility.18 For walking speed (see fig 1A), we would expect the between-methods difference to yield a maximum range of error of .14m/s (range, .09 to −.05m/s) with a mean difference of .02m/s 95% of the time. The expected difference between methods for calculating walking speed (.02±.07m/s) does not exceed our estimate of the MDC (.07m/s). Therefore, either the GAITRite or the stopwatch–footfall method would be valid for estimating walking speed. We believe the primary factor accounting for the variability between the GAITRite and stopwatch–footfall methods when calculating walking speed is the inherent measurement latency when using the stopwatch. The timing of the start and completion of the walking distance was based on the examiner’s visual observation of a subject crossing an imaginary vertical plane, whereas the GAITRite system used electronic sensors embedded in the mat. Furthermore, because the mean walking speed of our subjects was within reported normal values we would expect greater measurement error between the 2 methods due to timing.19 We believe patients with slower walking speeds would break the imaginary planes marking the start and finish points on the GAITRite mat less abruptly and thus account for smaller timing latencies. Hence we would expect smaller timing errors when testing subjects with slower walking speeds.
For cadence and stride length, we found clinically unacceptable agreement between the 2 methods. For example, 95% of the time we would expect the between-methods difference for cadence to yield a maximum range of error of about 23 steps/min (range, −24.3 to −1.5 steps/min) (see fig 1B) with a mean difference of −12.9 steps/min. Moreover, 95% of the time we would expect a maximum range of error of .36m for stride length (range, .01−.37m) (see fig 1 C) with a mean difference of .19m. Examination of the Bland and Altman plots shows that systematic error exists with the stopwatch–footfall count method (see figs 1B, 1C). The expected differences between methods for calculating cadence (12.9±11.9 steps/min) and stride length (.19±.18m) do exceed the MDCs (12.6 steps/min for cadence, 14m for stride length); therefore, the stopwatch–footfall count method does not appear valid. Using values obtained from the GAITRite as the criterion standard (see table 1) and the visual display of the differences between the 2 methods (see figs 1B, 1C), we would expect the stopwatch–footfall method to underestimate the true value of cadence and overestimate the true value of stride length. Error associated with the measurement of cadence can be attributed to a combination of timing error and error when counting complete footfalls, whereas error associated with stride length is related only to error when counting complete footfalls. We were unable to directly compare the number of footfalls between the GAITRite system and the stopwatch–footfall count technique. For example, the GAITRite system does not record the number of footfalls over the entire length of the mat (5.26m), because the active pick-up distance is about 4.35m for the system we used. In contrast, the footfall count technique used a walking distance of 5.26m. Although the measurement error associated with calculation of cadence and stride length using the stopwatch–footfall method is clinically unacceptable when applied to subjects walking at normal speeds, we would expect less measurement error in subjects with gait pathology, because they typically walk slower with shorter step lengths. Hence a person walking with shorter step lengths will take more steps over the same 5.26m, so error due to ignoring partial steps would be less. Furthermore, with slower walking speeds timing error would be expected to have a smaller effect on the calculation of both walking speed and cadence. Compared with the GAITRite method, the stopwatch–footfall count technique also has several weaknesses.18 Unlike the GAITRite system, the stopwatch–footfall count technique will not provide a clinician with information about symmetry between the right and left sides during the gait cycle.
Study Limitations
The results of the present study may lack external validity. For example, 1 limitation of the current study is that our data were obtained from young adult participants, so the results may lack external validity when applied to children, who have different walking speeds and shorter stride lengths. According to Craik and Dutterer,19 walking performance at a self-selected speed for a person with symmetric foot placements occurs at a walking speed ranging from 1.2 to 1.5m/s and a cadence of 100 to 120 steps/min. Our data agree with these reported ranges (see table 1). Investigators have reported a variety of values for step length, which equals one half of the stride length. Gabell and Nayak,20 using an instrumented walkway, reported that average step length ranged from .65 to .91m for a group of 32 young adults aged 21 to 47 years. Blanke and Hageman21 reported a mean step length of .88±.07m for 12 young men aged 20 to 32 years, whereas Hageman and Blanke22 reported mean step length of .81±.05m for 13 young women aged 20 to 35 years. Likewise, these values, which were obtained by high-speed cinematography, are comparable to our value of average step length (.81m) for young adults obtained by the stopwatch–footfall count technique. Another limitation of this study is that the data were obtained from participants with no observable gait dysfunction who walked at their self-selected walking speeds only. Additional studies need to be performed on participants with known gait impairments, those who range from young children to older adults, and healthy subjects who voluntarily walk at slow speeds. Last, we did not compare the agreement of measurements of walking speed, cadence, and stride length obtained with the GAITRite system with those obtained with videotape. Although less expensive than the GAITRite system, the videotape method potentially could have better agreement with the GAITRite system when measuring cadence and stride length than the stopwatch–footfall count method.
Conclusions
This study showed that the stopwatch–footfall count method is valid for estimating walking speed on the basis of MDC in a group of healthy participants walking at their self-selected speeds. Clinicians who require the precise measurement of cadence and stride length should consider using the GAITRite system as opposed to the stopwatch–footfall method.
Suppliers
References
- Is this walk normal?. Age Ageing. 1993;22:27–30
- . Clinical gait analysis: theory and practice. St. Louis: Elsevier; 2006;
- . Assessing reliability of measurement of gait velocity. Physiotherapy. 2003;89:313–317
- . Concurrent related validity of the GAITRite walkway system for quantification of the spatial and temporal parameters of gait. Gait Posture. 2003;17:68–74
- . Evaluation of an instrumented walkway for measurement of the kinematic parameters of gait. Gait Posture. 2000;12:134–138
- . The validity and reliability of the GAITRite system’s measurements: a preliminary evaluation. Arch Phys Med Rehabil. 2001;82:419–425
- . Test-retest reliability of temporal and spatial gait characteristics measured with an instrumented walkway system (GAITRite). BMC Musculoskelet Disord. 2004;5:13
- . Validity of the GAITRite walkway system for the measurement of averaged and individual step parameters of gait. Gait Posture. 2005;22:317–321
- . Reliability of the GAITRite walkway system for the quantification of temporo-spatial parameters of gait in young and older people. Gait Posture. 2004;20:20–25
- . Gait analysis: an introduction. 3rd ed. Boston: Butterworth-Heinemann; 2002;
- . Intraclass correlation: use in assessing rater reliability. Psychol Bull. 1979;86:420–428
- . Statistical methods for assessing agreement between two methods of clinical measurement. Lancet. 1986;8:307–310
- . Reliable change and minimum important difference (MIP) proportions facilitated group responsiveness comparisons using individual threshold criteria. J Clin Epidemiol. 2004;57:1008–1018
- . Sensitivity to change of the Roland-Morris Back Pain Questionnaire: part 1. Phys Ther. 1998;78:1186–1196
- . Foundations of clinical research: applications to practice. 2nd ed. Upper Saddle River: Prentice Hall Health; 2000;
- . Hemiplegic gait (Relationships between walking speed and other temporal parameters). Am J Phys Med Rehabil. 1997;76:128–133
- . Walking speed as a basis for normal and abnormal gait measurements. J Biomech. 1977;10:261–277
- . Screening for balance and mobility impairment in elderly individuals living in residential care facilities. Phys Ther. 1995;75:462–469
- . Spatial and temporal characteristics of foot fall patterns. In: Craig RL, Oatis CA editor. Gait analysis: theory and application. St. Louis: Mosby; 1995;p. 143–158
- . The effect of age on variability in gait. J Gerontol. 1984;39:662–666
- . Comparison of gait of young men and elderly men. Phys Ther. 1989;69:144–148
- . Comparison of gait of young women and elderly women. Phys Ther. 1986;66:1382–1387
No commercial party having a direct financial interest in the results of the research supporting this article will confer a benefit upon the author(s) or upon any organization with which the author(s) is/are associated.
PII: S0003-9993(06)01344-X
doi:10.1016/j.apmr.2006.09.012
© 2006 American Congress of Rehabilitation Medicine and the American Academy of Physical Medicine and Rehabilitation. Published by Elsevier Inc. All rights reserved.
Volume 87, Issue 12 , Pages 1648-1652, December 2006
