Temporal and Spatial Characteristics of Gait During Performance of the Dynamic Gait Index in People With and People Without Balance or Vestibular Disorders

Abstract

Background and Purpose: Understanding underlying gait characteristics during performance of the Dynamic Gait Index (DGI) could potentially guide interventions. The purpose of this study was to describe the characteristics and reliability of gait performance during the level walking items of the DGI in people with balance or vestibular dysfunction. The study was a cross-sectional investigation with 2-group comparisons.

Subjects and Methods: Forty-seven subjects (mean age=59.2 years, SD=8.5, range=24–90) participated in the study; 26 were control subjects, and 21 were subjects with balance or vestibular dysfunction. Three trials of each level gait item were administered to subjects as they ambulated on an instrumented walkway. Test-retest reliability was determined by use of an intraclass correlation coefficient (3,1) 2-way random-effects model for gait parameters associated with continuous walking and the item requiring turning and stopping quickly. Mean gait parameter differences between control subjects and subjects with balance or vestibular disorders were compared by use of a multivariate analysis of variance for each gait task.

Results: The reliability of most gait parameters during DGI performance were fair to excellent between trials. Subjects with balance or vestibular disorders demonstrated differences in gait characteristics compared with control subjects. The heterogeneity of the group of subjects with balance or vestibular disorders does not permit inferences to be drawn regarding the relationship between gait and any specific balance or vestibular diagnosis. The results are most pertinent to people with chronic balance or vestibular disorders.

Discussion and Conclusion: Gait parameters underlying dynamic walking appeared to be relatively reliable across multiple trials and distinguished subjects with balance or vestibular disorders. Evaluating a person's performance on items of the DGI may be useful in identifying gait deviations and in evaluating gait improvements as a result of interventions.

People with balance or vestibular disorders frequently demonstrate gait disorders, frequently report postural instability, and fall more frequently than community-living adults without vestibular dysfunction.^1–4 People with vestibular disorders ambulate with a gait that has a wide base.⁵ When subjects with vestibular disorders walked at their self-selected speed, their stance width was similar to that of control subjects but at the expense of decreased speed.⁶ Increased stance width was noted when these same subjects were asked to walk at an externally paced speed (120 steps per minute).⁶ People with vestibular disorders ambulate with increased stance time variability,⁷ decreased gait speed,^6,8,9 and increased stride time variability.¹⁰

Balance and gait tasks requiring head motion often provoke symptoms of dizziness and postural instability in people with vestibular disorders. Static trunk balance during standing was shown to improve rapidly after a vestibular event, yet dynamic gait tasks, such as walking with head movements, continued to be abnormal at 3 months compared with those in control subjects.¹¹ In a study by Crane and Demer,¹² subjects with unilateral peripheral vestibular disorders demonstrated less vertical head translation than control subjects during treadmill walking. Although most of the previous studies focused on continuous walking, little is known about gait characteristics associated with active head movements during gait in people with vestibular or balance disorders.

Inferences often are made by clinicians about underlying gait dysfunction on the basis of observation of walking. Observational gait assessment tools assist physical therapists in defining gait deviations. The Dynamic Gait Index (DGI)¹³ was developed to quantify gait dysfunction in older adults during the performance of level walking as well as more complex functional gait tasks, such as head movements, changes in speed, moving over or around objects, and going up and down steps. The rating criteria of the DGI integrate gait deviations, such as a reduction in speed, imbalance, staggering, not staying within the path, and an inability to adjust steps while attempting functional gait tasks. Clinicians quantify a person's functional gait performance on the basis of observations of these selected gait characteristics as defined by the test developers.

Movements of the head up and down while walking are included in 3 clinical measures of dynamic balance: the DGI,¹³ the 4-item DGI,¹⁴ and the Functional Gait Assessment.¹⁵ Coupling of head or trunk stability and gaze stability during active head movements while ambulating appears to contribute to the gait dysfunction often seen in people with balance or vestibular disorders.⁹

Clinicians often do not have access to sophisticated measures of gait performance in clinical settings. It can be difficult to document interferences regarding gait deviations from observational gait analysis with tools such as the DGI. The DGI is a commonly used clinical measure of gait performance¹³ and is frequently used as an outcome measure in people with vestibular dysfunction.^4,14,16–19 It is hoped that this study will provide additional evidence of the construct validity of the DGI.

The purpose of this study was to describe the reliability and magnitude of measurements of the temporal and spatial gait characteristics associated with the performance of complex level walking items of the DGI in people with and people without balance or vestibular dysfunction. It was hypothesized that people with balance or vestibular disorders would demonstrate the characteristics of imbalance described by the developers of the DGI compared with people without diagnosed gait deviations. We defined the characteristics of imbalance as follows: decreased speed, increased gait stability ratio (GSR), decreased step length, increased mean step length variability, decreased cadence, increased double-support time as a percentage of the gait cycle (DS%), increased stance width, and increased stance width variability.

We expected that the differences between people with and people without balance or vestibular impairments would be greatest on walking task items that were shown to be more difficult in previous studies^14,17 through physical therapist evaluation of walking performance, such as walking with up-and-down head movements. The goal of this study was to describe test-retest reliability and to better define what underlying gait dysfunction was identified through the use of the DGI items. Through an understanding of the underlying gait characteristics of people with balance or vestibular disorders during performance of the DGI, it might be possible to develop a targeted intervention program to address the identified gait dysfunction.

Method

Participants

Testing took place at the Balance and Vestibular Clinic of the Centers for Rehab Services, which is a large metropolitan tertiary care institution, and at Lancaster General Hospital, a large rural outpatient facility. Subjects with balance or vestibular disorders were recruited from the existing clinic population at both institutions and were encouraged to bring a friend or spouse who would serve as a control subject. Subjects were not paid for their participation in the study.

The inclusion criteria for subjects with vestibular or balance disorders included having been diagnosed with balance or vestibular dysfunction by a physician and receiving or having previously received rehabilitation because of complaints of balance dysfunction, dizziness, or both. All subjects were required to be between the ages of 21 and 90 years, to have the ability to walk at least 30 m with a cane, and to independently provide informed consent. Formalized cognitive testing was not performed, but all subjects had to be able to follow 3-step commands. Physical therapists were informed of the inclusion criteria, and subjects were recruited from among people who met these criteria. The control subjects also met the inclusion criteria, except that they had no reported history of dizziness or balance dysfunction.

Twenty-one subjects with balance or vestibular disorders (13 women and 8 men; mean age=60 years, SD= 19.8, range=24–90; mean level gait speed of 1.2 m/s, SD=0.3) and 26 control subjects (19 women and 7 men; mean age=58.5, SD=17.8, range=24–89; mean level gait speed of 1.4 m/s, SD=0.2) participated in the study. Subject diagnoses at both sites included unilateral vestibular hypofunction (n=8), multisensory disequilibrium (n=6), Ménière disease (n=2), central vestibular dysfunction (n=2), central plus unilateral hypofunction (n=1), cervicogenic dizziness (n=1), and head trauma (n=1). Twenty-eight of the subjects were recruited from an urban area, and 19 were recruited from a rural setting, with the subjects in the latter group being 8 years older (P<.04). Nine subjects were currently undergoing physical therapy intervention for symptoms of dizziness or imbalance, and the remaining 12 subjects had been discharged from care. The length of their symptoms was variable and ranged from 3 to 840 months. One subject with a 7-decade history of Ménière disease was experiencing an acute exacerbation.

Twelve subjects with a diagnosis of balance or vestibular disorders had undergone vestibular testing at the tertiary care center, with the following results: 7 abnormal rotational chair findings, 7 abnormal caloric results, 2 abnormal oculomotor results, 3 abnormal positional results, and 1 abnormal computerized posturography result. Nineteen subjects with a diagnosis of balance or vestibular disorders reported dizziness that resulted in complaints of functional disability, and 16 subjects had balance-related disability, as measured by the Dizziness Handicap Inventory (DHI)²⁰ and the Activities-specific Balance Confidence (ABC) Scale.²¹ The DHI scores range from 0 to 100, with 0 denoting the lowest dizziness handicap and 100 indicating the worst score possible. The subjects with vestibular disorders had a mean DHI score of 41 (95% confidence interval [CI]=29–53), and the control subjects had a mean DHI score of 4 (95% CI=1–7). The ABC Scale scores range from 0 to 100, with 100 indicating complete confidence in performing each of 16 different balance activities. The subjects with balance or vestibular disorders had a mean ABC Scale score of 64 (95% CI=52–77), and the control subjects had a mean ABC Scale score of 93 (95% CI=89–97).

There were no mean differences between the groups in either height (P<.90) or weight (P<.96). One subject could only complete one trial of the DGI because of fatigue, and one control subject completed 2 replicates of the DGI before having to leave because of transportation issues.

DGI Test Protocol

The DGI comprises 8 different gait tasks that are commonly performed, including walking, walking at different speeds, walking with head movements in pitch and yaw, walking around and over objects, turning and stopping quickly, and ambulation on stairs. All 8 DGI items were performed, but only the level gait items are reported here. Subjects are assigned an ordinal score (0, 1, 2, or 3) corresponding to severe, moderate, mild, or normal performance on each task, with a total possible score of 24. Reliability indexes for the total DGI score in people with unilateral vestibular disorders have been reported to range from .64 to .88.^16,19

The 8 walking tasks of the DGI were administered to all subjects on 3 successive trials, except for stair climbing, which was performed on the first trial only. Subjects performed the walking tasks in their preferred walking footwear. Subjects were given the option of using a straight cane during testing if that was their accustomed walking pattern. Only one subject used a cane during testing. The DGI was administered by use of the protocol and instructions described by Shumway-Cook and Woollacott.¹³ An experienced physical therapist provided all subject instructions and supervision or guarding as needed during test performance. Subjects were given up to a 1-minute rest period between performances of the individual DGI items at each administration. They were allowed to rest, as needed, between total DGI battery administrations. Testing was terminated at subject request.

Gait Parameter Data Collection Protocol

The level walking items of the DGI were performed centered on an 8- × 0.9-m GAITRite instrumented walkway^* that had an active recording area of 0.73 × 0.61 m with 27,648 sensors. The sampling rate was 80 Hz. Data from the electronic walkway were recorded on a Windows-based personal computer and converted into temporal and spatial gait parameters based on footfall patterns. Straight-line walking reliability measurements were previously reported for the GAITRite system.^22,23 Intraclass correlation coefficients (ICCs) were reported to be between .88 and .99.^22,23 To our knowledge, there is no previous work related to the measurement of temporal and spatial characteristics of gait during performance of the type of walking tasks that are included in the DGI. Floor markers were used to standardize command cue timing and obstacle placement for all trials. Walking was initiated 3 m before the front walkway border and was terminated 3 m beyond the back walkway border for all continuous walking items to ensure that the DGI commands were initiated on the walkway while a subject was walking at a constant speed.

The following gait parameters were recorded: speed (meters per second), cadence (steps per minute), step length (centimeters), stance width (centimeters), and DS%. For the turn-and-stop item, potentially overlapping steps were separated and identified in the data by visual inspection. Following the step identification process for the turn-and-stop item, speed, step count (number of steps taken during the entire walk-and-turn sequence), cadence, and ambulation time (seconds) for the entire walk-and-turn sequence were recorded. Gait parameters were collected and reduced by use of GAITRite acquisition software, version 3.4.^*

Data Reduction and Statistical Analysis

Test-retest reliability for measurements of multiple performances of the DGI continuous walking tasks was estimated for the following parameters: speed, cadence, step length, and DS%. Reliability for parameter measurements across multiple performances of the turn-and-stop task was analyzed by use of speed, step count, cadence, and ambulation time. Reliability was determined by use of an ICC (3,1) 2-way random-effects model to estimate the ratio of subject variance to total variance: between subject/[(between subject + within subject)/between trials].^24,25

Subject performance for all gait tasks was described by use of the mean and standard deviation for the parameters identified in the test-retest analysis, averaged across all trials. Additionally, the following parameters were identified for description of the performance of the continuous walking tasks: GSR (steps per meter),²⁶ stance width, step length variability, and stance width variability. Cromwell et al²⁷ suggested that an increase in the GSR is an indication of reduced dynamic walking ability because a greater proportion of the gait cycle is spent in double-limb support. Step length variability and stance width variability were estimated by use of the coefficient of variation (CV) (SD/mean, percent) for all measurements of step length and width at the point of heel contact.

Mean gait parameter differences between subjects with balance or vestibular disorders and control subjects on all gait tasks averaged across multiple trials were estimated by use of multivariate analysis of variance (MANOVA) with a type I error rate of .05. Multivariate analysis of variance deals with multiple dependent variables by combining them in a linear manner to produce a combination that best separates the independent variable groups and reduces the increased likelihood of a type I error that can result from multiple individual group comparisons.

If the MANOVA results indicated a significant overall trend attributable to subject group across multiple dependent gait variables, then group differences in individual gait parameters were tested by use of a 2-sample t test. Differences in gait linear or temporal parameters and variability parameters (CV for step length and stance width) between subject groups were tested by use of separate MANOVA models for the continuous walking tasks. A separate MANOVA model was also used to estimate group effects on the parameters measured during the turn-and-stop task (speed, step count, cadence, and ambulation time).

The relative difficulty of individual continuous walking tasks for subjects because of balance or vestibular disorders was estimated by use of the average effect size differences between subjects with balance or vestibular disorders and control subjects (mean difference/pooled SDs) across the following parameters: speed, cadence, DS%, step length, and GSR.²⁸ Items were ranked for difficulty in descending order of the average effect size differences between subjects with balance or vestibular disorders and control subjects across these 5 parameters. Effect sizes can be interpreted as the amount of distributional nonoverlap between subjects with and subjects without balance or vestibular disorders. Effect sizes of 0.2 and lower are considered small effect sizes and are indicative of less than 14.7% nonoverlap. Effect sizes of 0.3 to 0.6 are considered medium effect sizes and are indicative of 21% to 38% nonoverlap. Effect sizes of 0.6 and higher are considered large effect sizes and are indicative of greater than 38% nonoverlap.²⁸ Statistical analyses were performed by use of SPSS, version 13.0.^†

Results

Test-Retest Reliability of Gait Parameter Measurements of DGI Task Performance

Estimates of reliability determined by use of the ICC (3,1) for selected gait parameters with repeated performance of DGI tasks are shown in Table 1 for the continuous walking tasks and in Table 2 for the turn-and-stop task.

Table 1.

Intraclass Correlation Coefficients (ICC[3,1]) and 95% Confidence Intervals (CIs) for the Test-Retest Reliability of Repeated Measurements of Selected Gait Parameters During Continuous Walking Tasks of the Dynamic Gait Index (DGI)^a

^aAll results were significant at P<.05.

Table 2.

Intraclass Correlation Coefficients (ICC[3,1]) and Confidence Intervals (CIs) for the Test-Retest Reliability of Repeated Measurements of Selected Gait Parameters During the Turn-and-Stop Walking Task

Excellent test-retest reliability for measurements of repeated performances of the DGI continuous walking tasks (ICC≥.75) was shown for subjects with balance or vestibular disorders for all parameters except walking around obstacles and DS% for the items vertical head turns and walking around obstacles. The parameter DS% showed lower estimates of reliability for control subjects than for subjects with balance or vestibular disorders for all tasks except walking with vertical head turns. Reliability estimates for measurements of all parameters except speed was fair to good (.41–.69) for subjects with balance or vestibular disorders for the task of walking around obstacles.²⁴ Lower estimates of reliability were shown for subjects with balance or vestibular disorders than for control subjects with repeated performances of the task of walking around obstacles on all parameters except DS%.

Table 2 shows estimates of reliability for measurements of the gait parameters speed, step count, cadence, and ambulation time across repeated performances of the turn-and-stop task for subjects with balance or vestibular disorders and control subjects. Fair to good reliability for the measurement of ambulation time was shown for subjects with balance or vestibular disorders but not for control subjects. Reliability for control subjects was poor. Excellent reliability for the measurement of speed (ICC=.76) was shown for subjects with balance or vestibular disorders with repeated performances of the turn-and-stop task. The parameters step count and cadence showed relatively poor reliability for repeated performances of the turn-and-stop task for both groups.

Gait Parameter Subject Group Mean Differences

The results of the MANOVA for linear or temporal gait parameters indicated a significant (P<.05) trend attributable to group across continuous walking tasks except for DS% (level walking, changing speeds, vertical head turns, and walking around obstacles), stance width (horizontal head turns and walking around obstacles), and right step length (walking around obstacles). The results of the MANOVA for step length variability and stance width variability (CV) indicated a significant overall effect of subject group on step length variability with changing speeds (F=5.47; df=1,45; P<.03), horizontal head turns (F=13.38; df=1,45; P<.01), vertical head turns (F=8.89; df=1,45; P<.01), and stepping over obstacles (F=9.38; df=1,45; P<.01). There was no significant group effect on stance width variability with any of the continuous walking tasks.

Table 3 shows the mean and 95% CI for all temporal and spatial gait parameters for the continuous ambulation variables for subjects with and subjects without balance or vestibular disorders. Subjects with balance or vestibular disorders showed significantly decreased gait speed (P<.02) and increased GSR with level walking and all other continuous walking tasks (P<.02). The effect size for differences in speed between subjects with balance or vestibular disorders and control subjects was .82 for level walking and ranged from .77 for changing speeds to .91 for walking with horizontal head turns. Figure 1A shows the mean and 95% CI for speed across the continuous walking tasks for subjects with balance or vestibular disorders and control subjects. Similarly, Figure 1B shows the mean and 95% CI for GSR across the continuous walking tasks for subjects with balance or vestibular disorders and control subjects. The effect size for differences in GSR between subjects with balance or vestibular disorders and control subjects was .83 for level walking and ranged from .76 for walking around obstacles to .83 for walking with horizontal head turns and stepping over obstacles.

Table 3.

Mean and 95% Confidence Interval (CI) for Temporal and Spatial Gait Parameters for Subjects With Balance or Vestibular Disorders and Control Subjects Performing Continuous Walking Tasks of the Dynamic Gait Index (DGI)

^aP<.05 for comparisons of the 2 groups.

Figure 1.

Mean and 95% confidence interval (CI) for gait speed (meters per second) (A) and gait stability ratio (steps per meter) (B) for subjects with balance or vestibular disorders and control subjects performing Dynamic Gait Index (DGI) walking tasks.

Figures 2A and 2B show the mean and 95% CI for left and right step lengths, respectively, for subjects with balance or vestibular disorders and control subjects performing the continuous walking tasks. Subjects with balance or vestibular disorders showed shorter step lengths bilaterally for all continuous walking tasks except right step length while moving around objects. The effect size (average of right and left) for differences in step length between subjects with balance or vestibular disorders and control subjects was .72 for level walking and ranged from .55 for stepping over objects to .84 for horizontal head turns.

Figure 2.

Mean and 95% confidence interval (CI) for left (A) and right (B) step lengths (centimeters) for subjects with balance or vestibular disorders and control subjects performing Dynamic Gait Index (DGI) walking tasks.

Subjects with balance and vestibular disorders showed significantly increased mean step length variability, as measured by the CV in step length, while performing the tasks of changing speeds (right leg only; P<.02), horizontal head turns (P<.01), vertical head turns (P<.03), and stepping over obstacles (P<.02).

Subjects with balance or vestibular disorders also showed significantly decreased cadence (P<.05) compared with control subjects while performing all walking tasks except the turn-and-stop task. The effect size for differences in mean cadence between subjects with balance or vestibular disorders and control subjects was .65 for level walking and ranged from .58 for changing speeds to .86 for walking around obstacles.

Subjects with balance or vestibular disorders showed significantly increased DS% compared with control subjects only when performing the task of stepping over objects (P=.02). The effect size for differences in DS% between subjects with balance or vestibular disorders and control subjects while stepping over objects was .71.

There was a significantly increased mean stance width in subjects with balance or vestibular disorders compared with control subjects while performing the tasks of level walking, changing speeds, walking with vertical head turns, and stepping over obstacles (P<.05). There was no significant difference between subjects with balance or vestibular disorders and control subjects in stance width variability, as measured by the CV in stance width, for any of the continuous walking tasks.

The mean and 95% CI for speed, step count, cadence, and ambulation time for all subjects while performing the turn-and-stop task are shown in Table 4. The results of the MANOVA for the parameters measured during the turn-and-stop task indicated an effect attributable to subject group only for ambulation time (F=26.5; df=1,45; P<.05). The subsequent t test indicated that the difference in group means for this parameter was not significant.

Table 4.

Mean and 95% Confidence Interval (CI) for Temporal Gait Parameters for Subjects With Balance or Vestibular Disorders and Control Subjects Performing the Turn-and-Stop Walking Task of the Dynamic Gait Index

Figure 3 shows the average effect size differences between subjects with balance or vestibular disorders and control subjects for the gait parameters speed, cadence, step length (average of left and right), GSR, and DS% for each DGI continuous walking task. When these effect size differences were used as indicators of item difficulty for subjects with balance or vestibular disorders, the DGI tasks of walking with horizontal head turns and stepping over obstacles showed the greatest differences between subjects with balance or vestibular disorders and control subjects. Level walking and walking with vertical head turns appeared to represent equivalent and intermediate levels of difficulty. Changing speeds and walking around obstacles appeared to represent lower levels of difficulty.

Figure 3.

Average effect size mean difference (in SDs) between subjects with balance or vestibular disorders and control subjects for the gait parameters speed, cadence, step length (average of left and right), gait stability ratio, and double-support time (% of gait cycle) during performance of Dynamic Gait Index (DGI) continuous walking tasks.

Discussion

The primary goal of the present study was to describe reliability and the underlying gait differences demonstrated by people with balance or vestibular dysfunction during performance of the complex walking tasks represented on the DGI. The results of the present study support the use of the continuous walking items of the DGI as a clinical tool, because the gait characteristics of subjects with balance or vestibular disorders, as measured by use of a computerized gait mat, were different from those of control subjects.

The hypothesis that subjects with vestibular disorders would demonstrate reliable walking scores was supported across the 3 walking trials. Intraclass correlation coefficients were in the fair to high ranges for gait speed, cadence, double-limb support time, and step length, suggesting that both subjects with balance or vestibular disorders and control subjects performed similarly between walking trials of the DGI items. Overall, the underlying gait characteristics of the DGI were similar for the continuous walking items except for walking around objects, suggesting that clinicians need to observe a subject performing the DGI once. The parameter DS% showed lower reliability in control subjects for 3 of the continuous walking items: level walking, horizontal head turns, and stepping over obstacles.

The gait parameters for the turn-and-stop item of the DGI appeared to be less reliable than those for the continuous items of the DGI for both subjects with balance or vestibular disorders and control subjects, especially the variables step count and cadence. Reliability for the measurement of gait speed for the turn-and-stop item was excellent, and reliability for the measurement of ambulation time was good for the subjects with balance or vestibular disorders and poor for the control subjects. One possible explanation for the decreased reliability across trials is that subject performance might have changed over the 3 repetitions. Subject performance might have changed over the 3 repetitions on the basis of subjects’ previous experience with the task, because all participants were cued to turn and stop at the same location. The ability of the subjects with balance or vestibular disorders to predict when they were to turn and stop might have altered their performance. All subjects had a break between each of the 3 administrations of the task. In addition, subjects with balance or vestibular disorders often report experiencing symptoms with quick movements. Therefore, subject performance might have changed over time as subjects attempted to decrease their symptoms of dizziness.

Subjects with vestibular disorders consistently took fewer steps over each successive trial. The lowest trial step count mean for the subjects with balance or vestibular disorders (6.3 steps) for the turn-and-stop task was greater than the highest step count mean for the control subjects (6.1 steps), although this difference was not statistically significant. The low between-subject variation amplified the within-subject and between-trial variations for the control subjects, thus resulting in the low ICC of .28 for step counts. The mean step counts for the subjects with vestibular disorders were 6.9, 6.6, and 6.3 for trials 1, 2, and 3, respectively. These data suggest that the subjects with balance or vestibular disorders were continuing to improve over the 3 trials. It is unknown how many additional trials would be needed for the step count scores to plateau, suggesting that step count is unpredictable for the turn-and-stop item of the DGI. Thigpen et al²⁹ reported that older adults with self-reported difficulty turning took more steps during the 180-degree turning task included in the Timed “Up & Go” Test than older and young control subjects.

In summary, as was hypothesized, the subjects with balance or vestibular disorders demonstrated decreased gait speed, increased GSR, decreased step length, increased mean step length variability, and decreased cadence compared with the control subjects when performing the DGI continuous walking items. Decreased gait speed was previously reported in subjects with vestibular disorders compared with control subjects.^6,9,30–34 Additionally, gait speed and GSR were different in subjects with balance or vestibular disorders and control subjects. These data are of clinical significance because GSR can easily be recorded during a clinical gait assessment.²⁶ Evidence from Cromwell and Newton²⁶ suggested that GSR is a better indicator of balance deficits than gait speed alone. Furthermore, it is possible that GSR could be used to demonstrate change over time in people with vestibular disorders. This is an area for further investigation. Recording steps per meter in addition to recording gait speed might be a useful adjunct that could be obtained during performance of the DGI, because it could aid clinicians in quantifying gait deviations in people with balance or vestibular dysfunction.

Stance width was increased in the subjects with balance or vestibular disorders compared with the control subjects during the performance of all of the DGI continuous walking items except walking around objects; for the latter, stance width was decreased in the subjects with balance or vestibular disorders. Three subjects with balance or vestibular disorders had a negative mean stance width value, indicative of a crossover step (adducted) pattern as they altered their line of progression around the object. This crossover step pattern was seen only in the subjects with balance or vestibular disorders and was not seen in any control subject. The hypothesis that the subjects with balance or vestibular disorders would have increased stance width variability was not supported. Krebs et al⁶ reported that subjects with balance or vestibular disorders did not have an increase in interfoot distance at their preferred gait speed; when those subjects were asked to walk at 120 steps per minute, their interfoot distance increased, possibly to increase their stability while walking faster. Allum and Adkin¹¹ suggested that subjects with balance or vestibular disorders perform gait tasks more slowly than control subjects so that they can better control their trunk with a “stiffening” response that might decrease trunk sway.

The effect size differences between the subjects with balance or vestibular disorders and the control subjects across the continuous walking tasks of the DGI were considered for 5 gait parameters: speed, cadence, step length (average of left and right), GSR, and DS%. The difference in the averages of these parameter effect sizes between items was considered an indicator of walking task difficulty. These results confirmed the findings of Marchetti and Whitney regarding task difficulty, which were derived from an item response and factor analysis of a clinician's observational ratings of subject performance.¹⁴ The findings of Marchetti and Whitney¹⁴ demonstrated redundancy in item difficulty, suggesting that some DGI items could be eliminated, including the turn-and-stop item, stepping over and around objects, and stair climbing. Walking while performing yaw plane head movements in the present study appeared to be the most difficult task for people with vestibular or balance impairments.

In the present study, the effect size differences between items also supported the findings that a walking task battery of 4 items can be effectively used to observationally screen people for gait dysfunction in order to identify fall risk and the nature of the gait deviation. The 4 items include level walking, walking while changing speeds, and walking with yaw and pitch plane head movements. Our present description of gait parameters showed that the temporal and spatial characteristics of gait in subjects with balance or vestibular disorders during performance of the turn-and-stop gait task and walking around obstacles may be unstable across repeated administrations.

Stepping over obstacles was more difficult for the subjects with balance or vestibular disorders than for the control subjects. This finding is counter to previous findings suggesting that the level of difficulty of stepping over obstacles was the same as that of straight line walking or walking with changing speeds.¹⁴ The relationship between clinician ratings in the study of Marchetti and Whitney¹⁴ and gait parameters associated with stepping over objects requires further exploration given these conflicting findings.

Hall and Herdman¹⁶ reported that mean preferred gait velocities for people with peripheral hypofunction ranged between 1.2 and 1.3 m/s, with an ICC(3,1) of .88. In the present study, the mean gait speed was 1.2 m/s, with an ICC(3,1) of .95, for subjects with balance or vestibular disorders and various diagnoses. The use of the instrumented gait mat most likely increased the accuracy of the measurements. Patten et al⁹ reported a mean preferred gait speed of .98 m/s for people with vestibular disorders. Decreased step length has been reported^31,34 for people with vestibular disease and may contribute to decreased gait speed.

McGibbon et al³⁵ reported step instability (inconsistent stepping patterns) for a sample of people with unilateral or bilateral vestibular disease. During the task of walking around obstacles, subjects with balance or vestibular disorders adducted past midline or crossed over their steps, suggesting an inconsistent pattern, because the adduction pattern was never demonstrated in control subjects. The wide variability in our data for the task of stepping around objects most likely reflects the crossing of the limbs over the center of progression. In the present study, the subjects with balance or vestibular disorders did not perform the task of stepping around objects reliably across trials. Although the task of walking around objects does not make a strong contribution to the DGI test battery because of low test-retest reliability, knowledge of the skill of walking around objects may be useful and may be worthy of separate testing. Subjects with balance or vestibular disorders were less consistent on this task than control subjects, suggesting that it might be a clinically useful task, with the understanding that subject performance might vary across repetitions.

Our data demonstrated the presence of speed changes and other gait deviations in the performance of the DGI items for subjects with balance or vestibular disorders who were seen at a balance clinic in a rural setting and an urban tertiary care center. Future work will examine the relationship between clinician observational scoring of performance on the DGI and measurement of temporal and spatial gait parameters for people with balance or vestibular dysfunction.

Conclusion

For subjects with balance or vestibular disorders and control subjects, fair to excellent reliability in repeated measurements of most gait parameters across multiple trials of gait tasks described by the DGI was shown. Performance of the continuous walking tasks of the DGI was characterized by gait parameter differences that could distinguish subjects with balance or vestibular disorders from control subjects. The various effect size differences between the subjects with balance or vestibular disorders and the control subjects on the DGI continuous walking tasks confirmed a hierarchical structure for the level of difficulty inherent in the performance of the DGI items for people with balance or vestibular dysfunction. It appears that one performance of the DGI items provides valid results for analysis by a physical therapist.