Abstract
Background The modified Dynamic Gait Index (mDGI), developed from a person-environment model of mobility disability, measures mobility function relative to specific environmental demands. The framework for interpreting mDGI scores relative to specific environmental dimensions has not been investigated.
Objective The aim of this study was to examine the person-environmental model underlying the development and interpretation of mDGI scores.
Design This was a cross-sectional, descriptive study.
Methods There were 794 participants in the study, including 140 controls. Out of the total study population, 239 had sustained a stroke, 140 had vestibular dysfunction, 100 had sustained a traumatic brain injury, 91 had gait abnormality, and 84 had Parkinson disease. Exploratory factor analysis was used to investigate whether mDGI scores supported the 4 environmental dimensions.
Results Factor analysis showed that, with some exceptions, tasks loaded on 4 underlying factors, partially supporting the underlying environmental model.
Limitations Limitations of this study included the uneven sample sizes in the 6 groups.
Conclusions Support for the environmental framework underlying the mDGI extends its usefulness as a clinical measure of functional mobility by providing a rationale for interpretation of scores that can be used to direct treatment and infer change in mobility function.
Community ambulation requires not only the ability to walk at a minimum gait speed but also the ability to adapt gait to diverse and complex task demands.1,2 The Dynamic Gait Index (DGI) was developed to examine functional gait, including both unimpeded gait and more complex walking tasks requiring the ability to modify and adapt gait to both expected and unexpected environmental challenges. The DGI is based on a person-environment model of mobility disability in which environmental demands are categorized into 8 dimensions: distance, temporal, ambient, terrain, physical load, attention, postural transitions, and density, representing the external demands required for an individual to be mobile within a particular environment.1 The DGI includes tasks that examine walking in 4 of the 8 environmental dimensions: temporal, postural transitions, terrain, and density. A modified version (mDGI) was recently developed and tested on a large and diverse group of individuals with and without mobility disability.3,4 Results from these studies advocated the retention of the original 8 tasks but expanding the scoring system to evaluate 3 correlated but unique facets of walking performance: gait pattern (GP), level of assistance (LOA), and time.3 Previous research on the mDGI has focused on its psychometric properties and its validity for use with individuals from different medical diagnostic groups.3,4 However, research examining the 4-factor environmental model underlying the DGI and mDGI is lacking.
Thus, the purpose of this study was to examine the structural validity of the environmental dimensions framework upon which the mDGI was built. We expected that specific task scores would reflect 4 latent factors representing the 4 environmental dimensions. Specifically, we expected scores from the usual pace and change pace tasks to load on the temporal dimension. We expected that scores from the horizontal head turn, vertical head turn, and pivot turn tasks would load on the postural transitions dimension; that scores from the around obstacles and the over obstacles tasks would load on the density dimension; and that the stairs task would stand alone as a measure of the terrain dimension. It was hoped that examining the underlying framework used to interpret DGI and mDGI scores would improve the clinical usefulness of the test as a measure of mobility function in both geriatric and neurologic populations.
Method
An in-depth review of the methods used to investigate the reliability and validity of the mDGI was previously reported.3 A brief overview of the study's methods is presented here.
Recruitment
An email was sent via the American Physical Therapy Association Neurology Section's electronic mailing list to recruit potential clinical sites. From within each of these sites, participants with neurologic impairments who were currently receiving physical therapy for balance and mobility problems were evaluated using the DGI and mDGI. Inclusion criteria included the ability to walk 6.1 m (20 ft) without physical assistance of another person. The use of an assistive device was permitted. For the control sample, a convenience sample of adults was recruited from volunteers responding to a flyer posted in the University of Washington Department of Rehabilitation Medicine and in retirement communities in the greater Seattle-Bellevue area. Inclusion criteria were: age between 15 and 99 years, no neurologic diagnosis, ability to walk without the physical assistance of another person for a distance of 6.1 m, and ability to give informed consent. Participants provided informed consent prior to testing. The analyses presented in this article used a subset of data (794 participants) presented in a previous publication.3
Modification of the DGI
All 8 items from the original DGI were retained, with minor modifications made to distance ambulated (6.1 m) and 4 test items (change pace, over obstacles, pivot turn, and stairs).3 The original scoring system was modified to establish ordinal scores for 3 separate aspects of walking performance: GP (0–3), LOA (0–2), and time (0–3). Performance scores at the task level were calculated by adding the scores for time, GP, and LOA, resulting in a score ranging from 0 to 8 for each of the 8 tasks. A total score for each of the 3 facets of performance was calculated, characterizing walking performance with respect to time (range=0–24), GP (range=0–24), and LOA (range=0–16). A total score for the mDGI also was calculated by combining the 3 performance scores, for a total score range of 0 to 64.
Investigating the Environmental Dimensions as Explanations for Performance on mDGI Tasks
To investigate whether mDGI scores supported the 4 environmental dimensions that were the basis for DGI tasks, we conducted an exploratory factor analysis (EFA) using IBM SPSS software, version 19 (IBM Corp, Armonk, New York). Factor analysis is a procedure that uses correlations among scores from different variables to identify groups of scores that may be explained through latent (underlying) common factors.5,6 We investigated the viability of factor analysis using 2 tests: the Kaiser-Meyer-Olkin (KMO) test and the Bartlett test of sphericity. The KMO test indicates the proportion of variance in the data that may be caused by underlying factors. Values close to 1.0 suggest that factor analysis may be useful. If the KMO test value is less than .50, a factor analysis is unlikely to be useful in evaluating the dimensionality of the data. The Bartlett test evaluates the correlation matrix to determine whether it is an identity matrix (ie, the variables are unrelated). The distribution of correlation coefficients with the Bartlett test is approximately a chi-square distribution; therefore, results are interpreted as χ2 values. If the test is significant at P<.05, a factor analysis may be useful.5,6
After the initial solution, an orthogonal (varimax) rotation was used to determine which variables group together. Factor loadings are correlations between variable scores and the underlying latent factors, and are used to assess the strength of the association between each variable and each latent factor. The original model for the DGI was based on 4 environmental dimensions; therefore, we set the number of latent factors at 4 and used EFA to explore how the task scores grouped relative to those 4 factors. An exploratory rather than confirmatory factor analysis was chosen because it was felt we did not have sufficient items to warrant a confirmatory analysis. To optimize our ability to identify distinct latent factors, we used a varimax rotation, which assumes that the latent factors are not correlated with one another. We included the control cohort in the EFA because the mDGI is designed to examine mobility performance across a spectrum of individuals with and without mobility limitations. We used a criterion of 0.5 to identify significant factor loadings for each of the 4 factors with an eye to determining whether task scores grouped as expected.5,6 A factor loading of 0.5 represents 25% of shared variance between the rotated factor and the mDGI subscore.
Role of the Funding Source
This study was supported by a grant from the Walter C. and Anita Stolov Research Fund, Department of Rehabilitation Medicine, University of Washington.
Results
Participants/Sociodemographics
The analyses presented in this paper used a subset of data (794 participants) presented in a previous publication.3 The subset included individuals with the following diagnoses: stroke (n=239), vestibular dysfunction (n=140), traumatic brain injury (TBI) (including head injury and concussion) (n=100), gait abnormality (patients referred for physical therapy under the diagnostic code 781.2 of the International Classification of Diseases7) (n=91), and Parkinson disease (n=84), as well as the nonneurologically impaired control cohort (n=140). Table 1 summarizes sociodemographics by diagnostic group. Table 2 presents scores (mean, standard deviation, and minimum/maximum) on the 8 mDGI tasks for the 6 groups.
Sociodemographics of Sample by Group
Modified Dynamic Gait Index Task Scores in the 6 Groups
Patterns of Performance as a Function of Environmental Dimensions
For the data used in this study, the KMO and Bartlett tests supported implementation of a factor analysis (KMO test=.95, Bartlett test=8034.75; P<.01). Given these results, we proceeded with the EFA.
Results from the factor analysis showed that a single-factor solution explained 86% of the variance in mDGI task scores. The 4-factor solution explained 95% of the variance in task scores. After rotation, each factor explained the following variability in scores: factor 1=29%, factor 2=28%, factor 3=20%, and factor 4=18%.
Table 3 shows the final rotated component matrix and the factor loadings (ie, the correlations between task scores and the underlying factors) for the task scores. For the most part, the correlations suggested that task scores loaded as hypothesized. Scores from 3 tasks loaded on the first factor: horizontal head turns, vertical head turns, and pivot turn. These 3 tasks were designed to measure the postural transition environmental dimension; therefore, we named the first factor “posture.” Scores from 4 tasks loaded on the second factor: usual pace, change pace, pivot turn, and around obstacles. Given that usual pace and change pace were intended to measure the temporal environmental dimension, we named the second factor “temporal.” Neither pivot turn nor around obstacles tasks were originally intended to measure the temporal dimension, yet loaded on (ie, were strongly correlated with) this factor.
Dynamic Gait Index (DGI) Task Factor Loadingsa
One task, stairs, loaded on the third factor; as the stairs task was originally intended to measure the terrain environmental dimension, we named the third factor “terrain.” Finally, one task, over obstacles, loaded on the fourth factor. This task was intended to measure the density environmental dimension (avoidance of static and dynamic obstacles); therefore, we named the fourth factor “density.” In the original design of the DGI, the around obstacles task was intended to measure the density environmental dimension; however, this analysis did not support that design. In addition, the pivot turn task loaded on 2 factors: temporal and postural.
Discussion
The utility of any clinical measure requires clear guidelines for administering, scoring, and interpreting results. Previous articles have focused on expanding and clarifying the scoring system of the mDGI.3,4 The current study extended this research by investigating the underlying environmental dimensions framework used to select the tasks of the DGI (and mDGI) and to interpret scores. In addition, this study explored the clinical implications of this model in the assessment and treatment of mobility disability.
Evidence for an Environmental Model of the DGI
The DGI, as well as the mDGI, was designed to measure overall ability to engage in complex walking tasks requiring dynamic balance. However, walking takes place in a variety of environmental contexts. Therefore, the DGI was designed to represent a range of those contexts. The theory underlying the choice of tasks involves 8 environmental dimensions (attention, temporal, distance, ambient, physical load, terrain, density, and postural transitions).1 Four of the 8 dimensions are represented in the DGI. One major goal of this study was to obtain evidence to support the validity of the theory underlying the measure. In the EFA, a single-factor solution explained 86% of the variance. Therefore, one factor is sufficient to represent the primary function of the measure—to assess dynamic balance underlying complex walking task performance. Evidence to support the environmental dimensions selected as the focus for the DGI was found when a 4-factor solution was run. This model explained 95% of the variance and allowed examination of the validity of the environmental model. For the most part, tasks loaded as expected on the 4 environmental dimensions that were the basis for development of the DGI.
Patterns of Performance as a Function of the 4-Factor Environmental Model
As expected, horizontal head turn, vertical head turn, and pivot turn loaded together on (correlated with) a single factor. However, the pivot turn task loaded on an additional factor. This finding had not been expected. In the original design of the DGI, these tasks were thought to represent the postural transitions dimension. In the current study, scores on 1 or more of the 3 tasks in the postural transitions dimension were significantly different from the control group in all 5 groups. This finding is consistent with previous research demonstrating that, compared with older adults without disability, older adults with mobility disability make fewer postural transitions during community ambulation.8 This finding suggests that an inability to meet demands in the postural transition dimension is an important contribution to the development of mobility disability in both geriatric and neurologic populations. Several researchers have suggested that the vertical and horizontal head turn and pivot turn tasks reflect a subset of tasks that place special demands on the vestibular system and, therefore, should be in a separate category from the broader category of postural transitions.9–11 Our data do not support this proposal. The range of performance on these tasks across all groups was 0 to 8, indicating that not all individuals within the vestibular group performed poorly on these 3 tasks (as shown by a score of >6), whereas some individuals in the other diagnostic groups did (as shown by a score of <2). Thus, we do not believe that poor performance on these tasks can be attributed solely to problems within the vestibular system, and we suggest they remain in the broader category of postural transitions.
Consistent with the original proposed framework, scores on the usual pace and change pace tasks loaded together on a second factor: the temporal dimension factor. However, both the around obstacles and pivot turn scores did as well. This finding was not consistent with the original framework, which placed the pivot turn task within the postural transition dimension and the around obstacles task within the density dimension. Factor analysis is a process by which observed scores are grouped based on statistically similar patterns. The finding that these 4 tasks grouped together in the factor analysis suggests that the physiological demands for these 4 tasks were similar. All 4 tasks are performed on a firm, flat surface, with patients pacing themselves. Gait pattern scores for these tasks have the lowest (easiest) difficulty parameters on the underlying Rasch scale (see Rasch person-item map figure in Shumway-Cook et al3), suggesting that all 4 of these tasks require minimal modification to the GP. In contrast, over obstacles and stairs tasks require modifying the GP in order to manage changes in terrain. Horizontal and vertical head turn tasks require listening for and timing movement in response to the tester's cues. Thus, these 4 tasks are all more challenging, although for different reasons.
Finally, one task, over obstacles, loaded on the fourth factor. This task was intended to measure the density dimension (avoidance of static and dynamic obstacles); therefore, we named this factor “density.” In the original design of the DGI, the around obstacles task also was intended to measure the density dimension; however, this analysis did not support that design. As discussed above, it is evident that the pattern of scores for the around obstacles task was more similar to score patterns for usual pace, change pace, and pivot turn tasks than the pattern of scores for the over obstacles task.
Clinical Applications to Intervention
This research provides partial support for the underlying environmental model upon which tasks for the original DGI were selected. Support for this model is important in light of the proposed framework for treating mobility disability based on information from the DGI.2 In this framework, performance on the 8 DGI tasks is used to identify the specific environmental dimensions contributing to mobility disability within an individual. This information guides the selection of therapeutic strategies to improve mobility function within the specific dimensions identified through testing. The DGI tasks are meant to be indicators of function within a specific environmental dimension, not an all-inclusive list of tasks within that dimension. Thus, when a limitation is identified within a specific dimension, a range of tasks, not just those identified in the mDGI, should be incorporated into a therapeutic program. For example, previous research has indicated that community mobility requires many postural transitions, including starts and stops; change in posture (eg, sit-to-stand, sit-to-walk); changing direction; walking forward, backward, and sideways; modifying step width and length; reorientation of head independent of a change in direction; leaning over or forward; and reaching. Postural transitions are considered an integral part of community mobility and impose demands on the balance control system over and beyond those encountered during steady state walking.1,8 Therefore, a therapeutic program to improve function in the postural transitions dimension should include practicing not just the tasks specified in the DGI (eg, horizontal and vertical head turns and pivot turns) but also any task within this dimension that is impaired.
A limitation of the DGI (and by extension the mDGI) is that the current selection of tasks represents only 4 of the 8 environmental dimensions proposed by Patla and Shumway-Cook.1 The 4 dimensions not found in the DGI were the distance dimension (the ability to walk a minimum distance of 400 m), ambient dimension (the ability to walk in various lighting and weather conditions), attention dimension (the ability to walk under multitask or novel conditions), and physical load dimension (the ability to walk and carry static or dynamic loads and the ability to push or pull heavy objects such as doors). Thus, therapeutic programs to improve mobility function based solely on performance relative to the 4 dimensions from the DGI (or mDGI) may be incomplete and should consider an individual's performance on not only a broader range of tasks within the 4 represented dimensions but also tasks in the other 4 dimensions. Finally, we suggest that future efforts to modify the mDGI should focus on the addition of mobility tasks in the nonrepresented environmental dimensions rather than the addition of more tasks in environmental dimensions already represented in the measure.
Limitations
This study had several limitations. First, there were different sample sizes in each of the groups included in this analysis. The results may be biased by the particular individuals in each group, particularly for the groups with the smaller sample sizes. Individuals recruited for the current patient samples were volunteers and were actively receiving physical therapy. Therefore, their results may not be representative of the performances of the larger patient population. Another limitation was that several groups included in the original sample were not included in this analysis due to insufficient sample size. Therefore, the results cannot be generalized to all patient populations. A third limitation is that specific information related to each medical diagnosis (eg, time since onset, severity of condition) was not available.
Future Research
Future research should focus on the sensitivity and specificity of the mDGI in predicting fall risk using prospective fall data. In addition, research establishing the minimal clinically important difference is needed. Increasing our understanding of the performance facet scores is needed, including the treatment strategies used to affect each aspect of performance and how change in one aspect of performance affects the others as well as the total score.
In conclusion, scores from the mDGI provide insight into some of the underlying environmental dimensions contributing to mobility disability in an individual patient. This information can be used to determine priorities and select treatment strategies to improve mobility function, specifically the ability to adapt gait to environmental demands associated with walking in the community. The strength of the research on the mDGI supports its utility in both clinical and research settings.
Footnotes
All authors provided concept/idea/research design and writing. Dr Shumway-Cook and Dr Matsuda provided data collection. Dr Shumway-Cook and Dr Taylor provided data analysis. Dr Shumway-Cook provided project management, participants, and facilities/equipment.
The University of Washington Human Subjects Division approved all procedures.
This study was supported by a grant from the Walter C. and Anita Stolov Research Fund, Department of Rehabilitation Medicine, University of Washington.
- Received February 14, 2014.
- Accepted November 26, 2014.
- © 2015 American Physical Therapy Association
Issue highlights
