Abstract
Background The Functional Gait Assessment (FGA), a measure of walking balance ability, was developed to eliminate the ceiling effect observed in the Dynamic Gait Index (DGI). Three presumably more difficult tasks were added and 1 easier task was removed from the original 8 DGI tasks. The effects of these modifications on item hierarchy have not previously been analyzed.
Objective The purpose of this study was to determine: (1) the ordering of the 10 FGA tasks and the extent to which they map along a clinically logical difficulty continuum, (2) whether the spread of tasks is sufficient to measure patients of varying functional ability levels without a ceiling effect, (3) where the 3 added tasks locate along the task difficulty continuum, and (4) the psychometric properties of the individual FGA tasks.
Design A retrospective chart review was conducted.
Methods Functional Gait Assessment scores from 179 older adults referred for physical therapy for balance retraining were analyzed by Rasch modeling.
Results The FGA task hierarchy met clinical expectations, with the exception of the “walking on level” task, which locates in the middle of the difficulty continuum. There was no ceiling effect. Two of the 3 added tasks were the most difficult FGA tasks. Performance on the most difficult task (“gait with narrow base of support”) demonstrated greater variability than predicted by the Rasch model.
Limitations The sample was limited to older adults who were community dwelling and independently ambulating. Findings cannot be generalized to other patient groups.
Conclusions The revised scoring criteria of the FGA may have affected item hierarchy. The results suggest that the FGA is a measure of walking balance ability in older adults that is clinically appropriate and has construct validity. Administration of the FGA may be modified further to improve administration efficiency.
Assessment of balance while walking is an essential component of a comprehensive examination of postural control. The Functional Gait Assessment (FGA), developed by Wrisley et al,1 is a clinical outcome measure often used to specifically assess balance while walking. The FGA was developed as a modification of the Dynamic Gait Index (DGI)2 in order to address some of the shortcomings of the DGI. In particular, the DGI had been shown to have a ceiling effect in people with vestibular disorders.3 In response to the known ceiling effect, the FGA was developed with the addition of 3 items that were expected to be more difficult, particularly for people with vestibular disorders, and the exclusion of 1 of the original 8 items of the DGI (“walk around obstacles”) that was thought to be of insufficient difficulty. The 3 additional items were “ambulating backwards,” “gait with eyes closed,” and “gait with narrow base of support.”1 The developers also modified the instructions for scoring of several items with the intention of reducing ambiguity, so that written instructions alone would be sufficient for clinicians to reliably administer the test. The developers found the FGA to have good interrater and intrarater reliability, with intraclass correlation coefficients (ICCs) of .86 and .74, respectively, and good internal consistency, with a Cronbach alpha of .79. They also established concurrent validity with other balance measures used for people with vestibular disorders, with Spearman rank order correlations ranging from .11 to. 67.1
Since its initial development, the psychometric properties of the FGA have been reported in a number of different patient populations, including older adults4,5 and people with Parkinson disease,4,6–11 stroke,12,13 and vestibular disorders.14 The psychometric properties examined include interrater and intrarater reliability (ICCs ranging from .83 to .90),5,11–13 criterion validity,4–6 and construct validity,4,6–10 including score criteria associated with increased risk for falls in older adults (≤22 points)4 and in people with Parkinson disease (≤15 points11 and ≤18 points10). Age-related norms also have been reported.5 Responsiveness to change has been examined in the FGA with the determination of the minimal detectable difference in people with vestibular disorders14 and the minimal clinically important difference in older adults.15 Despite this rather extensive study of the properties of the FGA, several psychometric properties of the FGA have not yet been examined. Further analysis may reveal important psychometric characteristics of the FGA, the knowledge of which may make it more clinically applicable and interpretable as a measure of walking balance ability.
The FGA was formulated as a measure of walking balance ability that presumably spans a continuum of successively more difficult tasks and that is composed of tasks that are hierarchical in their nature (ie, a person must be able to do easier tasks before the harder tasks can be accomplished). Although this was likely the intent for the FGA, the extent to which the FGA assesses a progression of walking balance tasks that proceed from easier tasks to more difficult tasks in a ladder-like progression spanning a continuum of clinically defensible and functionally relevant tasks has not yet been specifically demonstrated. Such a calibration of walking balance tasks according to difficulty would be clinically useful. For example, the hierarchical structure of the tasks may be used to establish the logical order of administration of items, beginning with the easiest tasks and proceeding through tasks in order of difficulty. Such an examination of the tasks has already been established for the DGI16 but not for the FGA. In addition, a clear understanding of the task hierarchy would inform patient management and treatment planning with respect to the sequence of interventional tasks that should be implemented based on a patient's performance at a point in time.
The developers of the FGA also likely intended to include tasks adequate to measure peoples' abilities along a meaningful functional continuum. Ideally, a scale should include items that measure distinct ability levels in a way such that there are no redundant items at a given level along the continuum, making the scale most efficient to administer with the fewest tasks necessary. Similarly, in order to measure people of varying ability levels, items should be evenly distributed along the continuum without gaps where there are no items measuring people at a certain ability level. If redundant items or gaps are discovered, these findings can provide the justification for modifying an instrument by the elimination of redundant items or addition of harder or easier items. Specifically in regard to the FGA, further analysis will determine whether the addition of the tasks “ambulating backwards,” “gait with eyes closed,” and “gait with narrow base of support” improved the spread of task difficulties and the underlying construct definition and effectively eliminated the ceiling effect seen with the DGI.1
In addition to examining task hierarchy, there is a need for examination of the psychometric properties of individual items of the FGA—that is, to determine the extent to which patients perform in a predictable way on each individual task. In general, patients who are more able would be expected to score better on more difficult tasks than patients who are less able would be expected to do. Conversely, patients who are less able may succeed on easier tasks but should perform more poorly on difficult items. An analysis of these expectations can reveal patterns of inconsistency in task performance, which may be due to either problematic scoring of a task or unexpected patient performance. This form of intensive FGA item analysis is indicated because the scoring procedures for several FGA items were changed from the original scoring scheme of the DGI,1 and these changes may have affected item difficulty.
The purpose of this study was to examine FGA scores derived from a sample of older adults referred for physical therapy for balance training to determine: (1) the ordering of FGA tasks and the extent to which they map along a meaningful and logical difficulty continuum, (2) whether the spread of tasks is sufficient to measure patients of varying functional ability level without a ceiling effect, (3) where the 3 tasks added by Wrisley et al1 (“ambulating backwards,” “gait with eyes closed,” and “gait with narrow base of support”) locate along the task difficulty continuum, and (4) the psychometric properties of the individual FGA tasks.
Method
Participants
Participants were patients who received physical therapy for balance retraining in 1 of 3 participating outpatient physical therapy clinics (ie, an independent clinic, a clinic associated with a tertiary care hospital, and a clinic associated with a community hospital). Time allotted for initial evaluations ranged from 45 to 60 minutes. Therapists completed the FGA in approximately 10 minutes. Study eligibility criteria were those established for the parent study.15 Inclusion criteria for patients were a minimum of 60 years of age, able to walk independently with or without an assistive device, and referral for physical therapy for balance retraining. The single exclusion criterion was a diagnosis of benign positional paroxysmal vertigo (BPPV). Patients with BPPV were excluded because the purpose of the parent study was to assess change over time, and most patients with BPPV were expected to recover after only 1 or 2 treatments. Data from patient medical records were manually retrieved, including initial FGA scores and patient demographics (sex, age, and reason for physical therapy referral). In total, 179 records were analyzed.
Procedure
All patients had the FGA performed as part of their initial physical therapy assessment. The 10 item FGA was administered according to the protocol described by Wrisley et al1 (see Appendix for FGA protocol and scoring). Each FGA item was scored on a 0 to 3 scale, with a score of 0 indicating “severe impairment” and a score of 3 considered “normal” and with a possible maximum total score ranging from 0 to 30. The developers of the FGA1 modified the scoring of several items compared with the scoring used for the DGI. See Appendix for a comparison of DGI and FGA scoring criteria.1
Data Analysis
The data were analyzed using Rasch modeling. Rasch modeling17 is an item response theory (IRT) approach to the examination of the psychometric characteristics of an assessment instrument. The Rasch approach is typically applied during the development of an instrument as a means of guiding the construction of rating scales and justifying the inclusion or exclusion of specific test items.17,18 It is still reasonable, however, to conduct a post hoc investigation of the psychometric properties of an instrument such as the FGA based on the Rasch model. Post hoc Rasch analysis has been applied to the DGI in 2 different patient groups.16,19
Through a Rasch analysis, ordinal scores such as those derived from the FGA are converted to interval measures. This conversion allows for the positioning of items along a continuum of difficulty in an ordered and measured sequence that indicates the relative distance between items in interval units. Positioning the test items (or a test item) allows estimates to be made of a patient's functional ability that is relative to the other items in a sequential, interval scale.
The Rasch rating scale model17,20,21 was used for the analysis of the FGA scale. This model is appropriate when the response categories (severe impairment [0], moderate impairment [1], mild impairment [2], normal [3]) are intended to have the same meaning for all items. That is, the understanding of what differentiates a “moderate impairment” rating from the higher “mild impairment” rating is assumed to hold regardless of the specific gait task.
The rating scale model generates an “ability to balance while walking” estimate for each patient, a “difficulty of scoring normal” estimate for each task, and a “threshold” estimate representing the difficulty of scoring in successively higher performance categories. These 3 sets of estimates are reported in interval units of measure called “logits” (log-odds units).20,22 As shown below, the patient ability and task difficulty estimates simultaneously portray the locations of the patients and tasks on the continuum defining the functional gait assessment construct. The WINSTEPS software package (Winsteps, Chicago, Illinois) was used for the analysis.23
In the logit metric, higher-scoring patients (many “normal” scores on the tasks) have high positive-valued ability estimates; lower-scoring patients (fewer “normal” scores) have low negative-valued estimates. Harder FGA tasks (fewer “normal” scores by the patients) have high positive difficulty estimates, and easier FGA tasks (many “normal” scores) have low negative estimates. Higher levels of performance that are relatively harder to attain (from “mild impairment” to “normal”) have high positive threshold estimates, and lower levels of achievement that are relatively easier to move out of (from “severe impairment” to “moderate impairment”) have negative estimates.
Based on these estimates, the model generates a predicted score for each patient on each task. The difference between the score a patient actually receives on a task and the score predicted by the model is a “residual.”20 A positive-valued residual occurs when the patient scores higher than predicted, and negative residuals occur when the patient's observed score is lower than predicted. Rasch model goodness-of-fit analyses rely principally on summary statistics generated from these residuals.
The residuals may be standardized, weighted by the variance of the predicted scores, and then summed to produce the “ZSTD Infit” in the WINSTEPS software. The unstandardized, variance-weighted version (the “mean square Infit”) also is typically reported. These information-weighted fit statistics (ie, “Infit”) detect unexpected scores for patients with ability levels closely matched to the difficulty levels of the task they have performed. When the residuals are not variance-weighted but are still summed, they form the “ZSTD Outfit” and “mean square Outfit” statistics. These outlier identification fit statistics detect unexpected scores for patients with ability levels much higher or lower than the difficulty levels of the tasks.
We typically use flexible criteria of +3 for the ZSTD statistics and +1.4 for the mean square statistics to flag misfit between the patients' actual and predicted scores on a task. These criteria are set relatively low in our analysis in order to avoid missing surprising scores that might suggest problems with the items or unexpected success or failure on the part of the patient. Although large negative ZSTD (<−3.0) and small positive mean square (<0.7) statistics suggest score variation on an item that is more consistent than predicted by the model, our attention is primarily focused on those items that generated unexpected or inconsistent scores.
Finally, a principal component analysis (PCA) of the patient-by-item residual matrix24,25 and a parallel analysis of simulated data were conducted as checks on the assumed unidimensional structure20 of the FGA. If the patients' scores fit the Rasch model, the residuals on the items will be uncorrelated with each other.20 A plot of the first 2 principal component item loadings extracted from a patient-by-item residual matrix will then resemble a circular “zero-like” pattern.24 Furthermore, the corresponding first 2 component eigenvalues will be comparable in magnitude to the first 2 eigenvalues extracted from a parallel analysis of simulated data.25
Results
Descriptive demographic statistics generated for FGA score, age, sex, and reason for physical therapy referral using IBM SPSS Statistics version 21 (IBM Corp, Armonk, New York) revealed that 36% of the patients were male and that they were, on average, 79.0 years of age (SD=7.0). The median FGA score was 18, with a range of scores from 5 to 29. The most frequent reason for physical therapy referral was gait instability (69.3%), followed by dizziness (6.7%) and impaired balance (5.0%). See Table 1 for participant descriptive demographics.
Participant Descriptive Information (N=179)a
The category response frequencies are reported in Table 2, along with their corresponding category threshold estimates (“taus”). These estimates reveal that it is relatively easy to move from “severe impairment” to “moderate impairment” (−2.03) but much harder to move from “mild impairment” to “normal” (2.24). This finding supports the use of the rating scale model for these Likert-like items and confirms that the FGA administrators were consistent in their understanding and use of the scoring categories.
Category Response Frequenciesa
The person score reliability20 (ie, Cronbach alpha) is .83. The item separation reliability20 is .99. These Rasch model findings support the overall consistency of the patient scores on the tasks and the well-spread variability in the task difficulty estimates.
Figure 1 presents the “Person-Item Map,” which is the simultaneous locating of patient walking balance ability and task difficulty estimates along the functional gait continuum. The vertical line represents the ordered progression of the patients and the FGA items, placed adjacent to their corresponding logit value and FGA score. At the bottom, to the left of the line, are lower-scoring patients, and to their right are the easier tasks. These patients should do moderately well on these easier items but have greater difficulty on the harder items. At the top, to the left of the line, are higher-scoring patients, and to their right are the harder tasks. These patients should have no problem with the easier items and only mild, if any, difficulty with the harder items. Table 3 contains the possible FGA scores and their corresponding logit estimates of walking balance ability.
Person-item map. Patients are plotted on the map with those who are less able locating at the bottom of the map and those who are more able locating at the top of the map. Each “#” represents a count of 2 people, and each “.” is a count of 1 person. Associated logit values and Functional Gait Assessment (FGA) scores are indicated to the left of the map. FGA items are plotted on the right side of the map, with increasing difficulty from bottom to top. The mean ability and mean difficulty of items are both denoted by “M.” “S”=1 standard deviation from the mean; “T”=2 standard deviations from the mean.
Equivalence of FGA Raw Scores and Logit Estimates of Abilitya
Patient walking balance ability ranged from −2.66 logits (FGA score=5) to +5.22 logits (FGA score=29), with a mean logit value of 0.6 (FGA score=18) (SD=1.36). Task difficulty effectively spanned across person ability levels, with the easiest task (“change in gait speed”) at −1.43 logits to the most difficult task (“ambulate with narrow base of support”) at +2.98 logits. The mean logit value of 0 serves as the anchor point for the person-item map.20 Six of the FGA items locate below the task mean, and 4 FGA items locate above it. Three FGA items—item 10 (“steps”), item 3 (“gait with horizontal head turns”), and item 9 (“ambulating backwards”)—cluster at or near the same difficulty level, with logit values of −0.21 to −0.25.
Two of the 3 items added by Wrisley et al1 were the most difficult tasks (item 7 [“gait with narrow base of support”] and item 8 [“gait with eyes closed”]). We also analyzed the effect of excluding the 3 items added by Wrisley et al,1 and the results are shown in the person-item map in Figure 2. Without these 3 tasks, a marked ceiling effect is evident, and the construct validity of the scale is weakened at the upper level of the task hierarchy.
Person-item map without Functional Gait Assessment (FGA) items 7 (“gait with narrow base of support”), 8 (“gait with eyes closed”), and 9 (“ambulating backwards”). Patients are plotted on the map with those who are less able locating at the bottom of the map and those who are more able locating at the top of the map. Each “#” represents a count of 2 people, and each “.” is a count of 1 person. Associated logit values are indicated to the left of the map. FGA items are plotted on the right side of the map, with increasing difficulty from bottom to top. The mean ability and mean difficulty of items are both denoted by “M.” “S”=1 standard deviation from the mean; “T”=2 standard deviations from the mean.
Goodness-of-fit statistics and logit values for each task are listed in Table 4. Only one of the FGA items demonstrated a noticeable misfit to the Rasch model. Item 7 (“gait with narrow base of support”) exceeded the mean square criterion of 1.4 and the standardized z criterion of 3.0 for both Infit and Outfit. On closer examination of the data, it was determined that 32 patients exceeded our misfit criteria, but of those 32, 16 demonstrated misfit on only item 7. Twelve patients were expected to score 0 but scored 2 or 3, and 4 patients were expected to score 2 or 3 but scored 0 or 1. Items 9 (“ambulating backwards”) and 10 (“steps”), in contrast, both generated Infit mean squares less than 0.70, suggesting that these items provoked less variability in scores than expected and may, in a statistical sense, be redundant and contribute little information in measuring the construct of balance while walking.
Individual Item Statistics (N=179)a
The principal component loading plot for the residuals was consistent, with a “zero-like” pattern, although item 3 (“gait with horizontal head turns”) and item 4 (“gait with vertical head turns”) loaded relatively high on component 1, and item 7 (the misfitting item identified earlier) loaded relatively high on component 2. The slight nonrandom variation remaining in the residuals for these 3 items was reflected in the eigenvalue analysis. The first 2 residual eigenvalues were 1.79 and 1.47, and the corresponding simulated data eigenvalues were 1.38 and 1.26, respectively. These findings indicate that the residuals, although not fully random, do not contain evidence of a strong secondary component independent of balance walking ability and, critically, that the FGA is “essentially unidimensional.”26
Discussion
This study presents the analysis of the FGA by Rasch modeling. To our knowledge, this is the first analysis of the FGA by this method. The Rasch approach allowed FGA items to be plotted along an interval scale in sequence based on the difficulty of the task performed. Although the hierarchical ordering of tasks of the DGI was known based on 2 previous Rasch analyses,16,19 it was not known what effect the additional scoring criteria of the FGA might have on the ordering of tasks compared with that seen with the DGI. The current findings demonstrated that the location of FGA items did generally situate along a meaningful and logical difficulty continuum with easier tasks, such as item 2 (“change in gait speed”) locating at the bottom of the person-item map and more difficult tasks, requiring a higher degree of postural control, such as item 6 (“step over obstacle”), item 8 (“gait with eyes closed”), and item 7 (“gait with narrow base of support”) all locating above the item mean.
One obvious exception to the items mapping in a logical order was the counterintuitive location of item 1 (“gait on level surface”) at +0.18 logits, just above the item mean. In a similar sample of older adults (mean age of 75 years compared with 78 years in the present study), by Rasch analysis of the DGI, Chiu et al16 found item 1 (“gait on level surface”) to be the easiest of the DGI tasks. Also, in a sample of younger people (mean age=56.7 years) with vestibular and balance impairments, Marchetti and Whitney,19 through Rasch analysis of the DGI, located item 1 as the third easiest task after “change in gait speed” and “stepping over obstacle.” One possible reason that the patients in the present study found item 1 (“gait on level surface”) more difficult to perform may be due to the change in scoring criteria for this item in the FGA compared with the scoring in the DGI. Wrisley et al1 added the amount of deviation from a 30.48-cm-wide (12-in-wide) pathway as an additional variable to the scoring rubric for the task. Patients who adopted a wider base of support as a compensation for instability would score lower on this task in the FGA than they would according to the DGI criteria. This additional criterion may have made it more difficult for some patients to achieve high scores on the task, and, as a result, the task located as a moderately difficult task.
Knowing the hierarchy of task difficulty of a clinical assessment could lead to a change in how the FGA is administered. It makes clinical sense to have a standardized approach to administering tests, including the order in which items are tested. Knowing the task hierarchy enables this ordering of test item administrations. Furthermore, knowledge of the task hierarchy could change administration of the FGA in different ways. For example, a person could start with administration of a mid-level task such as “walking on level surface,” and, if the patient succeeds or fails, more difficult or easy tasks should be administered, as indicated. This approach to item administration would make administration of the items more targeted and efficient. In a similar way, task hierarchy can inform treatment progression, where patients are progressed through incrementally more difficult tasks based on their initial level of performance. This approach to task training is consistent with principles of motor learning, where the ideal level of challenge (not too easy or too difficult) is provided to optimize learning and neuroplasticity.27
The current findings further validate1 that the FGA is not limited by the ceiling effect seen with the DGI,3 as the person-item map illustrates a spread of tasks generally corresponding to varied patients' ability levels along the continuum. Two of the 3 items added by Wrisley et al1 (“gait with narrow base of support” and “gait with eyes closed”) were the 2 most difficult tasks, so their expectations about the difficulty of these items were confirmed. However, one of the items that Wrisley et al1 added that they thought would be more difficult to perform is item 9 (“walking backwards”). This item actually located just below the mean at −0.25 logits in a group of moderately difficult tasks. The FGA developers may have thought that item 9 would be challenging because potential obstacles could not be visually anticipated and the patient could not fixate on a target he or she was moving toward. These findings suggest that perhaps the knowledge that there were no obstacles in the test walkway and still having visual reference points of the marked walkway available reduced the actual versus expected difficulty of the task.
The current analysis also included the association of logit estimates of person ability with his or her equivalent FGA raw scores. Criterion scores, like those associated with fall risk, can be located on the person-item map, as has been done with the Berg Balance Scale.28 A total score of 22, identified as a criterion score for fall risk in older adults,4 locates at +1.7 logits, between items 8 (“gait with eyes closed”) and 7 (“gait with narrow base of support”) (see dashed horizontal line on person-item map in Fig. 1). Therefore, if a patient succeeds on item 8, there is a high probability of that patient achieving an FGA total score greater than 22, ruling out fall risk.
The PCA of the residuals indicates that, in general, the FGA is “essentially unidimensional”26 in measuring a single, unified construct: balance while walking. Interestingly, the PCA also revealed that items 3 and 4 (“gait with horizontal head turns” and “gait with vertical head turns”) were unexpectedly correlated. This finding suggests that the head turn action in these 2 items may reflect an unintended source of influence on an individual's balance while walking. That is, as other authors16 have suggested, given the activation of the vestibular system in particular with head turns, we suspect these items may identify a subset of patients with vestibular impairment.
The psychometric statistics of 7 individual FGA items were within the acceptable range, indicating that they were useful for measuring the construct of balance while walking. Item 7 (“gait with narrow base of support”) generated notable misfit statistics, indicating unexpected performance on the item. Instability while standing in a narrow base of support has been associated with impairments in controlling mediolateral hip muscular forces29 and diminished proprioception30 in older adults. It is possible that individuals with these specific impairments were able to perform reasonably well on other FGA items, and this task identifies a specific subset of postural control impairments. These findings are consistent with those of the PCA, where item 7 loaded high on residual component 2, suggesting an unintended influence on an individual's balance while walking. Despite these findings, we would not recommend dropping item 7 from the FGA, because it appears to identify a specific patient subgroup and contributes to eliminating the ceiling effect seen in the DGI.
There are 2 items—item 9 (“ambulating backwards”) and item 10 (“steps”)—with low (<0.7) Infit mean square statistics, indicating that there was less variability in scoring on these tasks than the Rasch model would have predicted. These statistics indicate that these tasks may not be useful for differentiation between walking balance ability levels among individuals. Furthermore, these 2 tasks were redundant with the task “gait with horizontal head turns,” with all 3 tasks locating at the same level of the difficulty. The lack of variability on item 9 (“walking backwards”) is likely related to the same factors discussed earlier regarding its location on the difficulty continuum. For item 10 (“steps”), patients may have used a handrail out of habit, but not necessarily to meet functional task demands, thus promoting a higher than expected frequency of “mild impairment” (score of 2) responses according to the grading criteria. A differential item functioning analysis31 with and without assistive devices (eg, stair rail, canes) may be conducted as a next step in analysis of the FGA.
Application of these findings is limited to the population of community-dwelling older adults with known balance impairments, and findings cannot be generalized to all patient groups. Chiu et al16 and Marchetti and Whitney19 have already demonstrated that the hierarchy of DGI items differs in older adults compared with patients with primary vestibular disorders. There is a need, therefore, to establish an FGA item hierarchy in different patient populations, as individual postural control impairments may influence item ordering and psychometric properties.
In conclusion, the results suggest that the FGA is a measure of walking balance ability that is clinically appropriate and has construct validity for older adults with balance impairments. The items generally follow a logical and clinically meaningful difficulty continuum. The scoring criteria added by Wrisley et al1 may have resulted in increasing the difficulty of some FGA items. The addition of 2 difficult items to the FGA has eliminated the ceiling effect seen in the DGI.
Appendix.
Functional Gait Assessment Protocol and Scoringa
a Adapted and reproduced from: Wrisley DM, Marchetti GF, Kuharsky DK, Whitney SL. Reliability, internal consistency, and validity of data obtained with the Functional Gait Assessment. Phys Ther. 2004;84:906–918, with permission of the American Physical Therapy Association. The Functional Gait Assessment (FGA) was adapted from the Dynamic Gait Index (DGA) from: Shumway-Cook A, Woollacott MH. Motor Control: Theory and Practical Applications. Baltimore, MD: Lippincott Williams & Wilkins; 1995. Modified and reprinted with permission of the authors.
Footnotes
Dr Beninato provided concept/idea/research design, data collection, and project management. Both authors provided writing and data analysis.
The study was approved by the institutional review boards for human subject research of each of the 3 participating institutions.
- Received March 20, 2015.
- Accepted August 22, 2015.
- © 2016 American Physical Therapy Association