Abstract
Background Evaluation of walking capacity and risk of falls in people with multiple sclerosis often are performed in rehabilitation. The Dynamic Gait Index (DGI) evaluates walking during different tasks, but the feasibility in identifying people at risk for falls needs to be further investigated.
Objective The objective of this study was to investigate (1) the construct validity (known groups, convergent, and discriminant) of the DGI and (2) the accuracy of predicting falls and establishing a cutoff point to identify fallers.
Design This trial was a multicenter, cross-sectional study.
Methods A convenience sample was composed of 81 people with multiple sclerosis with subjective gait and balance impairment who were able to walk 100 m (comparable to Expanded Disability Status Scale 1–6). Mean age of the participants was 49 years; 76% were women. The 25-Foot Timed Walk Test, Timed “Up & Go” Test, Four Square Step Test, Timed Sit-to-Stand Test, MS Walking Scale, Multiple Sclerosis Impact Scale, and self-reported falls during the previous 2 months were used for validation, to establish cutoff points for identifying fallers, and to investigate predictive values.
Results Significantly lower DGI scores (P≤.001) were found for participants reporting falls (n=31). High sensitivity (87%) in identifying fallers was found, with a cutoff score ≤19. The positive predictive value was 50%, and the negative predictive value was 87%. The positive likelihood ratio was 1.77, and the negative likelihood ratio was 0.26. The convergent validity was moderate to strong (ρ=0.58–0.80), with the highest correlation coefficient found for the 25-Foot Timed Walk Test. Discriminant validity was shown with low correlation for the psychological subscale of the Multiple Sclerosis Impact Scale.
Limitations The sample included ambulatory people participating in a randomized controlled trial investigating balance training.
Conclusions The DGI is a valid measure of dynamic balance during walking for ambulatory people with multiple sclerosis. With the cutoff point of ≤19, sensitivity was high in identifying people at risk of falls.
Multiple sclerosis (MS) is a chronic, progressive inflammatory disease that primarily affects the central nervous system. It is the most common neurological disorder in young adults and affects women more than twice as often as men.1 A majority (in some studies as many as 85%) of people with MS have difficulty with walking and have ambulatory deficits.2,3 Limited walking capacities can be seen as slower walking speed, shorter stride length, and restricted dynamic balance.2,4,5 Accidental falls are common in people with MS. Recent surveys report that more than 50% of people with MS have had falls and that at least one third are recurrent fallers.6–8 Evaluation of walking capacity, dynamic balance, and the risk of falls, therefore, are commonly performed in rehabilitation settings and provide information in justifying the need for individual treatments and planning treatments.
A measure that can be used for evaluating ambulatory and balance deficits is the Dynamic Gait Index (DGI), developed by Shumway-Cook and Woollacott.9 The DGI is a multitask, 1-dimensional outcome measure developed to investigate mobility and dynamic balance during walking in older people and to evaluate the risk of falling. The measure consists of 8 items with different walking tasks such as changing walking speed, walking and turning the head, walking around or over obstacles, and walking with pivot turn. Published translations include Danish10 and Finnish.11 With the use of Rasch analysis, Chiu et al12 showed that the 8 items of the DGI represented a single construct. Reliability was high in studies of older people,10 stroke,13,14 vestibular disorders,15 Parkinson disease,16 and MS.17,18 In items with a more subjective scoring such as walking with head turns, lower inter-rater and test-retest reliability were seen.13,18 Ceiling effects of 5% to 10% have been reported.11,14,19
Construct validity reflects the ability of a measure to measure a concept.20 The concept of the DGI is dynamic balance during walking tasks. According to Portney and Watkins,20 evidence on construct validation can be gathered by a variety of methods, among them known-groups validity (ability of a measure to discriminate between individuals), convergent validity (2 measures believed to reflect the same phenomenon), and discriminant validity (2 measures believed to assess different characteristics).
Falls are a major concern for people with MS, and it is of value to investigate known-groups validity of the DGI, if the measure discriminates fallers from nonfallers. Cattaneo et al19 found that fallers (falls in the month before the test) scored significantly lower on the DGI than did nonfallers. Also, in studies including older people, individuals with a history of falls had significantly lower scores on the DGI than did people with no falls.21,22 Different cutoff scores for identifying fallers have been suggested. Shumway-Cook et al,22 in a study of older adults, found that a score of ≤19 classified 59% of those with a history of falls, whereas Cattaneo et al,19 in a study of MS, found that a score of 12 identified 45% of those with falls. Identifying people with a risk of falls is clinically important, and the known-groups validity of the DGI needs to be further investigated. The feasibility of the DGI in predicting fallers versus nonfallers has not been investigated.
Convergent validity of the DGI has been confirmed in studies of older adults who were healthy21 and those with stroke13,14 or vestibular disorders23 through high correlation coefficients with other measures of balance and walking capacity. In people with MS, Cattaneo et al19 showed acceptable correlation coefficients between the DGI and the Berg Balance Scale (ρ=.78), the Timed “Up & Go” Test (TUG) (ρ=−.72), the Deambulation Index (ρ=−.80), and the Activities-specific Balance Confidence Scale (ρ=.54). McConvey and Bennett18 reported a high correlation (r=−.80) between the DGI and a timed 6.1-m walk for individuals with MS. Convergent validity of a Swedish version of DGI has not been assessed. Cattaneo et al19 investigated discriminant validity of the DGI and found low correlation coefficients for the Dizziness Handicap Inventory (ρ=−.39). However, discriminant construct validity of the DGI in relationship with specific measures for MS has not been investigated.
The aim of the present study was to investigate construct validity of the DGI in ambulatory people with MS in Sweden: known groups (fallers versus non-fallers), convergent, and discriminant. The aim also was to investigate the accuracy of predicting falls and establishing a cutoff point to identify fallers.20 Validity was investigated with the use of both commonly used clinical measures (TUG,24 Timed 25-Foot Walk Test,25 Four Square Step Test [FSST],26 and Timed Sit-to Stand Test,27) and diagnosis-specific measures (MS Walking Scale3 and Multiple Sclerosis Impact Scale [MSIS]).28 The MSIS measures the impact of the disease, and the results can be divided into 2 subscales: physical and psychological. On the basis of earlier studies,18,19 our hypothesis was that correlation coefficients would be moderate between scores on the DGI and timed tests. The convergent validity of the DGI with the MS Walking Scale and the MSIS physical subscale has not been established. Discriminative validity of the DGI was investigated with the use of the psychological subscale of the MSIS, in which a low correlation coefficient was expected.
Method
Design
A multicenter, cross-sectional data collection was conducted from September 2010 to June 2011.
Study Sample
Patients were recruited to participate in a randomized controlled trial investigating balance training for people with MS.29 The randomized clinical trial29 was registered in a Swedish clinical trials database (FoU i Sverige, ID 48641) and in Clinical Trials (NCT01299025). Data from the first assessment before randomization were used in the present validation study. The inclusion criteria were a diagnosis of MS by a neurologist according to the McDonald criteria,30 subjectively and objectively experienced balance and walking limitations, and ability to walk 100 m with an assistive device, the latter condition being comparable to Expanded Disability Status Scale (EDSS) scores of 1.0 to 6.0.31 The exclusion criteria were cognitive dysfunction or difficulty understanding Swedish that prohibited the administration of self-reporting scales.
Participants were identified from the Swedish MS Registry or from being known at one of the participating centers, which were located in 4 county council areas, both rural and urban. The research physical therapists responsible for the randomized controlled trial at each hospital/primary health care center provided verbal and written information about the study to patients fulfilling the criteria.
Eighty-one participants from 4 hospitals and 2 primary health care centers in the central part of Sweden were included in this study: 37 from the Örebro University Hospital/Nora and Brickegården primary health care centers in Örebro county council, 20 from the Västmanland Hospital in Västerås, 14 from the Mälar Hospital in Eskilstuna, and 10 from the Central Hospital in Karlstad (Fig. 1).
Flowchart of the recruitment of participants.
Translation Process
The DGI was translated into Swedish by 2 of the authors of this study (A.F. and Y.E.N.), with permission from the test developer. The translation process followed the recommendations of Streiner and Norman.32 The Swedish translation then was back-translated to English by a translator whose native language is English. The Swedish translation is close to the original presentation of the DGI.9
Measures
Demographic characteristics, use of assistive walking device indoors and outdoors, and the number of self-reported falls during the previous 2 months were registered. A fall was defined as an unexpected contact with any part of the body with the ground.33
The DGI consists of 8 items: walking on level surface, walking when changing speed, walking with horizontal head turns, walking with vertical head turns, walking with pivot turn, walking with stepping over obstacles, walking around obstacles, and stair climbing. Performance on each item is rated on a 4-point scale ranging from 0 (severe impairment) to 3 (normal walking ability without a walking aid). The maximum total score is 24.9
The TUG is a well-known measure of functional mobility.24 Originally developed for elderly people, it has been shown to be a valid measure of functional mobility in ambulatory people with MS.19,34 In the TUG, the time taken to rise from a chair, walk 3 m, turn around a cone, walk back, and sit down was measured. Each participant performed 1 practice trial to become familiar with the test, and the time on the second trial was recorded. Fourteen participants used a walking aid.
The 25-Foot Timed Walk Test is a part of the Multiple Sclerosis Functional Composite.25 From a static start, the participants walked a distance of 25 ft (7.62 m) at a fast but safe speed. Use of a walking aid was allowed, and 9 participants took this option. Two trials were performed, and the mean value was used.
The FSST measures the ability to quickly step over obstacles in different directions.6,26 Four sticks of 2.5-cm height were placed on the floor at 90-degree angles to each other. The participants were asked to perform a sequence of stepping into the squares (forward, sideways, backward, and sideways), first clockwise and then back without touching the sticks, and the time was recorded. Two trials were performed, and the mean value was used. Nine participants used a cane or crutch as support.
The Timed Sit-to-Stand Test was used to assess functional muscle strength in the lower extremities.27 Participants were asked to rise 10 times from a chair with handrails, and the time was measured.
The MS Walking Scale is a self-report scale on which the participants rate limitations of their walking ability for the last 2 weeks caused by MS.3 The scale includes 12 items, and each item is rated from “not at all limited” to “extremely limited.” A score of 0 to 100 is then calculated, with a higher score representing more severe limitation. A Swedish validated translation was used.35
The MSIS is a self-reported measure of the impact of MS, on which participants rate how much they had been limited or bothered by symptoms of MS during the last 2 weeks.28 The MSIS includes 29 items, and each item is rated from 1 (“not at all”) to 5 (“extremely limited/bothered”). The first 20 items are summed into a physical subscale, and the last 9 items are summed into a psychological subscale. A Swedish translation of the MSIS was used.36
Procedure
All measures were performed in a standardized order at a single test occasion between September 2010 and June 2011: the DGI, the TUG, the 25-Foot Timed Walk Test, the FSST, and the Timed Sit-to-Stand Test. The self-report scales, MS Walking Scale and MSIS, were filled in by the participants at the test occasion. Five research physical therapists, specially trained for this study, performed the data collection. All of the physical therapists had more than 10 years of experience in neurological physical therapy. All participants signed an informed consent form.
Data Analysis
Descriptive statistics were used to describe the characteristics of the study population. Because of skewed data distribution, medians and interquartile ranges (IQR) were primarily used to present the results. The Mann-Whitney U test was used for analyzing differences in DGI scores between individuals reporting falls or not reporting falls. Logistic regression analysis was carried out to determine the cutoff point that predicted the probability to be classified as a faller of ≥0.5. Predictive values were calculated. A positive predictive value (PV+) is the proportion of individuals with test results below the cutoff point who were correctly classified as fallers. A negative predictive value (PV−) is the proportion of individuals with scores above the cutoff point who were correctly classified as nonfallers. The likelihood ratio summarizes how many times more a person who experiences falls has results on the DGI worse than or equal to the cutoff. A positive likelihood ratio (LR+) is the true positive rate divided by the false positive rate. The negative likelihood ratio (LR−) is the false negative rate divided by the true negative rate. To evaluate convergent validity of the DGI, Spearman correlation coefficients were calculated between the DGI score and scores on the TUG, the 25-Foot Timed Walk Test, the FSST, the Timed Sit-to Stand Test, the MS Walking Scale, and the physical subscale of the MSIS. Discriminant validity was evaluated with the use of Spearman correlation coefficients between scores on the DGI and the psychological subscale of the MSIS. Correlation coefficients <.30 were interpreted as weak, .30 to .59 as moderate, and ≥.60 as strong.37 Probability values ≤.05 were considered statistically significant. The SPSS version 15.0 (SPSS Inc, Chicago, Illinois) was used for data analysis.
Role of the Funding Source
This work was funded by the Uppsala-Örebro Regional Research Council, the Research Committee of Örebro County Council, and the Norrbacka-Eugenia Foundation.
Results
Demographic characteristics and use of assistive walking devices of the 81 participants are shown in Table 1. The ages of the participants ranged from 19 to 73 years. The majority were women. All participants lived in their own housing. Fourteen participants (17.3%) reported comorbidity such as back pain and arthritis in the knee, but these disorders did not prevent them from fulfilling the tests.
Participant Characteristics (N=81)
Medians and ranges of scores on the DGI are presented in Table 2. The easiest items to perform were item 5 (“Gait with pivot turn”) and item 7 (“Step around obstacles”), with 64% and 70% of the participants, respectively, achieving the maximum item score of 3. Distribution of scores was skewed, with one third of the participants achieving a total score of ≥21.
Dynamic Gait Index: Median, Range of Scores, and Numbers With Maximum Score of 3 (N=81)
Participants reporting 1 or more falls during the previous 2 months (n=30; 37.0%) had significantly (P≤.001) lower scores on the DGI than those who reported no falls. The median score on the DGI was for 15.5 (IQR=12–18) for fallers and 20 (16–22) for nonfallers. Using logistic regression, with falls as the dependent variable and score on the DGI as independent variable, a cutoff value of 19 was calculated and presented on a receiver operating characteristic curve (Fig. 2). With the cutoff score of ≤19 on the DGI, sensitivity was 87% and specificity was 51%. The positive predictive value was 50%, and the negative predictive value was 87%. The LR+ was 1.77, and the LR− was 0.26.
Receiver operating characteristic curve with a cutoff point of 19, balancing sensitivity and specificity of the Dynamic Gait Index, for probability of identifying fallers.
The convergent validity overall was good, with moderate to strong correlation coefficients (ρ=−.582 to −.778), the lowest correlation coefficient found was for the MSIS physical subscale and the highest was for the 25-Foot Timed Walk Test (Tab. 3). Discriminant validity was shown with a low and nonsignificant correlation coefficient for the MSIS psychological subscale (ρ=−.022).
Spearman Correlation Coefficients Between the Dynamic Gait Index and the Other Measures, and Descriptive Statisticsa
Discussion
This study supports the construct validity of the DGI as a measure of dynamic balance during walking. Significantly lower score on the DGI was found for people reporting falls. With the use of a high cutoff value of ≤19, the probability to identify fallers was 87%. The negative predictive value was high, providing information that participants with scores of ≥20 can be classified as having a low risk of falls. Convergent validity was considered good, with a moderate to strong relationship between the DGI and the timed walking measures (25-Foot Timed Walk Test and TUG) and for tests of lower extremity function (FSST and Timed Sit-to-Stand Test). Discriminant validity also was established through the low correlation coefficient with the MSIS psychological subscale.
Similar to the findings of Cattaneo et al,19 our findings showed that individuals classified as fallers achieved a significantly lower score on the DGI compared with nonfallers; however, mean scores were higher in our study (18.6 for nonfallers, 15.3 for fallers) compared with 16.9 and 13.3, respectively, in the study by Cattaneo et al. To further investigate known-groups validity, a logistic regression was performed. The sensitivity of the scale was high when the cutoff score of ≤19 was used, thus being able to identify fallers. Cattaneo et al19 reported a lower sensitivity of 45% with a cutoff score of >12. In a study on community-dwelling older people, Shumway-Cook et al22 reported a sensitivity of 59% and a specificity of 64% with a cutoff score of ≤19. Measures that can discriminate individuals who are prone to falls are important in the rehabilitation of MS. If an individual can be identified as at risk of falls, interventions may be targeted to decrease the risk. In this analysis, we chose to maximize sensitivity to find a cutoff point that identified as many as possible with risk for falls. However, with a cutoff point of ≤19, the negative predictive value was high. A limitation in our study is the reliance on retrospectively collected data on falls. Prospective studies are needed on the validity of the DGI in predicting falls in people with MS.
Good convergent validity of the DGI was earlier reported through strong correlations with measures of physical mobility, the Berg Balance Scale, the TUG, and a timed short walk.18,19 In the present study, similar strong correlations were found for the TUG and the 25-Foot Timed Walk Test, supporting that these measures of physical mobility reflect the same phenomenon as does the DGI. Strong correlations also were found for the clinical measures (Timed Sit-to-Stand Test and FSST) (Tab. 3). The MS Walking Scale measures aspects of functioning beyond walking, such as running and stair-climbing, which also is reflected in items of the DGI. These findings together establish that the DGI is a valid measure of dynamic balance during walking. Our expectation that the correlation would be weak between the DGI and the psychological subscale of the MSIS was fulfilled. Cattaneo et al19 earlier reported weak correlations between the DGI and the Dizziness Handicap Inventory. These 2 studies support that the DGI does not assess psychological characteristics.
Performing the DGI took 5 to 10 minutes. All items apart from climbing stairs were performed in a hallway, which suggests that the DGI is feasible for use in clinical practice. The horizontal and vertical head-turns were the most difficult items to perform, similar to that in earlier findings.12,19 Several participants reported dizziness when performing these tasks. The item involving stepping over obstacles also presented difficulties for several participants. These tasks are commonly performed when walking around in the community and need to be evaluated in planning rehabilitation efforts. Two of the items in the DGI (“gait with pivot turn” and “step around obstacles”) showed a ceiling effect, with more than two thirds of the participants achieving the maximum score. In a Finnish study including people with neurological disorders, ceiling effects were found for the same 2 items.11 In addition, the overall distribution of scores was skewed, indicating that the DGI could be too easy to perform for ambulatory individuals. Wrisley et al38 presented a modification of the DGI aimed at reducing the ceiling effects seen in patients with vestibular disorders: the Functional Gait Assessment (FGA). The FGA includes all items of the DGI except the item “step around obstacles” and includes 3 additional items covering walking with eyes closed, in tandem, and backward. The DGI could be an appropriate tool for assessing walking in a clinical setting for people with more evident balance impairments. For people with MS with mild balance impairment, the FGA may be more suitable.
The study population included people with MS interested in participating in a study investigating balance exercise, which may affect the generalization of the results. Only those who still had the ability to walk at least 100 m were invited to participate; thus, the results cannot be generalized to people who are more severely disabled (EDSS >6). The participants lived in both rural and city settings, and the proportion of women was in line with the MS population in general, thus representing a general community-dwelling population in Sweden.
In conclusion, this study supports the good construct validity of the DGI as a measure of dynamic balance during walking. Falling is a major health problem among people with MS. With the use of a cutoff point of ≤19 on the DGI, sensitivity was high in identifying people at risk of falls, and appropriate interventions can be applied.
Footnotes
Dr Forsberg and Dr Nilsagård provided concept/idea/research design, project management, and fund procurement. All authors provided writing, data collection, study participants, facilities/equipment, and consultation (including review of manuscript before submission). Dr Forsberg and Ms Andreasson provided data analysis. Dr Forsberg provided institutional liaisons and clerical support.
Thanks are due to the physical therapists who assisted with data collection: Steven Allen, Brickegården Primary Health Care Centre; Anna Lövgren and Kerstin Eriksson, Nora Primary Health Care Centre; Anna Carling, Department of Physiotherapy, Örebro University Hospital; Helena Vesterlin, NeuroRehab Unit, Mälar Hospital in Eskilstuna; Ingrid Lundström and Ingmarie Westlund, Rehab Unit, Västmanland Hospital in Västerås; and Lena Sanner and Malin Nilsson, Rehab Unit, Central Hospital in Karlstad.
The study was approved by the Regional Ethics Committee in Uppsala-Örebro (ID 2010/263).
Part of the data were presented as a poster at the First International Symposium on Gait and Balance in Multiple Sclerosis; October 1, 2011; Portland, Oregon.
This work was funded by the Uppsala-Örebro Regional Research Council, the Research Committee of Örebro County Council, and the Norrbacka-Eugenia Foundation.
- Received July 10, 2012.
- Accepted April 25, 2013.
- © 2013 American Physical Therapy Association