| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Research Reports |
DJ Russell, MSc, is Assistant Professor, School of Rehabilitation Science, Faculty of Health Sciences, McMaster University, and Research Coordinator, CanChild, Centre for Childhood Disability Research, Room 408, Institute for Applied Health Sciences, McMaster University, 1400 Main St W, Hamilton, Ontario, Canada L8S 1C7 (russelld{at}fhs.mcmaster.ca). Address all correspondence to Ms Russell
LM Avery, BEng, is Research Assistant, CanChild, Centre for Childhood Disability Research, McMaster University
PL Rosenbaum, MD (FRCP), is Professor, Department of Pediatrics, McMaster University, Co-Director, CanChild, Centre for Childhood Disability Research, McMaster University, and Chief of Medical Staff, Bloorview MacMillan Centre, Toronto, Ontario, Canada
PS Raina, PhD, is Assistant Professor, Department of Health Care and Epidemiology, Centre for Community Health and Health Evaluation Research, BC Research Institute for Children's and Women's Health, University of British Columbia, Vancouver, British Columbia, Canada
SD Walter, PhD, is Professor, Clinical Epidemiology and Biostatistics, McMaster University, and Investigator, CanChild, Centre for Childhood Disability Research, McMaster University
RJ Palisano, PT, ScD, is Professor, Department of Rehabilitation Sciences, MCP Hahnemann University, Philadelphia, Pa, and Co-investigator, CanChild, Centre for Childhood Disability Research, McMaster University
Submitted June 22, 1999;
Accepted May 10, 2000
 | Abstract |
|---|
5 years of age) and whose CP is more severe. Data from an additional 19 children with CP who were assessed twice, one week apart, were used to examine test-retest reliability. Results. The overall changes in GMFM-66 scores over 12 months and a time x severity x age interaction supported our hypotheses. Test-retest reliability was high (intraclass correlation coefficient=.99). Conclusion and Discussion. This study demonstrated that the GMFM-66 has good psychometric properties. By providing a hierarchical structure and interval scaling, the GMFM-66 can provide a better understanding of motor development for children with CP than the 88 item GMFM and can improve the scoring and interpretation of data obtained with the GMFM.
Key Words: Cerebral palsy • Gross Motor Function Measure • Motor function • Rasch analysis • Validation
| Introduction |
|---|
 TopAbstract Introduction MethodResultsDiscussionSummary and ConclusionsReferences |
|---|
The 88 items of the GMFM are measured by observation of the child and scored on a 4-point ordinal scale (0=does not initiate, 1=initiates <10% of activity, 2=partially completes 10% to <100% of activity, 3=completes activity). The items are weighted equally and grouped into 5 dimensions: (1) lying and rolling (17 items), (2) sitting (20 items), (3) crawling and kneeling (14 items), (4) standing (13 items), and (5) walking, running, jumping (24 items). By the age of 5 years, children without motor delays can generally accomplish all of the items of the GMFM.1 Scores for each dimension are expressed as a percentage of the maximum score for that dimension. The total score is obtained by averaging the percentage scores across the 5 dimensions. The original intent of the GMFM developers was to have one measure that could be used for children across a spectrum of ability levels in order to make it possible for children with different gross motor abilities to enter clinical trials and be assessed with the same measurement tool.2 In addition, the measure needed to be useful for tracking individual children over time. Although items tend to increase in difficulty within a dimension, their ordering was based on clinical judgment and the literature, and the order had not been substantiated with data-based evidence. Items were grouped into dimensions primarily for ease of administration.
Evidence of construct validity of the measure's capacity to detect change in motor function over time was supported by several analyses of the scores of children who were administered the GMFM twice by the same therapist over a 5- to 7-month interval.1 For children with CP, change scores on the GMFM were correlated with parents' judgments of change (r=.54), the child's treating therapist's judgment of change (r=.65), and ratings of change made by therapists who were familiar with the GMFM but unfamiliar with the children by viewing pairs of videotapes, which were in random order (r=.82). As hypothesized in the original validation study,1 the change in GMFM scores was greatest in children recovering from acute head injury, less in preschool-aged children without motor delays, and least in children with CP. For the children with CP, the amount of change in GMFM scores was related to their age and severity of motor disability. The GMFM scores of children who were young (<3 years of age) and had "mild" CP changed more than the scores of children who were older (>6 years of age) and had "severe" CP. Furthermore, children who were judged by both their parents and therapists not to have changed did not demonstrate a change in GMFM scores, whereas children who were judged by both their parents and therapists to have changed demonstrated an increase in GMFM scores.
Bjornson and colleagues7 replicated the original GMFM validation study with children with spastic diplegia and quadriplegia and provided support for the construct validity of measurements of change in motor function obtained with the GMFM over 12- and 24-month periods. The responsiveness of the GMFM for infants under the age of 24 months with CP and motor delay has also been established.8 Researchers have used the GMFM with children with CP to assess the effectiveness of rhizotomy,9–11 intrathecal baclofen,12,13 physical therapy,14,15 horseback riding,16 therapeutic electrical stimulation,17 orthoses,18,19 strength training,20,21 and muscle tendon surgery,22 as well as to determine the correlation between these measurements and measurements of gait and physical fitness.23–26
As the GMFM has been used in a variety of clinical and research situations, its limitations have become more apparent. Some users of the GMFM have chosen to administer only those dimensions that are most relevant to their clients' current level of functional needs. This selective use of the GMFM allows fewer items to be administered and increases the measure's responsiveness to change by eliminating items that are not relevant to the therapeutic intervention or are unlikely to change as a result of intervention. However, the evidence for the reliability and the validity of the dimension scores is generally not as strong as for the measure as a whole.1
Another limitation has been the interpretation of the GMFM total score. Children with different skills and abilities within and between dimensions can receive the same total score. A further limitation is that scores of children functioning in the middle of the scale have greater potential to change than scores of children whose initial assessment is either very low or very high because more items are in the middle of the scale than at the extremes.
In an effort to improve the interpretability and the clinical usefulness of the GMFM, we applied the Rasch model of item analysis to the GMFM.27 Rasch analysis uses a one-parameter logistic model to derive an equal interval measure from the raw score.28 Rasch analysis originated in the areas of education and psychology, and during the last 5 to 10 years has been used to construct and validate measures used in rehabilitation.29–33 There are several reasons to apply the Rasch scaling to the GMFM. First, the items can be arranged in order of relative difficulty (hierarchical structure). Second, it makes it possible to create interval scales from what are believed to be ordinal scales, because scores take into account how much more difficult one item is to accomplish than the previous item. Third, Rasch analysis allows the elimination of items that do not fit the unidimensional construct (ie, misfitting items). The unidimensionality assumption is met when the items of the test demonstrate that only one ability is being measured (in our case, gross motor ability). Fourth, Rasch analysis allows calculation of a total score when not all items are administered. All of these factors would lead to improved scoring and interpretation of the GMFM.
Although Rasch scaling has potential advantages, potential challenges also needed to be considered. These challenges include the fact that the GMFM is already in widespread use and a revised measure would require modification of the manual and training materials. In addition, because Rasch analysis is used primarily to create unidimensional measures, it could identify items for removal that do not fit the model but are considered clinically relevant. Of major concern was whether the removal of items would affect the GMFM's responsiveness to change and require new validation of the modified scale. A computer program would also be necessary to analyze and interpret scores based on Rasch scaling.
With these considerations in mind, our group applied the Rasch partial credit model to the GMFM.27 The Rasch partial credit model does not make any assumptions about the difficulty of each response option, either within an item or between items. For example, it does not assume that the difficulty of going from a score of 1 to a score of 2 is the same difficulty as going from a score of 2 to a score of 3 or that the difficulty of going from a score of 1 to a score of 2 is the same for different items. Rasch modeling helped us to identify 66 items from the original 88-item GMFM that form a unidimensional hierarchical scale, the GMFM-66. Table 1 lists the original 88 GMFM items and indicates which items we removed for the GMFM-66.
|
| Method |
|---|
|
TopAbstractIntroductionMethodResultsDiscussionSummary and ConclusionsReferences |
|---|
|
|
Validity and Responsiveness
Face validity was assessed by examining the hierarchy of items (Tab. 4) and the GMFM-66 total scores for different groups of children to determine whether they made clinical sense (Tab. 2) based on what is known about CP. We expected that items from the lying and rolling and the sitting dimensions would be easier for children with CP to accomplish than items in standing and the walking, running, and jumping dimensions and would, therefore, have lower difficulty estimates. We also expected that children with hemiplegia would have higher scores, on average, than children with more limbs involved (eg, children with diplegia, triplegia, or quadriplegia) and that mean GMFM-66 scores would vary systematically by GMFCS level,34 with children in level I (mild disability) having the highest GMFM-66 scores.
|
5 years) and severity of motor disability, as assessed using the GMFCS ("mild"=levels I and II, "moderate"=level III, "severe"=levels IV and V).
Data Analysis
Test-retest reliability data were analyzed using an ICC based on the 1-way analysis of variance (ANOVA) model 1,137 for both the new and the original scoring methods. To test the difference in reliability between the 2 scoring methods, 95% confidence limits for the difference in ICCs were computed.38 To determine whether an effect of time existed and whether there were interactions of age and severity, a 3-way repeated measures ANOVA was calculated with time as the within-subject factor and age and severity as between-subject factors.
| Results |
|---|
|
TopAbstractIntroductionMethodResultsDiscussionSummary and ConclusionsReferences |
|---|
Validity
The items of the GMFM-66 in order of difficulty ascertained by Rasch analysis (ie, with an item score of 3) are presented in Table 4. To remove negative values and be more clinically interpretable, the linear estimates of difficulty expressed in logits (log-odds ratios) were transformed to a scale with values ranging from 0 to 100. The GMFM-66 has difficulty estimates ranging from 15.72 for the easiest item (item 21: "sits on mat, supported at thorax by therapist, lifts head upright, maintains 3 seconds") to 88.52 for the most difficult item (item 82: "stands on right foot, hops on right foot 10 times within a 24-inch circle"). The Rasch scale is in logits (log-odds units), which runs from negative infinity to positive infinity. A linear transformation was applied to the ability scale so that the minimum ability was 0 and the maximum ability was 100.28 In terms of the items on the GMFM-66, a score of 0 on the ability scale means that a child has a 100% probability of having a score of 0 on every item. A score of 100 means that a child has a 100% probability of having a score of 3 on every item. The item difficulty scale was transformed using the same linear transformation. This transformation placed the easiest item at a difficulty of 15.7 and the most difficult item at a difficulty of 88.52. An item's difficulty corresponds to the ability required for a score of 3 on that item. Therefore, for a child to be likely (P=.5) to have a score of 3 on the easiest item, he or she needs to have an ability estimate of 15.7.
Table 2 shows the mean GMFM-66 scores or "ability estimates" for the 537 children included in the Rasch analysis by age, sex, type and distribution of CP, and GMFCS level. The mean ability estimates are similar for males and females. Children with hemiplegia have the highest mean ability estimates, and children with 4-limb involvement have the lowest mean ability estimates. Children who were classified as having "mild" (ie, level I) motor impairment using the GMFCS had a mean ability estimate of 78.06, followed by mean ability estimates of 60.92 for children with level II motor impairment, 49.98 for children with level III motor impairment, 37.94 for children with level IV motor impairment, and 20.63 for children with level V motor impairment.
As hypothesized, there was an overall change in mean GMFM-66 scores for the sample of children assessed at baseline and at 12 months, from 52.76 to 54.61 (F=116.3; df=1,222; P<.0001). There was also a time x severity x age interaction (F=12.6; df=2,222; P<.0001). Figure 1 shows the 3-way interaction plotted as mean change scores by severity and age category. The means and standard deviations by group are reported in Table 5. Figure 1 shows that children under 5 years of age changed more than children aged 5 years and over. This change was greater for children whose GMFCS level was I or II than for children with other GMFCS levels. The mean change for children aged 5 years and over was approximately zero regardless of GMFCS level.
|
|
| Discussion |
|---|
|
TopAbstractIntroductionMethodResultsDiscussionSummary and ConclusionsReferences |
|---|
Test-retest reliability of the GMFM-66 scores was examined using data collected on the 88-item GMFM. The magnitude of the ICCs for the GMFM-88 and for the GMFM-66 was high. The ICCs may have been different had only the items of the GMFM-66 been administered; however, there is no reason to expect that the reliability estimates would have decreased.
Although we believe that the results support the construct validity of the GMFM-66 scores, the measure detected less change in the older children regardless of severity of CP. This finding suggests either that the GMFM-66 is not as sensitive to changes in motor function made by children with CP aged 5 years and over as it is to changes in motor function made by children under 5 years of age or that a smaller mean change is happening in children over age 5 years. This pattern of varying responsiveness for children over 5 years of age is similar to current findings from the models of motor growth using the 88-item GMFM.36 The emphasis of the GMFM is on motor abilities associated with gross motor development that typically are achieved by age 5 years in children without motor impairments. For children with CP over the age of 5 years, especially those who are unable to walk without assistive mobility devices, change may be associated with performance of motor functions at home, at school, in the community, and during social participation rather than acquisition of basic gross motor skills. In our view, therefore, measures of disability such as the Pediatric Evaluation Disability Inventory29 may be more appropriate for evaluating change in older children with CP.
The improvements of the GMFM-66 over the GMFM include (1) the ordering of items according to difficulty, (2) the interval properties of the scale allowing for improved interpretability of a total score and of change scores, and (3) the decrease in administration time. The difficulty estimates for each item on the GMFM-66 (Tab. 4) provide information that is unique to gross motor function of children with CP. Items for lying, rolling, and head control in supported sitting are the least difficult, whereas items for standing on one foot, jumping, and hopping are the most difficult. The difficulty estimates for items of sitting, crawling, kneeling, standing, and walking tend to overlap in the middle of the scale, indicating that motor abilities in these different areas may be developing simultaneously for children with CP. These findings have implications for determining outcomes and planning intervention.
Interpretation of a child's GMFM-66 score should be enhanced by using the item difficulty map (Fig. 2). The item map provides a graphic presentation of items ordered along a continuum of difficulty. As each step represents a single response option (eg, a score of 1, 2, or 3 on a GMFM item), there are 198 steps (66 items x 3 response options). Because items are now ordered in terms of their difficulty, the GMFM-66 clarifies the next task that a child is likely to accomplish and how difficult it might be to accomplish this task. For example, if a child has a GMFM-66 score of 80 and we look at the item map (Fig. 2), we would be fairly confident that he or she could accomplish all the items to the left of 80 and indicated by a black box (as completing the response option of 3). This includes all items in the lying and rolling, sitting, and crawling and kneeling categories and most items in the standing category (with the exception of items 57 and 58). The items to the right of a GMFM-66 score of 80 are those that the child is unlikely to have accomplished fully. The child would likely be able to accomplish items 74 and 87 (if not, the child would likely have a score of 2 on these 2 items, and these would be the next items that the child will likely complete). Item maps have been used with the Pediatric Evaluation of Disability Inventory29 and the School Function Assessment39 to facilitate clinical interpretation of summary scores. A further discussion of how item maps are useful for clinical interpretation of assessment scores can be found in the article by Coster et al.40 Transferring the GMFM-66 score onto the item map can help in the interpretation of assessment scores by presenting both the child's gross motor abilities and the difficulty of items that have not been achieved. The GMAE program35 provides a measure of error that is useful in assessing whether an individual's change score is significant.
|
Profiles of 2 children with CP illustrate differences between the original method of scoring the GMFM and the Rasched interval scale for scoring the GMFM-66 (Fig. 3). John is a 3-year 8-month-old boy with a diagnosis of CP with dystonic movements and primary involvement of the legs. He is classified at level I on the GMFCS. When initially tested, John achieved a GMFM score of 59 and a GMFM-66 score of 54. Upon retesting 6 months later, John achieved a GMFM score of 77 and a GMFM-66 score of 65.
|
Although John demonstrated a change score of 12% and Sara demonstrated a change score of 1% on the GMFM, they had similar change scores on the GMFM-66. Sara's larger change score on the GMFM-66 is attributable to the large spacing between certain item difficulties (Fig. 2). Although Sara did not improve on as many items as John, the items she accomplished were more spread out along the difficulty continuum.
The GMFM-66 should take less time to administer with 22 fewer items and the ability of the GMAE program35 to estimate a child's ability score even when all items have not been administered. A potential limitation of the GMFM-66 is the need for a computer program to score it. A computer program has been designed to convert the raw scores from either the GMFM-88 or the GMFM-66 into difficulty estimates.35 It will calculate the GMAE score based on the 66 items, with a standard error. There are no longer separate scores for each dimension. One major difference in the administration and scoring of the GMFM-66 than the scoring of the 88-item GMFM is the importance of differentiating a true score of 0 (child is unable to perform) from an item that was not tested. In the original administration and scoring guidelines, a score of 0 was assigned for items that were not administered or that the child did not perform during the assessment. The new GMFM score form has been modified to include the response "not tested." Although not all 66 items have to be administered to calculate a GMAE score, the more items that are administered, the more accurate the estimate of a child's gross motor ability.27
We believe that the GMFM-66 has applications for evaluating the effectiveness of interventions. Three studies,9–11 for example, have been published recently with conflicting results on the effectiveness of dorsal rhizotomy surgery for children with CP. All 3 studies used common outcome measures, including the GMFM. From discussions among the 3 groups of investigators, several hypotheses arose as to why there may have been differences in the study outcomes, including the possibility that the relative difficulty of items at different intervals of the GMFM is not reflected in the total score.9 It will be important to determine whether the results of these studies would be different using the GMFM-66, in which the change in score is interval and reflects the difficulty of items, regardless of a child's initial motor ability.
| Summary and Conclusions |
|---|
|
TopAbstractIntroductionMethodResultsDiscussionSummary and ConclusionsReferences |
|---|
| Footnotes |
|---|
The study was approved by the McMaster University Research Ethics Board.
This research was supported by grant R01 HD34947 from the National Institute for Child Health and Human Development, National Institutes of Health.
| References |
|---|
|
TopAbstractIntroductionMethodResultsDiscussionSummary and ConclusionsReferences |
|---|
CiteULike 
Complore 
Connotea 
Del.icio.us 
Digg 
Reddit 
Technorati What's this?
This article has been cited by other articles:
|
|
 |
|
  A Meyer-Heim, C Ammann-Reiffer, A Schmartz, J Schafer, F H Sennhauser, F Heinen, B Knecht, E Dabrowski, and I Borggraefe Improvement of walking abilities after robotic-assisted locomotion training in children with cerebral palsy Arch. Dis. Child., August 1, 2009; 94(8): 615 - 620. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 A. Lundkvist Josenby, G.-B. Jarnlo, C. Gummesson, and E. Nordmark Longitudinal Construct Validity of the GMFM-88 Total Score and Goal Total Score and the GMFM-66 Score in a 5-Year Follow-up Study Physical Therapy, April 1, 2009; 89(4): 342 - 350. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. R Pierce, M. F Barbe, A. E Barr, P. A Shewokis, and R. T Lauer Roles of Reflex Activity and Co-contraction During Assessments of Spasticity of the Knee Flexor and Knee Extensor Muscles in Children With Cerebral Palsy and Different Functional Levels Physical Therapy, October 1, 2008; 88(10): 1124 - 1134. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. E Hanna, D. J Bartlett, L. M Rivard, and D. J Russell Reference Curves for the Gross Motor Function Measure: Percentiles for Clinical Description and Tracking Over Time Among Children With Cerebral Palsy Physical Therapy, May 1, 2008; 88(5): 596 - 607. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 M. Linder-Lucht, V. Othmer, M. Walther, J. Vry, U. Michaelis, S. Stein, H. Weissenmayer, R. Korinthenberg, V. Mall, and and the Gross Motor Function Measure-Traumatic Bra Validation of the Gross Motor Function Measure for Use in Children and Adolescents With Traumatic Brain Injuries Pediatrics, October 1, 2007; 120(4): e880 - e886. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 T. D. Sanger, A. Bastian, J. Brunstrom, D. Damiano, M. Delgado, L. Dure, D. Gaebler-Spira, A. Hoon, J. W. Mink, S. Sherman-Levine, et al. Prospective Open-Label Clinical Trial of Trihexyphenidyl in Children With Secondary Dystonia due to Cerebral Palsy J Child Neurol, May 1, 2007; 22(5): 530 - 537. [Abstract] [PDF]
|
|
|
|
 |
|
 J. M Voorman, A. J Dallmeijer, C. Schuengel, D. L Knol, G. J Lankhorst, and J. G Becher Activities and participation of 9- to 13-year-old children with cerebral palsy Clinical Rehabilitation, November 1, 2006; 20(11): 937 - 948. [Abstract] [PDF]
|
|
|
|
|
|
H.-H. Wang, H.-F. Liao, and C.-L. Hsieh Reliability, Sensitivity to Change, and Responsiveness of the Peabody Developmental Motor Scales-Second Edition for Children With Cerebral Palsy Physical Therapy, October 1, 2006; 86(10): 1351 - 1359. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
Y.-P. Chiu, S. L Fritz, K. E Light, and C. A Velozo Use of Item Response Analysis to Investigate Measurement Properties and Clinical Validity of Data for the Dynamic Gait Index Physical Therapy, June 1, 2006; 86(6): 778 - 787. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
M. A Fragala-Pinkham, S. M Haley, J. Rabin, and V. S Kharasch A Fitness Program for Children With Disabilities Physical Therapy, November 1, 2005; 85(11): 1182 - 1200. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
B. R. Vohr, M. E. Msall, D. Wilson, L. L. Wright, S. McDonald, and W. K. Poole Spectrum of Gross Motor Function in Extremely Low Birth Weight Children With Cerebral Palsy at 18 Months of Age Pediatrics, July 1, 2005; 116(1): 123 - 129. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
B. L Tieman, R. J Palisano, E. J Gracely, and P. L Rosenbaum Gross Motor Capability and Performance of Mobility in Children With Cerebral Palsy: A Comparison Across Home, School, and Outdoors/Community Settings Physical Therapy, May 1, 2004; 84(5): 419 - 429. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 S. T. Iannaccone and L. S. Hynan Reliability of 4 Outcome Measures in Pediatric Spinal Muscular Atrophy Arch Neurol, August 1, 2003; 60(8): 1130 - 1136. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 P. L. Rosenbaum, S. D. Walter, S. E. Hanna, R. J. Palisano, D. J. Russell, P. Raina, E. Wood, D. J. Bartlett, and B. E. Galuppi Prognosis for Gross Motor Function in Cerebral Palsy: Creation of Motor Development Curves JAMA, September 18, 2002; 288(11): 1357 - 1363. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
J. W. Gorter, P. L Rosenbaum, and M. Ketelaar Reanalyzing the Data Physical Therapy, August 1, 2002; 82(8): 828 - 830. [Full Text]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
