The Gross Motor Performance Measure: validity and responsiveness of a measure of quality of movement.

Key Words: Cerebral palsy, Gross motor behavior, Measurement, Movement, Validity.

Research in the field of cerebral palsy (CP) has been hampered by a lack of reliable, valid, and responsive measures of physical function. In particular, before studies of treatment efficacy or effectiveness can be implemented, there is a need for good measures of gross motor behavior.[1]

Gross motor behavior in children with CP has been conceptualized as having two main features: function and performance.[2] We use the term "gross motor function" to describe the accomplishment of motor activities, or how much the child does, for example, standing independently for 10 seconds. In this context, "function" does not necessarily refer to activities that are purposeful to the child or performed in everyday settings. Instead, functional activities are defined as traditional gross motor milestone behaviors that can be tested in a standardized manner.

The term "gross motor performance" describes the quality of motor activities, or how well the child does the activity, for example, the degree of stability when standing. The importance of motor performance to overall motor behavior in children is reflected in the Work of Bobath,[3] Rood,[4] and Bly.[5] Numerous authors[6-8] have described the need to develop instruments to measure this feature.

Since 1984, researchers at McMaster University, Queen's University, and the Hugh MacMillan Rehabilitation Centre in Ontario, Canada, have conducted a series of studies to develop gross motor measures for use with children with CP. The purpose of these studies has been to design and validate instruments to measure change in motor function and performance over time.[9]

The Gross Motor Function Measure (GMFM) has demonstrated high levels of validity, reliability, and responsiveness m the assessment of motor function in children with CP.[10] The GMFM consists of 88 items organized into five dimensions: lying and rolling; sitting; crawling and kneeling; standing; and walking, running, and jumping. Although these items do not assess children within different environmental contexts, nor do they represent activities chosen by children themselves, they were chosen by therapists as important for developmental progress and amenable to change.[10,11] During a previous study, we determined that an additional instrument was required to measure performance, or the qualitative aspects of motor behavior that are characteristic of children with CP but that are not measured adequately by other motor measures.[12]

Initially, we developed the Gross Motor Performance Measure (GMPM) after reviewing the literature relative to attributes of gross motor performance. Definitions were written for 33 attributes.[13] Through nominal group process meetings, 30 developmental therapists from five children's treatment centres southern Ontario helped to reduce the number of attributes to 5 and to develop attribute scales.[13] An international panel of experts working in academic and clinical settings provided judgments on content validity of the GMPM using a Delphi consensus method. Results indicated that the GMPM had satisfactory completeness, clarity, and potential for evaluating changes in motor performance.[14]

The GMPM is a criterion-referenced observational instrument that can be administered with a minimum of equipment in less than 1 hour, depending on assessor skill, developmental stage, and cooperation of the child. The measure consists of 20 items derived from the GMFM, each which is matched with three designated attributes of performance. Possible attributes to be assessed include alignment, stability, coordination, weight shift, and dissociation. Children are assessed only on items in which they can achieve at least a partial GMFM score, meaning that they can initiate an activity, thereby allowing an assessment of motor quality. Thus, some children have motor performance assessed on all 20 items, whereas others are assessed on as few as 2 or 3 items. Attributes are scored on individualized five-point scales varying from 1 ("severely abnormal") to 5 ("consistently normal"). (See Appendixes 1 and 2 for attribute definitions, sample GMPM item, and scale.) Raw scores are summarized into a total score and subscale attribute scores. These scores are expressed as percentages of the maximum possible score relative to the number of items performed.

Validation Hypotheses

Because there is not an accepted criterion evaluative measure of overall motor performance for children with CP,[12] direct tests of criterion validity could not be conducted. Hypotheses were difficult to develop because there is minimal motor control theory applicable to this heterogeneous population.[15] In addition, there are no longitudinal studies of the natural history of motor performance in children with CP. Thus, we developed hypotheses from the knowledge base of investigators and therapists experienced in working with this population. In addition, experience with the GMFM development gave us information and ideas we could use to formulate hypotheses about potential change in motor performance.

We believe that construct validity of the GMPM could be tested by formulating a series of a priori hypotheses about how change scores on the GMPM would be associated with different study groups and with changes in function. We expected that differences in performance of motor activities in different groups of children could be assessed by comparing scores among groups of children with CP, children with other neurological problems, and children without neurological problems. We expected that the GMPM change scores would be greatest in a group of children with acute neurological injuries, such as head injury, as these children often recover rapidly from their injuries and frequently have marked improvements in performance attributes such as coordination and stability.

Change scores in children without disabilities were expected to be minimal because these children should already have normal quality of movement, although there may be slight changes in performance as children learn new motor behaviors such as walking.

Developmental changes in motor skills and performance generally slow as children mature; therefore, younger children with CP were expected to show more change in GMPM scores than older children. Change scores in children with CP were expected to be greater in those with milder impairments or disabilities and less in those with more severe problems. Classification of impairment severity into mild, moderate, and severe categories is a common, yet relatively untested, clinical and research practice. Recent evidence from plots of GMFM scores against clinician's assessments of "severity" in children with CP, however, shows distinct patterns of motor function in these different groups (Rosenbaum PL, Russell DJ; unpublished research). Although reliability data are not available, the face validity of this categorization of severity for motor function may extend to the assessment of gross motor performance, which is also a strong characteristic of movement of children with CP.

We anticipated that changes in the GMPM were likely to be positively correlated with changes in function as measured by the GMFM because improvement in a child's quality of movement, or performance, could contribute to change in the child's gross motor skills. We also recognized that the relationship between change in motor function and change in performance would not necessarily be strong.

Concurrent validity of the GMPM was tested through comparison with independent parent and therapist ratings of motor performance. In a previous study,[10] measures of parent and therapist ratings of children's motor function were demonstrated to be reliable in this population. We expected therapist ratings to correlate more highly with the GMPM than parent ratings because therapists were familiar with the concepts of performance used in the GMPM, whereas parents were not. Parents' judgments of changes were expected to be based on their evaluation of what the child does in daily life and thus more focused on function than performance.

Five hypotheses were formulated to test the validity of the GMPM:

1. Hypothesis 1 - Children with head injuries will have greater changes in GMPM scores over time than children in other groups. Nondisabled children will show smaller changes than children in other groups.

2. Hypothesis 2 - Children with CP classified by therapists as "mild" or "moderate" will have greater changes in GMPM scores over time than those with CP classified as "severe."

3. Hypothesis 3 - Infants with CP will show more changes in GMPM scores over time than toddlers and older children.

4. Hypothesis 4 - GMPM change scores will have a low to moderate correlation (r[greater than or equal to].30) with GMFM change scores.

5. Hypothesis 5 - Correlations between change scores on the GMPM instrument and therapist ratings of change will be moderate (r[greater than or equal to].55). Correlations of GMPM change scores with parent ratings of change will be slightly lower (r[greater than or equal to] .45).

Responsiveness

An evaluative instrument is responsive if there is a high likelihood of detecting a treatment effect even if that effect is small.[16] A responsive measure will allow one to be confident of detecting a change, through inspection of the change score, should a change actually occur. An appropriate measure of responsiveness can be obtained by relating the variability in test scores of stable individuals to the difference in scores of individuals who are changing.[16]

The objective of this validation study was to examine the measurement properties of the GMPM. Validity (construct, concurrent), reliability (interrater, intrarater, test-retest), and responsiveness of the instrument were investigated with the assumption the primary use of the GMPM would be to evaluate change in the performance of common gross motor activities over time. This assumption has implications for the criteria used to judge the measure.[17] The purpose of this article is to report the validity and responsiveness findings of that study. Results from reliability analyses are reported elsewhere (see article by Gowland and colleagues in this issue of Physical Therapy).

Method

On the basis of available information regarding variability in motor performance scores and possible effect sizes, we estimated that we would need a sample size of 180 children to appropriately examine our hypotheses. We initially planned to recruit 120 children with CP, 30 children who had sustained an acute head injury (HI), and 30 children without neurological problems from three children's treatment centers. The latter two groups were included to provide comparative samples in which changes in gross motor performance were likely to be either dramatic (HI) or minimal (no disability). We considered the sample size adequate for analysis because one-tailed 95% confidence intervals would vary from .74 to .84 if validation correlations of r > .80 were obtained for an overall effect. We contend that this correlation level is acceptable when using intraclass correlation coefficients (ICCs) and is based on our experience in previous validation studies of the GMFM.[7]

The final enrollment was 106 in the group of children with CP, 18 in the group of children with HI, and 29 m the group of children without neurological problems, for a total of 153 children. This final sample size was limited by the willingness of children and their families to participate in a lengthy study requiring multiple assessments. The children with CP and HI were receiving various forms and intensities of therapy. Because we wanted to study the ability of a measure to detect changes in motor performance, and not to attribute any changes in performance to therapeutic efficacy, we did not consider it important to classify children according to treatment status.

Children with CP were selected to provide a range of age and severity of neurological involvement. Children with either CP or HI were categorized into mild, moderate, and severe groups by the referring therapist. This categorization was based on criteria for severity that were in common use at the participating study centers. Children in the HI group were all in the early (<3 months postinjury) stages of recovery (Tab. 1).

Table 1. Validation Study Sample

                         Severity of Impairment
Diagnostic       Age
Group            (y)    Mild    Moderate   Severe   Total

Cerebral palsy   0-2      8       15          7        30
                 3-4     10       15          9        34
                 5-6      2       13          3        18
                 7-12     5       11          8        24
                 Total   25       54         27       106
Head injury      0-19     6        7          5        18
Nondisabled      0-2                                   16
                 3-6                                   13
                 Total                                 29

Physical therapists who worked regularly with children with neurological impairments participated in the study. Proficiency in administration of the GMFM was a requirement for participation because the GMPM and the GMFM are linked. Thirty therapists were initially trained in administration and scoring of the GMPM and GMFM in 2-day training sessions. Training was facilitated by using a series of videotapes of children with CP, including various ages, types, and levels of function, as well as nondisabled children. Therapists were trained by the investigators in the use of the measures to a criterion Kappa agreement level of at least [kappa]=0.6, indicating moderate agreement with the investigators' consensus-based ratings of the videotape,[18] before commencing to enroll children in the study. This training methodology was described in an earlier report.[19] Therapists also practiced administering the measure on non-study children to improve their familiarity with the instrument.

After receiving signed consent from parents or guardians, the children were assessed at the same time with the GMPM and GMFM on two occasions 4 to 6 months apart. The tests were administered on body occasions according to the protocols outlined in the GMFM and GMPM administration guidelines. Standardized parent rating scales and therapist rating scales were independently completed prior to each assessment to give separate parent and therapist judgments of each child's gross motor performance. For these measures, a Likert scale is used to judge the child's current motor performance and change in motor performance over time for the concurrent validity analysis. A similar set of questionnaires was used in the GMFM validation study and were shown to reliably assess change in gross motor function by parents (n=23, ICC=.92) and by therapists (n=23, ICC=.96).[11]

To assess responsiveness, parents and therapists were asked to rate the child's overall change in motor performance before the second administration of the test. A separate 15-point scale was used to rate the child's motor performance, from -7 (large negative change), through 0 (no change), to +7 (large positive change). We classified children as belonging to a "responsive" group if they were judged by either therapists or parents as having improved in performance (ie, a score between +2 and +7). A "stable" group was considered to be those children who were identified by either the parents or therapists as "not having changed" or as "changing only a very small amount" (ie, - 1, 0, + 1). A responsive measure should show little variability due to time in a "stable" group and greater variability due to time in a "responsive" group. Further details of the study design, including reliability assessment methods, are reported elsewhere.[19]

Data Analysis

Summary scores were calculated to create a GMPM total score and five individual attribute scores. A high score indicates a comparatively normal motor performance. To calculate an alignment score, for example, raw scores for the alignment attribute are summarized across all motor items in which alignment appears. This score is expressed as a percentage of the maximum possible score relative to the number of times the child was assessed on alignment. This method ensures that the level of motor function, or ability to do motor activities, does not affect the scoring of the performance of those activities. Thus, young children without disabilities may score low on function but high on performance. Older children with severe disabilities may score low on function and performance. Only total score data will be reported here.

Because 3 children were lost to follow-up and 7 children's data were incomplete, the data from 143 children were available. This sample composed 99 children with CP, 17 children with HI, and 27 children with no disabilities. Data transformation was not required, and no adjustment was made for multiple testing.

Hypotheses 1 (diagnostic differences), 2 (seventy differences) and 3 (age differences) were examined using t tests and analyses of variance (ANOVAs) to detect differential changes in GMPM scores over time. Hypotheses 4 (GMPM-GMFM correlations) and 5 (therapist/parent-GMPM correlations) were examined using a correlation analysis. In the responsiveness analysis, a repeated-measures ANOVA on time 1 and time 2 scores for both the "stable" group and the "responsive" group was done. The F statistics from these two ANOVAs were used to calculate an ICC for each group.[18] A Fisher Z transformation was used to compare the two correlation coefficients to determine whether they were different. The level of statistical significance was set at [P.sub.2][greater than or equal to].05 (two-tailed probability) for t-test and ANOVA analyses and at [P.sub.1[greater than or equal to].05 (one-tailed probability) for correlation analyses.

Results

Failure to enroll the full planned sample size had some effect on the statistical power available for certain subgroup analyses. Power, however, is also affected by the size of the effect and by acceptable levels of type I and II errors. Small effect sizes, such as those that may occur in motor performance, are more difficult to detect than larger differences. Similarly, in a heterogeneous population such as the children with CP in this study, the likelihood of detecting differences is decreased. On the basis of available information, we were able to estimate an appropriate sample size at the outset of the study. It was not possible, however, to predict the variability and relatively small effect sizes of motor performance in the children with CP. For this reason, and because we examined fewer children than we had planned, the power to detect differences in some analyses was less than expected. This methodological problem is common in preliminary studies of phenomena. Thus, we are cautious in the interpretation of these data and focus on trends that may direct future research.

Hypothesis 1

This hypothesis was strongly supported. Analysis of variance indicated that total scores for each diagnostic group differed from each other at time 1 and time 2 (Tab. 2). High scores for the nondisabled children reflected their normal movement quality. Lower scores for the children with HI reflected impairment of many attributes. Children with CP had the lowest scores, with impairment of movement quality in all attributes.

[TABULAR DATA OMITTED]

Changes in GMPM total scores (time 2-time 1) supported the hypothesis that the HI group would show more change than the CP group and that the nondisabled group would show less change. Paired t tests revealed that all groups changed between assessments, although the greater change in the HI group (7.3 units) reflects the rapid change in movement quality that may be seen in children recovering from head trauma. The CP group also showed change (2.3 units) but of smaller magnitude. This small degree of change illustrates the expected development in motor performance over a short period of time in a chronic condition such as CP. The nondisabled group was not expected to change dramatically in their motor performance (0.7 units), although their motor function was expected to increase.

Hypothesis 2

This hypothesis was supported by results that revealed that in the CP group total scores differed depending on the severity of condition (Tab. 3). Time 1 and time 2 scores were highest for the mild impairment group and lowest for the severe impairment group, This finding indicates that the classification of CP by severity may be based at least partly on movement quality. Analysis of variance indicated that change in total scores was not found to be dependent on severity (F=1.41; df=2,98; [P.sub.2]=.2491). We interpret this finding with caution because of the low statistical power (0.24) available to detect differences. Change in scores for both mild impairment (2.8 units) and moderate impairment (2.9 units) groups was small but statistically significant. The group with severe impairment showed no change (0.5 units) between assessments.

[TABULAR DATA OMITTED]

Hypothesis 3

The initial hypothesis, that changes in GMPM scores would be greatest for infants and less for older children with CP, was not supported by the data, yet this interpretation must be tempered by a general lack of power in the analyses. Analysis of variance indicated that total scores did not differ across age groups at either the first or second assessment (Tab. 4). Similarly, the change in total scores did not differ across age groups (F=0.23; df=3,98; [P.sub.2]=.8753). Paired t tests within each age group, however, revealed that scores for children older than 7 years did change (t[21]=3.09, [P.sub.2]=.0055). Power in these analyses varied from 0.07 to 0.21.

[TABULAR DATA OMITTED]

This initial finding was contrary to our expectation that the greatest improvement in motor performance would occur in younger children with CP. An ANOVA was conducted to determine whether there was an interaction of age and seventy of impairment. Mean change scores were calculated for each age severity subgroup (Tab. 5). Within seventy groups there was no difference in mean change scores for the various age groups. This finding is consistent with the previous results. Within age groups, however, the 3- to 4-year-old children showed differences between the mild, moderate, and severe impairment groups (F=4.47; df=2,32; [P.sub.2]=.0194). The positive change in the 3- to 4-year-old moderate impairment group (4.2 units) (t[14]=3.1, [P.sub.2]=.0085) was apparently balanced out in the earlier age analysis by the negative change (-2.5 units) in the 3- to 4-year-old severe impairment group. Similarly, the positive change in the >7-year-old moderate impairment group (4.1 units) (t[10]=2.9, [P.sub.2]=.0167) was sufficient to influence the results of the entire group of older children. Thus, the greatest change in motor performance occurred in children with moderate CP at the ages of 3 to 4 years and greater than 7 years. A comparatively large change in the 5-to 6-year-old mild impairment group (4.2 units) did not reach statistical significance. The small number of children (n=2) in this group likely affected the power (0.20) of the analysis to detect change.

[TABULAR DATA OMITTED]

The two-way ANOVA of age X severity indicated that the amount of change in each age group did not depend on the severity of CP (F=0.86; df=5,98; [P.sub.2]=.5109). These preceding analyses, however, may be questioned because of the small sample size in the subgroups and the small effect sizes, which resulted in a lack of statistical power (0.07-0.20).

Hypothesis 4

The hypothesis that change scores between the GMFM and GMPM would be moderately correlated was partially supported (Tab. 6). At the first assessment, GMFM total scores correlated with GMPM total scores for the HI group at .18, the CP group at .72, and the nondisabled group at .64. The low correlation for the HI group was due to consistently high GMFM scores and variable GMPM scores. Different severity groups of the CP sample yielded low, but usually significant, correlations from .25 to .45. Different age groups also yielded significant correlations from .56 to .89.

Table 6. Correlations Between Gross Motor Function Measure and Gross
  Motor
Performance Measure Total Scores for Various Study Groups(a)

                         n    Time 1    Time 2   Change

Group
  Nondisabled            27   .64(b)   .81(b)    .23
  Head injury            17   .18      .11       .18
  Cerebral palsy         99   .72(b)   .80(b)    .17
Severity of impairment
  Mild                   23   .25      .41(b)    .28(b)
  Moderate               53   .45(b)   .60(b)    .12
  Severe                 23   .39(b)   .61(b)   -.04
Age (y)
  0-2                    26   .56(b)   .64(b)    .36(b)
  3-4                    33   .82(b)   .91(b)    .29(b)
  5-6                    18   .85(b)   .85(b)   -.18
  >7                     22   .89(b)   .87(b)    .07

(a) Hypothesis r[greater than or equal to].30.
(b) One-tailed probability, [P.sub.1][less than or equal to]-.05.

At the second assessment, similar correlation patterns were observed, with higher correlations in the nondisabled group (r=.81) and the CP group (r=.80). The GMPM change scores, however, were not highly correlated with GMFM change scores and ranged from .17 to .23. Differential rates of change in function and performance could account for these low correlations. The low correlations may also be consistent with problems of low reliability of change scores, which compound any measurement errors at both assessment points.[21]

These data generally support our contention that motor performance is related to, yet distinct from, motor function. It appears that the GMFM and GMPM are measuring different constructs of motor behavior.

Hypothesis 5

The hypothesis that the correlations between the GMPM change scores and the therapist ratings of change would be moderate is not supported by the data (Tab. 7). Correlations between GMPM change scores and judgments of parents (r= -.06) and therapists (r=.22) were low. To determine whether judgments were being made on the basis of changes in gross motor function as we have defined it rather than performance, correlations were also calculated for GMFM change scores. These correlations were moderate and significant for parent ratings (r=.44) but not for therapist ratings (r=.40). These results indicate that structured assessments of gross motor performance as measured by the GMPM were not strongly linked with clinicians' perceptions of changes in quality of movement.

Table 7. Correlations Between
Change on Gross Motor Performance
Measure (GMPM) and Gross Motor Function
Measures (GMFM) and Change in
Quality Judged by Parents and
Therapists.

                  Parent       Therapist
                  (n=81)        (n=81)

Hypothesis          .45        .55
GMPM change        -.06        .22
GMFM change         .44(a)     .40

(a) One-tailed probability, [P.sub.1][less than or equal to].05.

Responsiveness

Through analysis of the therapists' responsiveness subscales, we identified subgroups of 36 "stable" and 60 "responsive" children with CP (Tab. 8). The "stable" group showed no difference between first and second assessments in total GMPM scores (F=2.13; df=1,35; [P.sub.2]=.1533). This finding indicates that the "stable" group had little variability in GMPM scores over time and were not changing in motor performance. The "responsive" group, however, showed change in GMPM scores between the first and second assessments (F=12.30; df=1,59; [P.sub.2]=.0009). Thus, changes in GMPM total scores accurately reflected therapist judgments of overall change in performance in children with CP.

[TABULAR DATA OMITTED]

An analysis of the parents' responsiveness subscales identified 17 "stable" and 66 "responsive" children with cerebral palsy. There were significant changes in GMPM total scores for both the "stable" group (F=9.45; df=1,16; [P.sub.2]=.0073) and the "responsive" group (F=8.97; df=1,65; P2=.0039). The "stable" group appeared to experience a greater change (4.6 units) than the "responsive" group (2.8 units). These data may indicate parents, lower specific ability to judge overall change in a construct such as motor performance.

Intraclass correlation coefficients calculated from the F statistics in this ANOVA showed a difference between the therapist-identified "stable" and "responsive" groups ([P.sub.2]<.001), but not for the parent-identified groups ([P.sub.2]>.05).

This responsiveness analysis demonstrates the ability of the GMPM to detect change in motor performance in children who have been judged by trained therapists to be changing in their overall quality of movement. This capability is crucial for evaluative instruments being used in clinical trials of treatment effectiveness.

Discussion and Conclusions

There is sufficient evidence from this study to satisfy a number of our original objectives in assessing the validity and responsiveness of a measure of quality of movement in a clinical setting. Attributes of gross motor performance can be defined, matched with relevant gross motor function items, and measured with a common scale. A GMPM total score can be generated that may be of use in describing overall motor performance.

An event could occur in assessment of very young children or children with severe disabilities, who cannot per form many motor function items, which could have implications for the interpretation of summary scores. The use of percentage scores for individual attributes may occasionally result in attribute scores based on one or a few observations. The contribution of this score to the total score may outweigh its overall importance to the construct of motor performance. Analysis of individual attribute data will allow us to evaluate the impact of this issue and subsequently to make necessary changes to the GMPM scoring procedures.

A limiting factor in this study was the low statistical power to detect differences between certain subgroups. Although we came close to enrolling a satisfactory number of children in the study, the number of children in the age X severity groups was often too low to allow us to be confident that we avoided type II errors. This issue, combined with the relatively small effect size, or degree of change in motor performance in children with CP, could lead to an inability to detect differences even if they actually exist. A small effect size may also have contributed to poor correlations between GMPM change scores and parent/therapist judgment ratings of change. These problems require us to interpret the data with some caution. Other studies have reported similar problems in comparing observational and biomechanical measures of motor performance in children with CP.[22]

Clinical hyphotheses were supported regarding the ability of the GMPM to detect changes in movement quality in children of different diagnostic and severity groups. The observation that major improvements in motor performance occurred in older children with CP was unexpected. This pattern of change in motor performance in the older group contrasts sharply with observed change in motor function in a similar population. Russell et al[10] reported that the greatest improvement in motor function in children with CP occurred when the children were young. The data from our study tentatively indicated that the greatest improvement in motor performance occurred in both the 3- to 4-year-old children with moderate CP and the older children with moderate CP who had presumably already acquired most of their basic motor skills. These preliminary findings should be investigated further in longitudinal studies with larger sample sizes. Other factors that may contribute to differences in motor quality, such as variations in treatment frequency, should also be controlled for in such studies.

Hypotheses regarding the relationship between the GAUM and GMPM were partially supported. Further analysis of the data will allow us to explore the critical relationship between function and performance in different severity and age groups of children with CP. The hypothesis regarding the relationship between GMPM scores and therapist and parent ratings of change was not well supported. The rating task required of the parents and therapists was to estimate the degree of overall change in movement quality of the child. There are several possible explanations for the low correlations observed. The GMPM scores of children with CP changed very little over the two assessments. Therapists' global impressions of these changes may have been influenced by concurrent changes in function as well as by changes in other aspects of motor quality not assessed by the GMPM (eg, speed of movement). In addition, a change that was observed in a single attribute on the GMPM may have influenced therapist global ratings. This change, however, may not have been reflected in the total GMPM score. Finally, because the GMPM sampled a small number of the qualitative attributes of movement, it may not have been realistic to expect that overall judgments of movement quality by parents and therapists would be correlated with GMPM change scores. Thus, further refinement of the concepts and improved training of therapists in distinguishing between motor function and performance may be required.

The responsiveness analysis, however, supported the use of the GMPM as a measure for evaluating change in motor performance. The initial analysis indicated that the GMPM scores changed in "responsive" groups of children with CP and indicated that when children are independently classified by therapists as either "stable" or "responsive," changes in the GMPM accurately reflect these groupings. Nonetheless, this finding contrasts with the inability of the therapist and parent rating scales to correlate with GMPM change scores. Alternate wording for "overall change in motor performance" used in the rating scales and the responsiveness subscale may account for the observed findings. This reinforces the conclusion that further work is necessary before the instrument's validity can be fully established.

We believe the GMPM represents an important attempt to construct and validate an observational measure of quality of movement for use with children with CP. Considerable conceptual, methodological, and practical challenges have been partially overcome. Conceptually, there have been advantages in utilizing an existing measure of gross motor function as a source of observable activities. There are also difficulties, however, for observers to distinguish between motor function and performance in scoring activities.

In practical terms, the need to simultaneously observe and rate three attributes of performance for each activity has proven a difficult task for therapists. Those therapists who assessed numerous children reported that this task became much easier with experience. A number of these issues must be resolved before the GMPM is ready for use as a measure in a clinical setting. At this time, the GMPM is more suitable for research in controlled settings with well-trained therapists.

Further study will be required to determine whether the evaluative properties of the GMPM, or its ability to detect change over time, are as useful as its discriminative properties, such as its ability to discriminate between different diagnostic, severity, and age categories. The availability of a measure of movement quality would allow a fuller assessment of gross motor behaviors in children with CP. Treatments such as neurodevelopmental therapy, use of "tone-reducing" anklefoot orthoses, and dorsal rhizotomy may eventually be tested more completely with the GMPM and GMPM as part of the evaluation protocol.

We are continuing to develop an understanding of motor behavior in children with CP. Future work in this area may involve refinement of the GMPM measure as a clinical tool and evaluation of observer training techniques to ensure that potential users are capable of learning and applying this instrument to the complex patterns of motor behavior in children with CP.

Finally, there are many issues of theoretical importance to motor control research that have been identified, such as the relationships between performance, function, age, and severity of impairment. There is also work to be done on validation of the GMPM with other kinematic measures of motor performance and with other populations of children with motor performance impairments.

Acknowledgments

We acknowledge the therapists, children, and parents who cooperated in this study. Special thanks are extended to the Children's Developmental Rehabilitation Program, Chedoke-McMaster Hospitals; the Hugh MacMillan Rehabilitation Centre; and the Niagara Peninsula Children's Centre. Diane Terris and Jim Chen assisted greatly with data management and analysis.

References

[1] Campbell SK. Measurement in developmental therapy: past, present, and future. In: Miller L, ed. Developing Norm-referenced Standardized Tests. Binghamton, NY: The Haworth Press Inc; 1989:1-13. [2] Hopkins B, Prechtl H. A qualitative approach to the development of movements during early infancy. In: Prechtl H, ed. Continuity of Neural Function From Prenatal to Postnatal Life. Philadelphia, Pa: JB Lippincott Co; 1984:143-145. [3] Bobath B, Bobath K. The neurodevelopmental treatment. In: Scrutton D, ed. Management of the Motor Disorders of Children With Cerebral Palsy: Clinics in Developmental Medicine (No. 10). Philadelphia, Pa: JB Lippincott Co; 1984:6-18. [4] Stockmeyer S. An interpretation of the approach of Rood to the treatment of neuromuscular dysfunction. Am J Phys Med. 1961;46: 900-956. [5] Bly L. The Components of Normal Movement During the First Year of Life and Abnormal Motor Development. Oak Park, Ill: Neuro-Developmental Treatment Association; 1983. [6] Campbell SK. Assessment of the child with central nervous system dysfunction. In Rothstein JM, ed. Measurement in Physical Therapy. New York, NY: Churchill Livingstone Inc; 1985:207-228. [7] Scherzer AL, Tscharnuter I. Early Diagnosis and Therapy in Cerebral Palsy. New York, NY: Marcel Dekker Inc; 1982:87-97. [8] Harris SR. Movement analysis: an aid to early diagnosis of cerebral palsy. Phys Ther. 1991; 71:215-221. [9] Rosenbaum PL, Russell DJ, Cadman DT, et al. Issues in measuring change in motor function in children with cerebral palsy: a special communication. Phys Ther. 1990;70: 125-131. [10] Russell DJ, Rosenbaum PL, Cadman DT, et al. The gross motor function measure: a means to evaluate the effects of physical therapy. Dev Med Child Neurol. 1989;31:341-352. [11] Russell DJ, Rosenbaum PL, Gowland C, et al. Manual Gross Motor Function Measure: A Measure of Gross Motor Function in Cerebral Palsy. Hamilton, Ontario, Canada: Chedoke-McMaster Hospitals; 1990. [12] Boyce WF, Russell DJ, Rosenbaum PL, et al. Measuring quality of movement: a review of instruments. Phys Ther. 1991;71:813-819. [13] Boyce WF, Goldsmith CH, Gowland C, et al. Development of a quality-4-movement measure for children with cerebral palsy. Phys Ther. 1991;71:820-832. [14] Boyce WF, Gowland C, Russell DJ, et al. Consensus methodologies in the development and content validation of a gross motor performance measure. Physiotherapy Canada. 1993;45:94-100. [15] Corcos DM. Strategies underlying the control of disordered movement. Phys Ther. 1991; 7:25-38. [16] Guyatt GH, Walter S, Norman G. Measuring change over time: assessing the usefulness of evaluative instruments. Journal of Chronic Disease. 1987;40:171-178. [17] Kirshner B, Guyatt G. A methodological framework for assessing health indices. Journal of Chronic Disease. 1985;38:27-36. [18] Russell DJ, Rosenbaum PL, Lane M, et al. Training users in the Gross Motor Function Measure: methodological and practical issues. Phys Ther. 1994;74:630-636. [19] Boyce WF, Gowland C, Rosenbaum PL, et al. Gross Motor Performance Measure for children with cerebral palsy: validation study design and preliminary findings. Can J Public Health. 1992;83(suppl 2):S34-S40. [20] Kramer HC, Karner AF. Statistical alternatives in assessing reliability, consistency and individual differences for quantitative measures: application to behavioral measures of neonates. Psychol Bull. 1976;83:914 -921. [21] Burckhardt C, Goodwin L, Prescott P. The measurement of change in nursing research: statistical considerations. Nurs Res. 1982;31: 53-55. [22] Reddihough D, Bach T, Burgess G, et al. Comparison of subjective and objective measures of movement performance of children with cerebral palsy. Dev Med Child Neurol. 1991;33:578-584.

Appendix 1. Gross Motor Performance Measure Attributes and Their Definitions

Alignment

The adjustment or arrangement of parts or segments of the body in relation to each other. Coordination

The smooth and controlled use of movements in motor performance. This takes into account timing, velocity, direction, force, and amplitude. Dissociated Movement

Isolated movement. Movement of one part of the body independent of another part (ie, one limb

independent of another limb, or one segment of the limb or trunk independent of another segment). Movement that combines components of different motor patterns (eg, extension of the hip with flexion of the knee). Stability

The active maintenance of a body position in the presence of disturbing forces. Weight Shift

Movement that involves a transfer of the body's center of gravity. This takes into account

amount and direction. The transference can take place in any of the four planes of the body (posteriorly, laterally, anteriorly, and vertically) or any combination of these.

Appendix 2. Gross Motor Performance Measure Item Example

The child is positioned and asked or encouraged to complete an activity

Gross Motor Function item

Sitting on mat with feet and small toy in front, child leans forward, touches toy, reerects without arm propping

0. Does not lean forward and reerect

1. Leans forward but cannot reerect

2. Leans forward, touches toy, reerects with arm propping

3. Leans forward, touches toy, reerects without arm propping

During three repetitions of this activity, the therapist assessess and scores

Gross Motor Performance Attributes

Alignment of Trunk and Pevis

1. Severely malaligned the majority of the time; majority of movement is execluded in the extremes of range inappropriate to the performance of the task 2. Moderatety malaligned the majority of the time 3. Mildly malaligned the majority of the time 4. Demonstrates one trial with completely normal aignment and no evidence of movement pathology 5. Consistentty normal alignment for all three trials

Dissociated Movements of the Reaching Upper Extremity 1. No dissociation the majortty of the time 2. Moderate impairment of dissociation the majority of the time; absence of efficient movement of one segment relative to another 3. Mild impairment of dissociation the majority of the time 4. At least one trial with completely normal dissociat on, with no evidence of movement pathology 5. Consistently normal dissociation for all three trials' efficient movement of one sagment relative to another within normal variations

Weight Shift 1. Severely abnormal weight shift the majority of the time; inappropriate position of center of gravity; inappropriate direction of weight shift 2. Moderately abnormal weight shift the majority of the time 3. Mildly abnormal weight shift the majority of the time; weight shift always in the appropriate direction 4. At least one trial with completety normal weight shift, with no-evidence of movement pathology 5. Consistently normal weight shift for all three trials