Advertisement

Abstract

Background and Purpose. The Pediatric Evaluation of Disability Inventory (PEDI) subscales are hierarchic in nature, with item placements within each subscale reflecting the general order of skill attainment in children without disabilities. The purpose of this study was to determine whether a hierarchical subscale developed in this study for children with acquired brain injuries (ABIs) corresponds to the generic PEDI subscales, and, if not, whether condition-specific (ABI-specific) PEDI subscales are more sensitive for measuring change. Subjects. Eighty-seven children and adolescents (mean age=9.2 years, SD=5.2, range=1–20) with ABI during inpatient rehabilitation admissions were included. Methods. Data were collected by retrospective chart review. Rasch one-parameter analyses were conducted to construct the ABI-specific PEDI scale focusing on the Mobility and Self-care domains (content areas) only. Each domain consists of a Functional Skills subscale and a Caregiver Assistance subscale. Thus, in all, 4 scales were constructed within the ABI-specific PEDI scale. Differences in item hierarchies and sensitivity between generic and ABI-specific PEDI subscales for each domain were then examined. Results. Both generic and ABI-specific PEDI scales were sensitive for measuring functional changes during inpatient rehabilitation hospitalization. Even though the generic and ABI-specific item hierarchies differed substantially, only one of the 4 ABI-specific PEDI subscales (Caregiver Assistance Self-care subscale) was more sensitive for measuring change than the generic PEDI scale. Discussion and Conclusion. The ABI-specific scales added relatively little improvement in sensitivity compared with the generic PEDI scales of the Mobility and Self-care domains. Thus, for group analyses, the authors recommend use of the generic PEDI subscales for children with ABI. Future work with ABI-specific subscales may improve the physical therapist's ability to describe an individual's pattern of functional recovery.

Rehabilitation providers are facing increasing pressure to provide evidence of improved function as part of meeting rehabilitation goals. The main goals of physical therapy intervention in pediatric rehabilitation, we believe, are to reduce barriers limiting the performance of daily routines and to facilitate the successful integration of children into the home and school environments. Functional status measures, which focus on assessment of behaviors needed for daily routines in the home and the community, in our opinion, are needed to evaluate the effectiveness of physical therapy intervention. The Pediatric Evaluation of Disability Inventory1 (PEDI) was developed in an effort to provide a comprehensive assessment of functional capabilities and performance in children and adolescents with disabilities and to be responsive for measuring changes in functioning as a result of episodes of physical therapy intervention.

The PEDI1,2 scale structure is depicted in Figure 1. The PEDI consists of 3 parts: (1) Functional Skills (197 items), (2) Caregiver Assistance (20 items), and (3) Modifications (20 items). Capability of a child is assessed by the identification of skills for which the child has demonstrated mastery and competence using the PEDI Functional Skills part. Level of performance is measured by the amount of caregiver assistance needed by the child to accomplish major functional activities included in the Caregiver Assistance part. Each part consists of items that focus on 3 content areas or domains: (1) Self-care, (2) Mobility, and (3) Social Function. Our study focuses only on the Self-care and Mobility domains. The Self-care domain includes activities such as eating, grooming, dressing, bathing and toileting. The Mobility domain includes activities such as floor mobility, ambulation, transfers, and mobility in different environments with or without adaptive equipment. Each domain contains a Functional Skills subscale, a Caregiver Assistance subscale, and a Modifications subscale. The subscales included in our study are shown in Figure 1. On the Functional Skills subscales, using a dichotomous (not capable/capable) scoring system, examiners indicate the child's capability to perform (without physical assistance) the 73 self-care tasks and the 59 mobility tasks. The Caregiver Assistance subscales consist of complex functional activities, such as lower-body dressing or car transfers, in which the amount of assistance provided is assessed on a 6-point ordinal scale (polytomous) ranging from “total assistance” (0) to “independent” (5).

Figure 1.

Pediatric Evaluation of Disability Inventory (PEDI) structure: parts, domains, subscales. The shaded boxes indicate the PEDI subscales used in our study.

Development and construction of all PEDI measurement scales are based on Rasch rating scale methodology.3,4 The Rasch analytic procedure is a one-parameter logistic solution, part of a family of item response theory (IRT) models.5 Rasch analysis creates an estimated position for each PEDI item (dichotomous) or for each rating scale category (polytomous items) on a hierarchy representing a continuum of a functional ability.

Data are used to construct these item positions. Items that are easier for most children (eg, crawling, sitting, indoor walking) are placed at the less difficult end of the measurement subscale, and items that are harder for most children (eg, walking outdoors, use of stairs) are placed at the more difficult end of the measurement subscale. Figure 2 is a schematic representation of the Functional skills Mobility subscale. The Rasch measurement scales enable an estimation of item positions and person scores on the same scale as depicted in Figure 2. Other instruments used commonly in pediatric physical therapy also have utilized a one-parameter Rasch model for scale validation and positioning items along constructs such as gross motor skills (Gross Motor Function Measure).6

Figure 2.

Schematic representation of Pediatric Evaluation of Disability Inventory (PEDI) Functional Skills Mobility subscale and item locations (selected items).

In theory, an instrument designed to measure function should have reproducible item locations in different samples and on different test occasions.7,8 The underlying assumption of Rasch scales is that item hierarchies are sample-free, that is, they are applicable across a variety of samples. An analogy can be made to a measuring device such as a ruler. The ruler is marked off in increments of distance (in centimeters), and these distances are fixed. That ruler can be used for any occasion for which a measure of distance (in centimeters) is needed. Similarly, in an application of a measurement scale for functional assessment, these item locations are considered fixed for any sample and thus are potentially valuable in the interpretation of change in different samples.2,9

In practice, item positions and the order of items within functional instruments across samples can vary slightly.1013 For instance, Haley et al previously found that the physical functioning items of the SF-36 varied slightly across people with different chronic conditions,11 and the subsequent variation in sensitivity of these measurements was tolerable, given the purpose of measuring change in physical functioning in large group analyses.12 One major unanswered question with the generic PEDI scale is whether its subscales are truly applicable to all samples. For example, are the PEDI item positions and order (hierarchy) applicable for measuring function in children with various conditions? If differences exist in the order of items across samples, this leads to the question of whether variability in item positions or item order across samples has an impact on the calculation of summary and change scores on the PEDI. To date, there is only one report of an examination of the stability of the PEDI item positions. Haley and colleagues14 reported only minor differences in item positions of the Caregiver Assistance Mobility and Self-care subscales of the PEDI when measurements from home and school settings were compared. This finding, we believe, indicates that the Caregiver Assistance subscales of the PEDI can be used to monitor progress over time in children when the initial and follow-up assessments are performed in different settings.

In our study, a sample of children with ABI was used to examine these 2 questions. We chose to study children with ABI because ABI is a leading cause of disability in childhood15,16 and accounts for a large number of pediatric rehabilitation hospital admissions per year in the United States.17 This cohort of children with disabilities would represent a large percentage of the caseload for rehabilitation specialists such as physical therapists and occupational therapists. A second reason we chose to study children with ABI was because of the observed variability in the course and outcome of recovery in this population. We would expect that this population would differ from the typically developing population in the order of skill attainment during recovery.

We believe there are major practical advantages in using a generic functional assessment instrument such as the PEDI that can be applied to children with a variety of conditions if PEDI measurements can be shown to be sample free and the order of items does not influence test sensitivity. We proposed to test the applicability of the generic PEDI in children with ABI. The purpose of our study was to test the congruence of functional item positions along Rasch measurement scales from a sample of children without functional disabilities and an ABI-specific sample. Further, if item positions among the measurement scales differ, we predicted that these item locations and order disparities would have little or no effect on the sensitivity of the PEDI summary scores to detect changes in children with ABI during inpatient rehabilitation hospitalization.

Method

Subjects

Subjects were 87 children (55 male and 32 female) recovering from brain injury who were admitted to the inpatient service at Franciscan Children's Hospital and Rehabilitation Center (Boston, Mass) between October 1994 and July 1997. The mean age of the subjects was 9.2 years (SD=5.2, range=1–20). The subjects were evaluated at admission and discharge by the therapy staff using the PEDI. The data were collected for this study by retrospective chart review. The PEDI is routinely administered to children with brain injury upon admission to and discharge from the hospital in which the study was conducted. Therapists view an instructional video, attend initial training sessions conducted by supervisory clinical staff, and complete the case studies in the PEDI manual1 in preparation for administering the PEDI in routine clinical practice. Annual in-services are conducted to review scoring and interpretation of test results for use in patient management. The Self-care domains are administered by an occupational therapist, and the Mobility domains are completed by a physical therapist. In almost all cases, the primary therapist scores both the admission and discharge assessments for any child.

We selected this particular cohort of children with ABI because we had complete item-level data available in the hospital records. The diagnostic classifications of ABI included traumatic brain injury (58.6%), seizures (4.6%), brain tumors (13.8%), hemorrhage (9.2%), anoxia (3.4%), and cerebral infarct (2.3%). Of the potentially available 94 records, 87 records were complete with admission and discharge PEDI administrations, although a few PEDI scales had missing or unusable data. Demographic data are presented in Table 1. We did not stratify the sample based on diagnostic classification within the broad category of ABI because of the effect of we believe such stratification would have in reducing the sample size needed for the creation of a reliable ABI measurement scale.

Table 1.

Subject Demographic Data (N=87)

Psychometric Properties of the PEDI

Several researchers1821 have described the reliability and validity of PEDI scores in a wide variety of clinical samples. Nichols and Case-Smith18 examined intrarater reliability of PEDI scores in a sample of children with varied diagnoses (n=23). Intraclass correlation coefficients (ICCs) ranged from .67 to .99 for the Functional Skills Mobility subscale and from .69 to 1.00 for the Functional Skills Self-care subscale. The ICCs ranged from .83 to .98 for the Caregiver Assistance Mobility subscale and from .68 to .90 for the Caregiver Assistance Self-care subscale. They also examined test-retest reliability via 2 interviews with the same parents within a week. The results, in our view, were fairly stable over the period of 1 week, providing support for the test-retest reliability of the PEDI scores.

Wright and Boschen19 evaluated the intrarater and interrater reliability of scores for multiple respondents (physical therapists, occupational therapists, parent, and teacher) when using the PEDI Functional Skills part for all 3 domains with 40 children with varying severities of cerebral palsy aged 3 to 7 years.20 The Functional Skills part of the PEDI yielded ICCs greater than .95 for all the total scores and ICCs greater than .80 for all 3 domains within each respondent group. Between the respondent groups, the Functional Skills Self-care subscale yielded ICCs of .87 and the Functional Skills Mobility subscale yielded ICCs of .81.

Haley at al20 examined the content validity of PEDI scores using a panel of 31 experts in the field of pediatric rehabilitation. The experts were requested to complete a questionnaire to determine whether the PEDI measured function and the extent to which the items on the PEDI sampled pediatric functional activities. The results of this study, we believe, served to validate and confirm the functional content of the PEDI. Feldman et al21 examined what they considered construct validity of the PEDI by testing whether it could be used to discriminate between children without functional disabilities (n=20) and children with functional disabilities (n=20). Their results showed that PEDI Mobility and Self-care domain scores were different between the children with disabilities and the children without disabilities.

PEDI Subscales Based on Data From a Sample of Children Without Functional Disabilties (Generic PEDI Scale)

During the initial standardization of the PEDI, data were collected from 412 children without functional who ranged in age from 6 months to 7.5 years. These data were used to define the “standard measurement scale” for newly acquired functional items. Details of the sampling approach are given in the PEDI manual.1 The Rasch model transforms raw scores into interval measures by using log-odd units (logits). Logits run from positive infinity to negative infinity. To make the scores more easily interpretable, a linear transformation is applied,4 which calibrates the scale from 0 to 100 (as shown in Fig. 2) instead of logit units. Thus, scores on each scale could range from 0 to 100, with scores approaching 0 indicating limited functional ability and scores approaching 100 indicating problem-free ability. Larger ability scores indicate an increase in the level of capability in performing a functional skill and an increase in the level of independence to perform complex functional activities. Using a simplified example with selected items from the Functional Skills Mobility subscale, the number of “capable” items is converted to a transformed score on a 0 to 100 scale. Using the example in Figure 2, if the child is capable of successfully completing only one item, it is most likely to be the item that is at the bottom of the hierarchy; in this case, the item would be “crawls, creeps,” and the child would be assigned a score of approximately 6. If a child is capable of successfully completing 6 items, the child's performance is converted to a score of approximately 50 (based on the expected performance of all items up to and including “walks outdoors”). Scores on the PEDI are directly related to the positions of the items along the measurement subscale.

The transformed scores do not take into account the age of the child, but they provide an estimate of the level of capability in each domain. For this reason, transformed scores theoretically can be used for children of all ages. The values of the transformed scores, derived from a sample without functional disabilities, are organized in the form of tables in the PEDI manual,1 such that each possible raw summary score has a corresponding transformed score value based on the item locations within that scale.

PEDI Scales Based on Data From ABI Sample (ABI-Specific PEDI)

Individual item scores of the Functional Skills and Caregiver Assistance Self-care and Mobility sub–scales of the PEDI obtained shortly after hospital admission and then at the time of discharge were analyzed using WINSTEPS,22,23 a Rasch computer program. One approach is to fix the calibrations of the discharge data.24 This means that the discharge data were used to create the ABI-specific item ruler. This was done because the variability of the items was greater at discharge, as some subjects improved small amounts, whereas others improved much more. Anchoring locations to the discharge data allowed construction of hierarchical scales based on order of item attainment following recovery. At the time of admission, many of the items were truncated due to the subjects' limited functional performance, and therefore admission scores could not be used to develop the hierarchical scales. Thus, Rasch analysis was used to estimate item locations based on discharge scores. The summary score estimates at admission and discharge were then computed using the measurement scale transformations based on the discharge item locations.

Data Analysis

To estimate the degree of correspondence of the item locations between the generic and ABI-specific scales, Pearson product moment correlation coefficients were used to evaluate the strength of the relationship between average logit item locations obtained from use of the 2 scales. A series of z tests4 were then performed on the respective pairs of item positions to identify those items that had different positions between the 2 scales (P<.05). A critical value of 1.96 was used to identify those items that had item positions that were different (P<.05) between the 2 scales.4 These individual tests were not corrected for a Type I error because their purpose is to locate any potential discrepancy between conditions. This strategy does not protect against making a Type I error but rather the commission of a Type II error, that is, overlooking an item pair with potential differences.4 We conducted separate analyses for each of the subscales within Mobility and Self-care domains.

We used paired t tests to evaluate whether the mean change scores between the admission and discharge data of our study sample were different from zero. The responsiveness of the 2 scales (ABI-specific and generic), to assess change, was then determined on the basis of relative precision (RP)12 estimates and effect size. Relative precision is a ratio of pair-wise F statistics (Fclinical/Fnormative). F is equal to t2; the value of t was obtained from the paired t tests. In our study, the RP estimates indicated, in proportional terms, how much more or less precise the ABI-specific PEDI scales were than the generic PEDI scales in detecting change in functional status in children with ABI.

Relative precision depends on 2 elements: the magnitude of difference between the 2 test occasions being compared and within-group variance. We hypothesized that the RP estimate would be 1; that is, both scales would be equally sensitive to change, or the magnitude of change measured by both scales would be equal. An RP estimate of greater than 1 would indicate that the ABI-specific scale is more sensitive to change in the functional status of children with brain injury than the generic scale. The level of significance was set at P<.05. The effect size for each scale was calculated as the ratio of the mean change score to the standard deviation of the score at admission. To evaluate the responsiveness of the 2 scales, data sets of only those children with complete admission and discharge scores were used for analysis.

Results

Strength of the Relationship of Item Locations

Correlations between the item locations obtained using the ABI-specific and generic scales were r=.77 (P<.01) for the Functional Skills Mobility subscale and r=.87 (P<.01) for the Functional Skills Self-care subscale. Correlations did not reach significance for the Caregiver Assistance Mobility and Self-care subscales.

Differences in Item Locations

Our results showed differences in item position and order between the generic and ABI-specific scales. The estimated item locations along a continuum for either the Self-care or Mobility domain suggest a relative placement of activity difficulties within a sample. In the convention we have adopted, item locations with small values mean that most children can accomplish the activity, whereas item locations with larger values indicate that the items are more difficult to accomplish. Overall, a greater number of the Mobility domain items differed in their item locations between the 2 scales as compared with the Self-care domain items (Tab. 2). In general, with the ABI-specific scale, items common to both the Functional Skills and Caregiver Assistance parts that yielded larger item location estimates included the locomotor items in the Mobility domain and eating items in the Self-care domain. This finding indicates that these items were more difficult to accomplish for most of the children. Items with smaller item location estimates included the transfer items and stairs in the Mobility domain and grooming in the Self-care domain, which indicates to us that these items were fairly simple to accomplish for most of the children. Items specific to the Functional Skills part with a smaller item location estimate included bathing and with a larger item location estimate included certain lower-body dressing skills for the Self-care domain. (For a full list of the specific item calibrations and differences across the 2 scales for each part or subscale, contact the first author [DHK].)

Table 2.

Correspondence of Item Calibrations Between the Generic and Acquired Brain Injury-Specific Scales

Relative Precision and Effect Size

Our results indicate that variability in item positions or item order across samples has an impact on the calculation of summary and change scores on the PEDI. The paired t tests revealed that the mean difference in the transformed scores at admission and at discharge within our study sample, computed separately for the ABI-specific and generic scales, were different from zero for the Mobility and Self-care domains. This finding indicates to us that both scales are sensitive to changes in functional status. Relative precision estimates and effect sizes for the 4 subscales are presented in Tables 3 and 4. A small loss of precision for the ABI-specific scale was noted for the Functional Skills Mobility and Self-care subscales, with RP estimates of 0.97 and 0.96, respectively (Tab. 3). The ABI-specific scale had an RP estimate gain of 4% (RP estimate=1.04) for the Caregiver Assistance Mobility subscale (Tab. 4) and an RP estimate gain of 20% (RP estimate=1.20) for the Caregiver Assistance Self-care subscale (Tab. 4). Effect size, we contend, is an alternate index of sensitivity to longitudinal change that relies on the within-subject variability at baseline (standard deviation of the scores at admission). Effect size findings were consistent with the RP estimates, with approximately the same effect size for both scales (generic and ABI-specific) except the Caregiver Assistance Self-care subscale. For this subscale, the effect sizes for the generic and ABI-specific scales were 0.58 and 0.69, respectively, indicating that the ABI-specific scale is more sensitive than the generic scale to change in the level of independence in the Self-care domain.

Table 3.

Relative Precision Estimates and Effect Sizes for Functional Skills Subscalesa

Table 4.

Relative Precision Estimates and Effect Sizes for Caregiver Assistance Subscalesa

Discussion

The PEDI was developed as a generic instrument to measure function in children and adolescents with a variety of disabling conditions. If the PEDI is to be useful for children with a variety of diagnoses, the item locations and item order should be relatively stable across various diagnoses. Our data suggest that the locations and order of item positions in children at early stages of recovery from ABI differ substantially when compared with the model developed for children without disabilities. This variance appears to have only minor scoring consequences, but we believe it may have important implications for understanding mobility and self-care recovery of children with ABI.

Scoring Implications

Our findings, we contend, have important implications for using the PEDI in clinical outcome studies where the magnitude of the change is examined. Our results suggest that the PEDI scores obtained by examining children without disabilities are acceptable for detecting change in children with ABI. Although a substantial number of items had different paired item locations across samples, the only major difference in RP estimates of the 2 scales was only on one subscale (Caregiver Assistance Self-care subscale). We maintained an alpha level of .05 when identifying item pairs that had calibrations that were different. This means that the Type I family-wise error rate most likely exceeded .05 because of the multiple tests, and consequently some of the differences we observed in item pairs may have been due to chance. This strategy, however, enabled us to identify any potential item that might be influenced by the difference in samples. One of the major influences on the differences in paired calibrations was more limited variability in the ABI-specific scale in comparison with the generic scale. Even though we used the discharge item calibrations (greater variability than the calibrations at admission) as an anchor for the ABI scale, yet the variability of the ABI calibrations approached only 70% to 80% of that of the sample without functional disabilities.

The scoring impact of the ABI-specific scale can be understood by examining the distribution of the rating scale categories for the Caregiver Assistance Self-care and Mobility subscales (Figs. 3 and 4, respectively). We have highlighted in the figures the positions of the scale categories for the item “bathing” on the Caregiver Assistance Self-care subscale and the item “stairs” on the Caregiver Assistance Mobility subscale. The relative distances between scale categories, however, is the same for all Self-care and Mobility domain items within each subscale. Each item on the Caregiver Assistance Self-care and Mobility subscales is scored on a 6-point scale, from “total assistance” (0) to “independent” (5). In Figures 3 and 4, we have identified scale locations for “maximum assistance=1,” “moderate assistance=2,” “minimal assistance=3,” “supervision=4,” and “independent=5.”

Figure 3.

Pediatric Evaluation of Disability Inventory (PEDI) Caregiver Assistance Self-care subscales and locations of rating scale categories for the generic and ABI-specific scales.

Figure 4.

Pediatric Evaluation of Disability Inventory (PEDI) Caregiver Assistance Mobility subscale and locations of rating scale categories for the generic-based and ABI-specific scales.

For the item “bathing” on the Caregiver Assistance Self-care subscale (Fig. 3), the distances between the rating categories of “supervision” and “independent” were greater on the ABI-specific scale than on the generic scale. When a child improves in bathing from “supervision” to “independent,” a change of approximately 25 units is recorded using the ABI-specific scale, whereas a change of only 10 units occurs in the generic scale. The ABI-specific scale depicts a greater interval between the more advanced levels of the scale and is, therefore, more sensitive in this range than the generic scale. In our sample, a number of children with ABI moved from supervision levels to independence. This difference in the distribution of the levels of the rating scale was observed most clearly in the Caregiver Assistance Self-care subscale and explains the improvement in sensitivity found in the ABI-specific scale versus the generic scale. Thus, it appears that the relative distances between scale points can have an impact on the scale sensitivity.

For the item “stairs” on the Caregiver Assistance scale of the Mobility domain (Fig. 4), there was a wide spread in the location of the scale categories for the ABI-specific scale as compared with the generic scale, but this difference was not sufficient to have an impact on scale sensitivity (RP estimate=1.04). Although the ABI-specific scale was not found to be more sensitive than the generic scale, the relative distances between its scale points may reflect the patterns of behavior specific to children with ABI. As shown in Figure 4, the location of the scale category “stairs-independent” was relatively higher on the ABI-specific scale than on the generic scale. This finding may reflect the need for more close supervision in children with ABI on stairs than in children without functional disabilities who may not exhibit the same degree of variability in safety behaviors around stairs. Future work with ABI-specific scale may improve the scale so that physical therapists can better describe individual patterns of functional recovery.

Clinical Implications

Our study indicates that early stages of recovery in children with ABI do not necessarily follow the same pattern of re-acquiring competence and independence in performance of functional activities as occur with children who are typically developing. This is indicated by the differences found in the order and spacing of the item locations on the Functional Skills and Caregiver Assistance Mobility and Self-care subscales between the ABI-specific and generic models. Our findings are consistent with those of several other researchers16,2527 who examined the short- and long-term manifestations in children with brain injury and found disruption of normal development secondary to multiple deficits in motor, communicative, cognitive, sensory, behavioral, and emotional systems.

Paired items with different calibrations between samples were closely inspected to identify the possible underlying themes. A group of items occurred relatively high in the hierarchic sequence for children with ABI as compared with children without functional disabilities on the PEDI Functional Skills part. This group included items such as indoor locomotion, outdoor locomotion, and tub transfers in the Mobility domain and items such as lower-body dressing and manipulation in the Self-care domain. Capability on these Mobility domain items is believed to require balance, appropriate reaction times to prevent falls, endurance, and peer-matched speed and muscle force, whereas success on the Self-care domain items is believed to require grasping strength, precision, eye-hand coordination, and bilateral coordination, which are common deficit areas after brain injury.28

A second group of items occurred relatively earlier in the sequence of items for the children with ABI than for the children without functional disabilities. These items included basic transfer skills, stairs, and bathing and grooming. We believe there are a couple of reasons for this pattern. Transfer skills pose contextual challenges related to height, such as the chair or surface height, size of the toilet, and height of the bed or crib. The sample with ABI consisted of children who were older (age range=1–20 years; 54% of the children were above 7.5 years of age) than the children without disabilities (age range=6 months to 7.5 years). The pattern seen in grooming and bathing skills also may be age-related, as hygiene tasks are expected of older children and are especially important to adolescents. The use of adaptive equipment such as sliding boards or tub benches may facilitate the relearning of transfer skills in children with ABI.

We noted similar patterns of item sequence differences for the Caregiver Assistance part. Items higher in the hierarchic sequence in children with ABI included indoor and outdoor locomotion and eating, all potentially related to concerns with safety and continued need for supervision. Coster and Haley29 have presented a conceptual model of pediatric disability that underscores the influence of “context” in the ability of a child to perform a functional task. Functional independence is highly dependent on opportunities presented by the social and physical environments that a child experiences.30 The differences in the item calibrations between the 2 scales also may be partially due to the different settings in which the sample of children with ABI and the sample of children without functional disabilities were assessed (hospital versus community, respectively).

The relatively small sample size of children with ABI limits the generalizability of our conclusions. We considered traditional Rasch model issues of item and person misfit, item and person separation, rating scale threshold estimates, and diagnostic analysis of the residuals to be irrelevant for the purpose of our study. As in the original development of the PEDI, misfit issues and residual analyses are crucial for establishing the soundness of the psychometric characteristics of a new measurement instrument. In such an application, we would argue that the fit of the data to the model should be aggressively undertaken from many different perspectives. In our analysis, however, no intent was presumed. Rather, the analyses were conducted to establish whether a difference in the ABI-specific and generic scales would have an impact on summary scores.

Despite the limitations of a relatively small sample, we believe our analyses have laid important groundwork for understanding the impact of condition-specific scales on summary scores. The statistical effect of the small sample was to yield standard errors of item estimates that were larger than would have been the case if more clinical data were available from more subjects. Furthermore, we recognize that a small sample size cannot adequately represent the diversity of functional patterns seen in the recovery of children with ABI. Researchers in the future should seek to verify the item hierarchy in a well-controlled prospective study with a larger group of children recovering from ABI and to establish how that knowledge can be translated to improve physical therapists' management of patients or clients.

Conclusion

Our results suggest that the generic PEDI Self-care and Mobility domains with dichotomous and ordinal rating systems can be calibrated using Rasch analysis to convert categorical data to an interval-like scale. This transformation, in our view, should enable clinicians to quantify functional change accurately. For group analyses in longitudinal outcome studies, the PEDI scales based on children without functional disabilities can be used to document changes in self-care and mobility in children with ABI. Differences in item locations between the 2 scales (the generic PEDI and the one developed for children with ABI) indicate that the order of skill attainment following recovery in children with ABI in the Self-care and Mobility domains differs from that developed for children without functional disabilities. In future studies, if the item hierarchy of the ABI-specific scale is verified on a larger sample, it may be instructive to study these patterns of recovery more closely to better understand, manage, and predict functional outcome in children with ABI.

Footnotes

  • All authors provided concept/idea/research design. Ms Kothari and Dr Haley provided writing and data analysis. Ms Dumas provided data collection, subjects, and facilities/equipment. Dr Haley, Dr Gill-Body, and Ms Dumas provided consultation (including review of manuscript before submission).

    Ms Kothari was a graduate student at the MGH Institute of Health Professions during this study, which was undertaken in partial fulfillment of the requirements for the post-professional degree of Master of Science in Physical Therapy.

    The Spaulding Rehabilitation Hospital Institutional Review Board and the Human Subjects Committee at Franciscan Children's Hospital and Rehabilitation Center approved this study.

  • Received December 6, 2002.
  • Accepted May 5, 2003.

References

View Abstract