Functioning of young patients with cerebral palsy: Rasch analysis of the pediatric evaluation of disability inventory computer adaptive test daily activity and mobility

Children, adolescents and young adults with cerebral palsy (CP) experience limitations in performing activities of daily living [1]. Parents and their children with CP often identify goals from the levels of activities and participation [2‐5].

Daily activities and mobility are important aspects of functioning which are essential to support children’s and adolescents’ participation in relevant contexts such as the school, home, and community [6]. Interventions focusing on daily activities and mobility for children, adolescents, and young adults with CP should be guided by measures that properly assess these outcomes [7, 8]. Rehabilitation practitioners seek instruments that can measure changes in the performance of relevant activities in children with CP. Such instruments must show good psychometric properties and be easy-to-use [9].

The Pediatric Evaluation of Disability Inventory Computer Adaptive Test (PEDI-CAT) [10] assesses functioning in children, adolescents and young adults with various health conditions. The PEDI-CAT is the new version of the Pediatric Evaluation of Disability Inventory (PEDI), which is frequently used as an outcome measure for children with CP [11‐13]. The PEDI-CAT was developed using current methodologies for test construction, including Item Banking, Item Response Theory (IRT), and Computer Adaptive Testing (CAT). The items are computer-selected from the responses provided by the respondent based on algorithms [10]. This strategy yields improved test score precision and a reduction in assessment time. As a consequence, procedures to test its validity must acknowledge and be congruent with these methodologies.

As the original PEDI assesses functioning in children from 0 up to 7 years, it has a restricted follow-up window [14]. A major contribution of the revised version of PEDI, the PEDI-CAT, is its expanded age coverage from childhood to adolescence and early adulthood, up to 20 years old. Because of these changes, it is important to investigate the validity of PEDI-CAT for older children, adolescents, and youth with CP, whose age ranges were not covered by the original version.

Strong psychometric properties of the PEDI-CAT have been documented for individuals with and without disabilities [15‐17]. Its application so far includes the evaluation of functioning in children and young people with autism spectrum disorders [18], complex medical conditions [19], HIV encephalopathy [20] and spinal muscular atrophy [21].

Investigation of the PEDI-CAT validity for children and youth with CP has used procedures from the Classical Test Theory (CTT). Frazier et al. [22] analyzed the relationship between PEDI-CAT domain scores and categories of CP type and severity. They found associations between CP type and severity with PEDI-CAT scores in all domains. Children with bilateral CP and those from the more severe category scored lower (raw and normative PEDI-CAT scores) than children from other categories. Shore et al. [23] examined two aspects of discriminant validity of the PEDI-CAT domains: (1) ability to differentiate the degree of independence in fine and gross motor function, using receiver operating characteristics (ROC) curve analysis, and (2) differentiation across all Gross Motor Function Classification System (GMFCS) and Manual Ability Classification System (MACS) levels, using general linear modeling. Their results showed that Mobility and Daily Activity domains were able to discriminate, respectively, all GMFCS and three of five MACS levels, whereas the other domains were able to discriminate fewer GMFCS and MACS functional levels. A recent longitudinal study by Burgess et al. [24] investigated self-care developmental trajectories from a population-based sample of children with cerebral palsy (CP), ages from 2 years and 6 months to 12 years of age, across all functional ability levels, according to the MACS classification system. Their results identified different PEDI-CAT DA scores for all five manual functions MACS levels, implying PEDI-CAT was able to distinguish different manual functional levels. To date, and to the best of our knowledge, no study has used Rasch analysis to evaluate the validity of the PEDI-CAT daily activity and mobility items in older children and youth with CP.

Rasch analysis is a probabilistic mathematical model derived from IRT, which transforms raw scores into linear continuous measures of a person’s ability and item difficulty. It allows people and items to be located along the same continuum and to be measured by the same unit, called logit [25]. This model is an alternative to traditional methods in psychometrics and has been widely used in recent studies evaluating the psychometric properties of different instruments [26]. Thus, the analysis of PEDI-CAT psychometric properties using the Rasch approach will further evaluate the validity of this instrument for children and youth with CP, and do so in a way that is consistent with the theoretical model on which it was based. This study aimed to (1) evaluate the fit of daily activity and mobility items and children with CP to the Rasch model, including scale dimensionality, item difficulty, and the reliability and adequacy of the response categories and (2) compare the performance in daily activities and mobility of older children, adolescents, and young adults with CP based on manual function and gross motor function limitations.

This study investigated the validity of PEDI-CAT using Rasch analysis to verify if the DA and MB domains of this instrument are suitable to evaluate the functioning of children and youth with CP. The results showed that 46 items of DA and 43 items of MB fit the Rasch Model. Reliability and PCA scores point to a good fit to the Rasch model, which is consistent with previous studies that evaluated the PEDI-CAT in IRT-based models with other clinical groups [10, 35].

The CCCs did not show the expected pattern for the PEDI-CAT response categories. 36.41% of responses of the DA domain scored 1 (Unable) and 43.99% scored 4 (Easy). For the MB domain, 49.44% of responses scored 1 (Unable) and 33.07% scored 4 (Easy). It suggests that parents or caregivers of children and youth with CP had difficulty distinguishing some response categories (i.e., the medium response categories of hard and a little hard). More frequently, parents had a tendency to choose the two extreme and opposing categories of response (e.g., unable and easy) when answering the PEDI-CAT items. Pasternak et al. [21], found similar results and suggested that, in future reviews of the instrument, the “hard” and “a little hard” response categories should be collapsed. Our results suggest that the middle response categories, which describe the process of skill acquisition, do not seem to be considered by the parents and caregivers when scoring the DA ad MB items. Parents seemed to base their evaluation of their children’s levels of ability as the activities that they can or cannot perform and partial performance or performance with effort were not recognized. In contrast, rehabilitation professionals are often interested in measuring and documenting intermediate stages of the skill acquisition process, when children evolve from being unable to perform an activity to being able to perform only part of an activity, sometimes with a lot of effort, to further being able to perform most of the activity and finally performing this activity with no difficulty. Further studies should verify that parents understand all possibilities of the scoring criteria of the new PEDI-CAT.

The item-person maps showed some inversions of CP severity levels along the two continua, especially in the DA domain. This result suggests that performance of daily activities and mobility activities [1, 23] is not directly determined only by impairment on the gross motor and manual function but, rather, by a combination of factors, including other body functions (e.g., cognitive functions), and personal and environmental factors [36]. It is possible, for example, that a child with CP who has a moderate level of motor function limitation may have environmental supports that allow him/her to create strategies to perform activities, in contrast to another child with a mild level of limitations who does not have the same supports. In that sense, it is possible that children with better manual skills (i.e. MACS I and II) who have cognitive deficits may show difficulty performing more complex daily activities (e.g. Ties shoelaces). Also, studies on prognoses of gross motor function and manual skill have shown that there is some instability of the GMFCS and MACS classification levels over time, that is, children classified as MACS level II, e.g. undergoing intensive rehabilitation, may further become MACS level I [37, 38].

Only five items showed DIF between people with mild and moderate motor function limitations and other two items revealed differences between individuals from moderate and severe GMFCS level. In the DA domain, the three items that differentiated mild and moderate severity levels were complex items from Dressing and Home Task areas. In the MB domain, differences between these severity levels were revealed by two items from the Running and Playing area, which require upper body strength (Moves across monkey bars; Pulls self out of swimming pool not using ladder). The number of items that presented DIF corresponds to 7.7% of the total items evaluated in this study. Coster et al. [35] found similar number of items in the DA domain of the PEDI-CAT that showed differential item functioning among children and youth with and without autism. The authors concluded that such small percentage required no modification in the PEDI-CAT parameters. However, studies have pointed out the need to, once DIF is found, use complementary methods to (1) analyze the magnitude of DIF, (2) analyze whether the magnitude of DIF results in an important clinical impact, considering the instrument as a whole, and (3) use analysis based on qualitative methods to interpret the reasons for the differential functioning of the item (i.e., conduct groups with area experts) [39, 40]. Thus, further studies are needed to verify in greater detail the reasons that led parents of children and youth with different severity levels of CP to respond differently to these items.

Although most items are located in the same region as most individuals, at the extremes of both continua there are some participants who are not aligned with any item. This suggests that the difficulty of the items is either too low or too high for the ability levels of some participants and that it may be necessary to add very easy items and very difficult items in both domains. In addition, the results showed that most items in the Running and Playing area are located at the top of the continuum. It might be useful, then, to develop other items in this area with lower levels of difficulty. This addition of items could help to better identify the limitations in performance in daily activities and mobility of children and youth with CP with mild and severe levels of motor function as recently also recommended by Burgess et al. [24]. One great advantage of instruments that use the CAT system is the possibility of adding new items to the item bank, thus allowing the instrument to become more complete and appropriate as results from new validation studies may indicate [41, 42].

Most DA items located at the bottom of the item-person map refer to the use of electronic devices such as the remote control, phone, and computer. The easiest MB items were simple tasks involving wheelchair management. This is not surprising as these activities involve minimal physical effort and allow the use of compensatory strategies to manage assistive technology devices [43‐45]. Conversely, the most difficult DA items were those that required fine bimanual coordination. In this study, most participants with mild manual function limitation had spastic hemiplegic CP type, which is characterized by a unilateral upper and lower extremity involvement. This may explain the difficulty encountered by these mild severity level participants in performing such bimanual activities [46].

PEDI-CAT items for individuals who use mobility devices (such as cane, crutches, walker) and wheelchair were consistent with participants’ GMFCS levels. In this classification system, people who make use of walking devices are classified in levels III, IV, or V of gross motor function. In the item-person MB map, disability items are located throughout the continuum of difficulty, and paired with participants classified as GMFCS levels III, IV, and V.

This study demonstrated that DA and MB PEDI-CAT items are able to discriminate individuals with CP of different functional levels, as we can observe through the location of individuals and items on the two ability/difficulty continua (Fig. 2). These results are in accordance with those reported by Frazier et al. [22] and Shore et al. [23], who used CTT-based analyzes to verify the validity of PEDI-CAT in discriminating functional levels of children with CP. Analyzes conducted by those studies compared GMFCS and MACS categories according to scores obtained by the participants in each PEDI-CAT domain. However, they were not able to identify which items or functional areas within the PEDI-CAT domains are more suitable for specific CP levels. By focusing on items and individuals and their relative difficulties using Rasch analysis, our study moves the validity investigation of the PEDI-CAT forward. This study had a small number of participants, especially older ones (e.g., adolescents and young adults) and with severe and moderate motor function limitations, which may limit the generalization of our results. We also acknowledge that the decision to curb the PEDI-CAT item pool represents a limitation of the present study, that needs to be addressed in future studies (Additional file 1).

Our results provide evidence that the PEDI-CAT DA and MB domains are valid to evaluate older children, adolescents and young adults with CP. We observed that parents had a tendency to choose between the two extreme and opposing PEDI-CAT response categories. There was a floor effect in both domains, that is, the difficulty level of some items was either too low or too high to measure the CP individuals’ abilities. It is recommended to add very easy items and very difficult items to the DA and MB domains, especially easiest items in the Running and Playing content area. Only few items presented DIF, indicating that the DA and MB domains are valid to evaluate CP participants with different manual abilities and gross motor function levels. Understanding the performance profile of older children, adolescents, and young adults with CP in daily activities and mobility, by focusing on items’ difficulties and individuals’ abilities support more appropriate evaluation and interventions for this population.