Use of the Pediatric Evaluation of Disability Inventory ‐ Computer Adaptive Test in Denmark, the Netherlands, and Norway

Abstract

Aim

To investigate whether the items of the Danish, Dutch, and Norwegian versions of the Pediatric Evaluation of Disability Inventory ‐ Computer Adaptive Test (PEDI‐CAT) align with the location‐order used in the original algorithm and to assess their structural validity.

Method

Three convenience samples without disability (0–20 years; Danish [n = 318], Dutch [n = 349], Norwegian [n = 362]) responded to the language‐specific versions. Item location‐order was estimated using the graded response model and structural validity was tested using confirmatory factor analysis and Rasch analysis.

Results

For most items, the item location‐order was largely consistent with the location‐order used in the original PEDI‐CAT algorithm. Items showing a different order were primarily related to the daily activity domain. However, the confirmatory factor analysis and Rasch analysis indicated poor model fit, multidimensionality, and local dependency. Additionally, the Rasch analysis revealed that some items were misfitting, with a few also showing signs of misfit in the original PEDI‐CAT version. Few items displayed differential item functioning by sex.

Interpretation

The Danish, Dutch, and Norwegian version of the PEDI‐CAT can be used to measure the degree of functioning or responsibility. However, clinicians should interpret the PEDI‐CAT results with caution due to evidence of multidimensionality, some misfit items, and differential item functioning by sex. Further research is warranted in a population of children and young people with disabilities.

Short abstract

This original article is commented by Fragala‐Pinkham on pages 1379–1380 of this issue.

Abbreviations

CFA

confirmatory factor analysis

DIF

differential item function

ICC

item characteristic curve

PEDI‐CAT

Pediatric Evaluation of Disability Inventory – Computer Adaptive Test

What this paper adds

• The algorithm of the PEDI‐CAT can be used in a Denmark, the Netherlands, and Norway.

• Results should be interpreted with caution if items showing misfit or differential item function are included.

• Further research is warranted in a larger population of children and young people with disabilities.

In the rehabilitation of children with disabilities, a key objective is to enhance their engagement in daily activities. ¹ Therefore, it is important for clinicians to assess and evaluate the functional performance of the child. The Pediatric Evaluation of Disability Inventory – Computer Adaptive Test (PEDI‐CAT) is a patient‐ (parent‐)reported outcome measure designed to detect and monitor delays in functional activity performance of children and young people (from birth – 20 years) with physical, cognitive, and/or behavioural disabilities, and can be used across diagnoses, conditions, and settings. ² The PEDI‐CAT items focus on the activity performance which refers to what the child actually does in their daily environment, and on the child's ability to perform each functional activity in a manner that is effective given their abilities and challenges. ² The PEDI‐CAT is a standardized, norm‐referenced test developed in 2011 as a revision and extension of the PEDI, ² which was found to have some limitations, such as a limited age range, and to be time‐consuming. ² , ³

The PEDI‐CAT comprises a scoring algorithm and a 276‐item bank of functional activities acquired by typically developing individuals from infancy through adolescence. The items are distributed across three domains related to the performance of daily functional skills: daily activities, mobility, and social/cognitive functioning. A fourth domain concerns responsibility and measures the extent to which level a child or adolescent (≥3 years of age) is engaged and takes responsibility in life situations (Table 1).

TABLE 1.

Overview of the Pediatric Evaluation of Disability Inventory ‐ Computer Adaptive Test (PEDI‐CAT) domains, their content, and response categories.

Domains	Content areas	Item example	Likert scoring scale
Daily activities 68 items	Getting dressed, keeping clean, home tasks, and eating and mealtime	Cuts vegetables or meat with a knife and fork	(1) Unable = Can't do, doesn't know how, or is too young (2) Hard = Does with a lot of help, extra time, or effort (3) A little hard = Does with a little help, extra time, or effort (4) Easy = Does with no help, extra time or effort, or child's skills are past this level (5) I don't know
Mobility 75 items	Basic movement and transfers, standing and walking, steps and inclines, and running and playing	Walks and carries a food tray
Social/cognitive 60 items	Interaction, communication, everyday cognition, and self‐management	Teaches another person a new game or activity by giving examples and explanations
Responsibility 51 items	Organization and planning, taking care of daily needs, health management, and staying safe	Managing kitchen appliances such as stove, microwave, or dishwasher safely	(1) Adult/caregiver has full responsibility; the child does not take any responsibility (2) Adult/caregiver has most responsibility and child takes a little responsibility (3) Adult/caregiver and child share responsibility about equally (4) Child has most responsibility with a little direction, supervision, or guidance from an adult/caregiver (5) Child takes full responsibility without any direction, supervision, or guidance from an adult/caregiver

The PEDI‐CAT was developed in the USA, and the item bank has been calibrated in a sample of 2908 American children and young people with and without disabilities. The PEDI‐CAT applies a statistical algorithm that personalizes the assessment by selecting relevant items to be administered on the basis of responses to previous items. ² This reduces the assessment‐burden and provides more precise scores than the original PEDI. ² , ⁴

For each domain the responses are summarized into scaled scores and normative T‐scores. In addition, the PEDI‐CAT provides an item map for each content area within each domain, where all items are presented in a hierarchical order according to their level of difficulty.

The psychometric properties of the PEDI‐CAT and its translations have been investigated, ² , ³ , ⁴ , ⁵ , ⁶ , ⁷ and have been explored in different diagnostic groups. ⁸ , ⁹ , ¹⁰ , ¹¹ , ¹² Since 2015, researchers in Denmark, the Netherlands, and Norway have collaborated on country‐specific translations and implementation processes of the PEDI‐CAT. When implementing translated versions of a CAT, such as PEDI‐CAT, the item location‐ordering must align with the original order, and the psychometric properties should be acceptable. ¹³ This has not been addressed for the three translated versions. It is essential to ensure that a translated instrument is adapted to the specific language and cultural context without altering its meaning. ¹⁴ When cross‐cultural validity can be established, comparisons across diverse populations can be trusted, and equitable health care practices implemented.

The overall aim of the study was therefore to investigate whether the algorithm in the PEDI‐CAT software could be applied in Denmark, the Netherlands, and Norway without adjustments. This was undertaken by addressing two aspects of cross‐cultural validity: (1) invariance of item locations across countries (i.e. whether the US‐developed algorithm worked as intended for the translated versions) and (2) structural validity (i.e. whether the translated versions measured what they are supposed to measure). ¹⁴ , ¹⁵ The specific aims were to: (1) examine whether the item location‐order of the Danish, Dutch, and Norwegian translations of the PEDI‐CAT aligned with the order presented in item maps for each PEDI‐CAT domain for the US sample, and (2) examine the structural validity of the three language versions of the PEDI‐CAT.

METHOD

Design

This was a methodological study assessing: (1) the item difficulty levels and (2) structural validity in the Danish, Dutch, and Norwegian versions of the PEDI‐CAT.

Translating procedure

The translation procedures for the three language versions followed a comparable forward–backward translation procedure ⁷ , ¹⁴ with the involvement of representatives from the respondent group and experts from the paediatric field (see Figures [Link], [Link], [Link] for details).

Recruitment

All three countries used convenience sampling. A national network of health care professionals and researchers was asked to distribute information about the PEDI‐CAT study to colleagues, friends, and family who were parents of typically developing children and young people (Denmark and Norway, 0–20 years; the Netherlands, 1–20 years). In the Danish and the Dutch recruitment, an inclusion criterion was that the participants be able to read and speak the language. The Netherlands and Norway also used social media, schools, and sports clubs for recruitment. In Norway, young people older than 16 years were contacted directly for self‐reports.

Data collection

Data collection took place between March 2016 and June 2021. In Denmark and Norway, all items of the PEDI‐CAT were administered as a paper/pen questionnaire and as an electronic survey respectively. In the Netherlands, data were collected through an electronic survey with 12 age‐blocks consistent with the PEDI‐CAT manual. The blocks included one‐third of all items from the original item bank according to the age of the child with overlapping items across blocks. In all three countries, the responsibility domain was only administered for children aged 3 years and older. All respondents provided information on birth date, sex, and area of residence for whom they responded, and parents additionally reported on their educational levels.

Ethics

The study was approved by the Danish Data Protection Agency (number 212–58‐0004) and the Data Protection Office at Oslo University Hospital (19/06525). In the Netherlands, the Medical Ethics Committee of the University Medical Center Utrecht declared no approval was necessary (17/496). All respondents were informed that participation was voluntary, that they could withdraw from the study at any stage and provided written informed consent.

Data analysis

Data from Denmark and Norway were adapted to match the selection of items according to the age‐blocks used in the Netherlands. The distribution of item responses was reported as frequencies and proportions. Items where all respondents used the highest response options and items where all respondents used lowest response were omitted from the analysis because they did not provide any information in the applied analyses. The response category ‘I don't know’ in daily activity, mobility, and social/cognitive was handled as missing data in the analysis.

In accordance with the procedures applied during the development of the PEDI‐CAT, ² , ⁴ item difficulties were estimated using the graded response model, ¹⁶ and psychometric properties were examined using confirmatory factor analysis (CFA) ¹⁷ and Rasch analysis. ¹⁵ , ¹⁸ These three methods are related applications of latent variable models. For all three the latent variable is the construct measured by the domain and for all three the observed PEDI‐CAT items are indicators of the value of this latent variable. What differs is the parametric model that specifies this relationship and the utility of each model in evaluating the usefulness of the PEDI‐CAT items and scoring algorithm. The graded response model tests the utility of the scoring algorithm, ¹⁶ CFA tests structural validity, ¹⁷ and Rasch analysis tests a range of measurement properties outlined below. ¹⁵ The graded response model can also be used to test validity.

Item difficulty by graded response model

Item parameters were estimated separately for each of the three language versions using the graded response model in SAS PROC IRT (SAS Institute, Cary, NC, USA), where the slope parameters were held constant. Invariance of item location‐order across countries was evaluated visually using item maps where the location estimates were plotted against the original US item location‐ordering reported in the PEDI‐CAT manual. ²

Unidimensionality by CFA

CFA assessed consistency of PEDI‐CAT measures within each domain mean square residual, root mean square error of approximation with a 95% confidence interval (CI), and comparative fit index. Model fit criteria were mean square residual less than 0.08, root mean square error of approximation less than 0.06, and comparative fit index greater than 0.95. ¹⁸ Full information maximum likelihood in SAS PROC CALIS was used for the CFA.

Rasch analysis

The analysis was conducted for each domain separately using RUMM2030 Plus. ¹⁹ The Rasch model analysis followed the recommendation of the Rasch Reporting Guideline for Rehabilitation Research (RULER). ²⁰ Respondents were categorized according to location into class intervals of approximately equal size with at least 50 in each group. ²¹ Observed and expected item scores in each of these intervals were compared.

Overall fit to the Rasch model was examined using χ ² interaction tests, with insignificant values indicating model fit. The overall fit of items and persons was assessed by examining fit residuals and reported as mean (standard deviation [SD]) for each domain, where values approaching 0.00 (1.0) indicate a good model fit. ²⁰ The reliability of PEDI‐CAT domains was assessed using the person separation index (values greater than 0.70 are considered acceptable). ²² Local dependency was indicated by item residual correlations greater than 0.2. Individual item fit was evaluated using: (1) item χ ² tests (insignificant tests indicate model fit), (2) item fit residual values (values between −2.5 and 2.5 indicate model fit), and (3) visual inspection of the item characteristic curve (ICC). Bonferroni adjustment was used to take multiple testing into account. Differential item functioning (DIF) ²³ is reported in accordance with the RULER guidelines ²⁰ when the distribution of the respondents in the subgroups was even and the sample size sufficient. DIF was assessed through two‐way analysis of variance tests of the residuals, with a statistically significant p‐value indicating item DIF, and through visual inspection of the ICC.

RESULTS

The cohort included 318 Danish, 349 Dutch, and 362 Norwegian PEDI‐CAT respondents. The mean age of the children was between 9 years 1 month and 9 years 7 months and the sexes were equally distributed (male, 47.0%–50.9%). The education level of the parents was unevenly distributed (Table 2).

TABLE 2.

Overview of the demographic data of the respondents.

	Denmark	the Netherlands	Norway
Number of respondents	318	349	362 a
Non‐immigrant respondents (%)	N/A	96.1	98.2
Sex of child, % (n)
Male	50.9 (162)	47 (164)	47.5 (172)
Female	49.1 (156)	53 (185)	52.5 (190)
Age of child, years:months, mean (SD)	9:7 (5:8)	9:1 (5:10)	9:1 (5:10)
Educational level of parents, % (n)
Lower	0.9 (3) b	1.7 (6)	0.7 (2) c
Intermediate	55.4 (179) b	20.4 (71)	6.5 (20) c
Higher	43.7 (141) b	77.9 (272)	92.8 (285) c

Note: Parental highest educational level according to ISCED: lower (ISCED 0–2), intermediate (ISCED 3–4), higher (ISCED 5–8).

Abbreviations: ISCED, International Standard Classification of Education; N/A, not applicable—no data on country of origin was obtained; SD, standard deviation.

^a 121 of the respondents older than 16 years are self‐reported.

^b Highest educational level of parents.

^c Only data from parent respondents.

For the domain of daily activities and mobility, extreme items were present in all the samples (Denmark, four; the Netherlands, eight; Norway, one; see Figures [Link], [Link], [Link]), and these were removed from the Rasch analysis. Individuals having extreme items with ceiling effects were present in all domains across all language versions, while the responses in the responsibility domain were well distributed (Figures [Link], [Link], [Link]).

Item locations

Figures [Link], [Link], [Link], [Link] include item location plots for the three language versions. In Figure 1, we present examples: one aligns with the original PEDI‐CAT item map for all language versions (Figure 1a) while the other does not (Figure 1b).

FIGURE 1.

Examples of item location‐order for the three language versions. For the item location‐order (y‐axis), the easiest items are on the bottom row, while the hardest ones are on the top. The placement of the coloured lines for each language version on the x‐axis illustrates the item difficulty, with the easiest items on the left and the hardest item on the right. (a) Item location‐orders for the content area ‘organisation and planning’ in the responsibility domain, showing items that align the expected item location. (b) Item location‐orders for the content area ‘home tasks’ in the daily activities domain, showing items that do not align the expected item location.

The item location was generally consistent with the PEDI‐CAT manual across content areas, except in the daily activity domain. Specifically, discrepancies were observed in the ‘home tasks’ content area for all language versions, with additional variations in specific content areas for the Dutch and Norwegian versions. In the mobility domain, differences were noted for one item across all language versions in the ‘steps and inclines’ and ‘running and playing’ content areas. Additionally, one item in the social/cognitive domain differed across all language versions.

Unidimensionality

Table 3 shows CFA model fit for the content area for each domain in each of the three language versions. Overall, the model fit was poor with indications of multidimensionality and modification indices pointing to evidence of local dependence. CFA models with correlated error terms showed better model fit to the data (Table S1).

TABLE 3.

Overview of the confirmatory factor analysis for content areas within each domain for all three language versions.

Domain	Subdomain	Sample	SRMR	RMSEA	95% CI		CFI
Daily activities	Home tasks	Denmark	0.194	0.086	0.075	0.097	0.912
		the Netherlands	0.315	0.212	0.181	0.246	0.785
		Norway	0.106	0.073	0.063	0.083	0.935
	Keeping clean	Denmark	0.294	0.326	0.294	0.360	0.799
		the Netherlands	0.321	0.088	0.056	0.123	0.979
		Norway	0.417	0.164	0.149	0.180	0.830
	Getting dressed	Denmark	0.672	0.136	0.128	0.145	0.833
		the Netherlands	0.377	0.127	0.112	0.144	0.943
		Norway	0.431	0.165	0.155	0.175	0.833
	Eating and mealtime	Denmark	0.256	0.142	0.131	0.153	0.873
		the Netherlands	0.209	0.130	0.109	0.151	0.931
		Norway	0.134	0.075	0.062	0.088	0.964
Mobility	Running and playing	Denmark	0.256	0.142	0.131	0.153	0.873
		the Netherlands	0.209	0.130	0.109	0.151	0.931
		Norway	0.134	0.075	0.062	0.088	0.964
	Steps and inclines	Denmark	0.256	0.142	0.131	0.153	0.873
		the Netherlands	0.209	0.130	0.109	0.151	0.931
		Norway	0.134	0.075	0.062	0.088	0.964
	Standing and walking	Denmark	0.256	0.142	0.131	0.153	0.873
		the Netherlands	0.209	0.130	0.109	0.151	0.931
		Norway	0.134	0.075	0.062	0.088	0.964
	Basic movement and transfers	Denmark	0.256	0.142	0.131	0.153	0.873
		the Netherlands	0.209	0.130	0.109	0.151	0.931
		Norway	0.134	0.075	0.062	0.088	0.964
Social/cognitive	Self‐management	Denmark	0.133	0.066	0.037	0.097	0.980
		the Netherlands	0.073	0.096	0.049	0.150	0.964
		Norway	0.124	0.093	0.068	0.120	0.940
	Interaction	Denmark	0.226	0.139	0.124	0.154	0.907
		the Netherlands	0.432	0.116	0.100	0.133	0.907
		Norway	0.158	0.110	0.096	0.124	0.922
	Communication	Denmark	0.360	0.156	0.138	0.175	0.864
		the Netherlands	0.110	0.130	0.105	0.157	0.953
		Norway	0.508	0.167	0.152	0.182	0.803
	Everyday cognition	Denmark	0.471	0.199	0.187	0.211	0.816
		the Netherlands	0.882	0.179	0.168	0.189	0.828
		Norway	0.544	0.180	0.170	0.189	0.824
Responsibility	Organization and planning	Denmark	0.097	0.203	0.194	0.211	0.771
		the Netherlands	0.099	0.177	0.169	0.185	0.810
		Norway	0.091	0.245	0.238	0.253	0.726
	Health management	Denmark	0.031	0.154	0.131	0.177	0.949
		the Netherlands	0.104	0.094	0.070	0.118	0.974
		Norway	0.022	0.136	0.114	0.159	0.967
	Taking care of daily needs	Denmark	0.091	0.176	0.165	0.187	0.836
		the Netherlands	0.085	0.154	0.143	0.165	0.863
		Norway	0.091	0.208	0.198	0.218	0.800
	Staying safe	Denmark	0.075	0.221	0.199	0.245	0.868
		the Netherlands	0.068	0.179	0.157	0.202	0.920
		Norway	0.056	0.198	0.176	0.220	0.904
	Fit criteria		≤0.08	≤0.06			≥0.95

Abbreviations: CFI, comparative fit index; CI, confidence interval; RMSEA, root mean square error of approximation; SRMR, standardized root mean square residual.

Item analysis

Table 4 presents overall fit statistics for the four domains across language versions. Ceiling effects were observed in 1.4% to 33% of respondents, with only the responsibility domain meeting the criterion of not more than 15%. Significant misfits in item–trait interaction were noted for all domains and language versions. The social/cognitive and responsibility domains exhibited misfits, supported by numerous misfitting items in these domains across language versions (Appendix [Link], [Link], [Link]). Person fit residuals were acceptable for all domains except responsibility in the Netherlands. Local dependency occurred for several item pairs across all domains and language versions. The person separation index indicated good discriminative ability (0.93–0.99) in all domains across language versions. Detailed item fit statistics are available in Tables [Link], [Link], [Link], revealing 16 misfitting items (2% of total) across language versions. Item DA060 (puts on a T‐shirt) in the Norwegian version was the only item with a significant negative fit residual, suggesting over‐discrimination, possibly indicating redundancy due to local dependency (Figure 2a). All other misfits showed significant positive fit residuals (>2.5, illustrated in Figure 2b), indicating under‐discrimination and suggesting potential multidimensionality.

TABLE 4.

Overall fit and item‐level fit to the Rasch model for the four domains and language versions.

Overall fit statistics									Item‐level fit statistics
	Number of respondents	Number of items	Respondents at extreme, n (%)	Items with <10 responses in each category, n (%)	Fit residuals, mean (SD)		Item–trait interaction, χ 2 (df), p	Person separation index, with extremes/excluding extremes	Misfitting items, statistic and visual in ICC	DIF items (uniform/non‐uniform)	Local dependency, n/total (number of empty item pairs)
					Item	Person
Daily activities
Denmark the Netherlands Norway	318 349 362	67 65 66	69 (21.7) 95 (27.2) 62 (17.1)	36 (53.7) 15 (23.1) 23 (68.1)	−0.05 (1.17) −0.31 (1.16) −0.21 (1.48)	−0.04 (0.45) −0.13 (0.77) −0.07 (0.72)	722.57 (268), <0.001 431.12 (260), <0.001 602.55 (264), <0.001	0.67/0.99 0.93/0.97 0.96/0.98	DA044, DA047 DA002 DA011, DA060, DA097	DA092/0 0/0 DA084/0	125/2045 (169) 197/1953 (332) 71/1954 (191)
Mobility
Denmark the Netherlands Norway	318 346 362	72 68 75	108 (33.0) 114 (32.9) 101 (27.9)	13 (18.1) 5 (7.4) 24 (30.7)	−0.24 (1.09) −0.43 (1.11) −0.29 (1.29)	−026 (0.90) −0.31 (0.79) −0.24 (0.70)	769.65 (287), <0.001 543.71 (299), <0.001 475.11 (272), <0.001	0.92/0.97 0.86/0.93 0.92/0.97	MB069 0 MB078	0/0 0/0 0/0	148/2269 (281) 168/2016 (371) 161/1645 (367)
Social/cognitive
Denmark the Netherlands Norway	318 339 362	60 60 60	73 (22.9) 74 (21.8) 66 (18.2)	27 (45) 21 (35.0) 33 (55.0)	−0.40 (1.49) −0.55 (1.40) −0.68 (1.74)	−0.30 (0.86) −0.39 (0.88) −0.34 (0.81)	497.58 (240), <0.001 435.36 (240), <0.001 490.31 (240), <0.001	0.93/0.97 0.93/0.97 0.95/0.98	SC057 SC002 SC014, SC057	0/0 0/0 0/0	74/1592 (178) 130/1770 (203) 100/1593 (177)
Responsibility
Denmark the Netherlands Norway	280 295 311	50 51 50	4 (1.4) 7 (2.4) 21(6.8)	42 (84.0) 43 (84.3) 38 (76.0)	−0.36 (1.61) −0.16 (2.14) −0.13 (1.64)	−0.26 (0.85) −0.27 (1.24) −0.19 (0.78)	587.79 (200), <0.001 413.14 (204), <0.001 389.05 (204), <0.001	0.99/0.99 0.98/0.99 0.98/0.99	0 RS014, RS020, RS022 RS022, RS041	RS015, RS048/0 0/0 RS049/0	42/1225 (0) 67/1275 (1) 46/1225 (0)
Fit criteria			<15%		0.0 (≤1.00)		p > 0.05	≥0.70	0	0	0

Note: Misfitting items in bold type indicate a visual misfit according to the RUMM manual. ²¹ All other misfitting items are in Tables [Link], [Link], [Link].

Only DIF by sex that fulfils the criteria of the RULER are presented. ²⁰ All other DIF by sex are in appendix [Link], [Link], [Link].

Abbreviations: df, degrees of freedom; DIF, differential item functioning; ICC, item characteristic curve; RULER, Rasch Reporting Guideline for Rehabilitation Research; SD, standard deviation.

FIGURE 2.

Item characteristic curves (ICCs) of two misfitting items of the Pediatric Evaluation of Disability Inventory ‐ Computer Adaptive Test (PEDI‐CAT). The curved line represents the expected score for the item and the dots represent the observed scores for the fire class intervals. (a) ICC plot for item DA060 (Norwegian sample) shows a high negative significant fit residual (−3.033). (b) ICC plot for item SC057 (Danish sample) represents items with a high positive and significant fit residual (4.864).

DIF

Using the RULER guidelines related to DIF analysis, ²⁰ it was only possible to investigate DIF by sex for some items owing to an uneven distribution of the respondents in the subgroups. Evidence of uniform DIF by sex (illustrated in Figure 3) was seen in five items (Denmark, three; Norway, two).

FIGURE 3.

Differential item functioning graph for sex illustrated by item RS048 (Norwegian sample). The red line represents female and the blue line male respondents according to the class intervals.

Table 5 shows an overview of items with an item location that differed from the original item order, misfit, or DIF by sex.

TABLE 5.

Overview of items with a different item location then the original item location‐order, item misfit (including the USA data), or DIF by sex in the language versions.

Item	Item location	Misfit	DIF by Gender
Daily activities
DA002 (Swallows pureed/ blended/ strained foods)		NL and USA
DA006 (Holds/eats Sandwich/burger)	NL and NO
DA031 (Turns water on/off at sink)	NO
DA044 (Shaves face, male only)		DK and USA
DA047 (Fastens necklace/chain)		DK
DA060 (Puts on a T‐shirt)		NO
DA075 (Puts on tights/pantyhose, female only)	NO
DA084 (Operates video game controller)	DK, NL, and NO		NO
DA086 (Uses computer mouse)	DK, NL, and NO
DA087 (Uses computer keyboard)	DK, NL, and NO
DA092 (Opens door lock using key)			DK
DA097 (Puts a bandage on a small cut on hand)		NO
DA100 (Removes bill from wallet)	DK, NL, and NO

Mobility
MB067 (Rides bicycle)	DK, NL, and NO
MB069 (Gets on/off bus)	DK, NL, and NO	DK
MB078 (Walks down a flight of stairs with handrail)		NO

Social/cognitive
SC002 (Uses several words/signs together)		NL
SC004 (Uses words or signs to ask questions)	DK, NL, and NO
SC014 (Uses appropriate language)		NL and USA
SC057 (When upset, responds appropriate)		DK, NO and USA

Responsibility
RS014 (Fixing simple meals)		NL
RS015 (Following a recipe)			DK
RS020 (Maintaining cleanliness of living space)		NL
RS022 (Putting items away after use)		NL and NO
RS041 (Locating services or supports)		NO
RS048 (Packing items for overnight stay)		USA	DK, NO

Abbreviations: DK, Denmark; NL, the Netherlands; NO, Norway; USA, The United state of America.

DISCUSSION

This study aimed to assess whether the item location‐order of three translated versions of the PEDI‐CAT aligns with the order in the original US version. ² Generally, the item location‐ordering was consistent with the original PEDI‐CAT. The study also investigated the structural validity of the three language versions, finding that the fit to both the Rasch and CFA models was not satisfactory, with evidence of multidimensionality, local dependency, some misfitting items, and DIF by sex. Despite these issues, reliability estimates showed that all domains could still discriminate individuals according to their functional ability, making the tool useful for research and clinical work.

Compared with the original PEDI‐CAT algorithm, the item location‐order in our study revealed lower difficulty levels for items related to computer and gaming, probably because of changes in technology since the development of the PEDI‐CAT. The item ‘removes bill from wallet’ was more challenging, possibly because of increased use of electronic payments. In all three language versions, the mobility domain items (rides bicycle, and getting on/off buses) were less difficult, suggesting cultural differences in transportation use. This observation suggests potential cultural differences influencing item difficulty, such as the common use of bicycles for transportation particularly in Denmark and the Netherlands compared with the USA. ²⁴ The social/cognitive and responsibility domains remained consistent with the original item locations. Notably, this study is the first to examine the item location of these two domains in languages other than English, highlighting the necessity for further research to confirm these findings.

Most items exhibiting different location‐order, misfit, or DIF by sex were in the daily activities domain. This domain has shown misfit in various diagnostic groups in earlier studies ⁷ , ⁸ revealing misfit among items, but most differ from those identified in the current investigation. This observation may suggest a potential cultural difference influencing the interpretation of the items, warranting further investigation. It is worth noting that the original PEDI‐CAT includes misfitting items (n = 25) across the four domains. These items were retained in the original item bank to avoid increasing the floor or ceiling effect, which could threaten the content validity. ² As depicted in Table 5, our study replicated misfit in 19% of these items, suggesting a need for further investigation or revisions.

Respondents demonstrated a tendency to underutilize certain response categories. Across the domains, the criterion of having at least 10 responses per category, as required for Rasch analysis, was met for only 7.4% to 84.3% of the items. This underutilization was most pronounced in the mobility domain. Generally, responses clustered at the extremes of the scale (i.e. ‘unable’ or ‘easy’), with fewer responses in the middle categories (‘hard’ or ‘a little hard’) (Tables [Link], [Link], [Link]). Unfortunately, the distribution of responses from the US normative sample is not reported in the manual ² or the publication on the development of the PEDI‐CAT. ³ Previous studies applying Rasch analysis to data from children with disabilities ⁸ , ¹² also observed a concentration of responses at the scale extremes, suggesting that the ‘hard’ and ‘a little hard’ response categories should be collapsed.

The CFA across all three language versions indicated multidimensionality, contrary to the unidimensionality found in the original PEDI‐CAT. This might be due to translation issues, requiring further investigation to ensure each domain measures a single construct. The current analyses also identified evidence of local dependency, suggesting that items are correlated beyond the measured construct. In the development of PEDI‐CAT, Haley et al. ² eliminated items exhibiting local dependency before the final calibration. However, a direct comparison of our results with those of Haley et al. is challenging owing to differences in analysis methodology. In our study, local dependency was observed for several item pairs, which may be attributed to the nature of PEDI‐CAT assessing functional performance throughout childhood and adolescence.

For clinical practice, the original PEDI‐CAT algorithm can be used in Denmark, the Netherlands, and Norway but caution is needed when interpreting results related to problematical items (Table 5). If any problematical items are selected during an assessment, the reliability of the score may be compromised, so it is advisable to inform parents of this limitation beforehand. When planning an intervention following a PEDI‐CAT assessment, the item map is helpful. These maps display items by domain/content area and include a vertical line indicating the child's scaled score with a confidence interval. Items in the shaded area represent what the child is expected to achieve. Clinicians can use this information to set individualized goals with the children and their parents. The total scores from the domains can be used to monitor the child's progress and evaluate intervention goals.

This is the first non‐US study examining the hierarchical order of item locations and structural validity across all PEDI‐CAT domains. The current study examined DIF by sex and found that only a few items exhibited DIF, but this was not possible for all items owing to an uneven distribution and small sample size. A larger sample size with at least 100 participants per subgroup is needed to obtain an accurate result in a DIF analysis. ²⁰ However, analysing DIF by age was not possible owing to the age‐blocks used in the data collection procedure in the Netherlands and the subsequent data cleaning procedures for the other countries. Scale improvement strategies, such as removing misfitting items, were not applied to avoid compromising content validity.

The most recent PEDI‐CAT manual describes the use of numerous variations of Rasch and item response theory models. ² The Rasch model is the most parsimonious member of the class of statistical models referred to as item response theory models. We chose this model for some of the analyses reported because it is readily applicable for the magnitude of sample sizes in the present study, and because implementations that work well with incomplete data are available. However, for other analyses we used another item response theory model. The presence of a discrimination parameter in more flexible item response theory models means that they will fit observed data better. However, the difference in item locations identified is unlikely to be an artefact because of our use of a parsimonious model.

The study had several limitations, including a high percentage of extreme responses, a small sample size, and the absence of a clinical population. While a larger sample size could yield more responses, it may not provide additional meaningful data owing to the skewness caused by the absence of a clinical population. Furthermore, the respondents were predominantly non‐immigrants with higher levels of education. Given the demographic changes in European populations, ²⁵ it is important to note that immigrant groups are diverse. They originate from various countries and differ in region, ethnicity, culture, education, socioeconomic status, and length of stay in Europe. ²⁵

Cross‐cultural validation and DIF analysis by language of data collected in the USA and the translated European versions are warranted. We recommend repeating the survey using all the items from the three European studies with larger samples that include children with disabilities and immigrant families to examine DIF related to these factors. Finally, future development could include age‐related algorithms and removal of items with DIF or misfit. Research is needed to assess the clinical utility of the PEDI‐CAT among different age groups and different groups of health care professionals and clinicians.

CONCLUSION

The item location‐ordering in the Danish, Dutch, and Norwegian language version of the PEDI‐CAT was largely consistent with the original US version. Structural validity analysis indicated acceptable reliability, evidence of multidimensionality, some misfit items, and DIF by sex. Despite these issues, the items of the different language versions can still be used to measure functioning or responsibility in children and young people, but results should be interpreted with caution. Further research is warranted.

FUNDING INFORMATION

This research received no specific grant from any funding agency in the public, commercial, or not‐for‐profit sectors.

CONFLICT OF INTEREST STATEMENT

The authors have stated that they had no interests that might be perceived as posing a conflict or bias.

Supporting information

Figure S1: Flowchart of the Danish translation procedure.

Figure S2: Flowchart of the Dutch translation procedure.

Figure S3: Flowchart of the Norwegian translation procedure.

Figure S4: Targeting the Danish sample.

Figure S5: Targeting the Dutch sample.

Figure S6: Targeting the Norwegian sample.

Figure S7: Daily activities.

Figure S8: Mobility.

Figure S9: Social/cognitive.

Figure S10: Responsibility.

Table S1: CFA with correlated error terms.

Table S2: Item fit statistics for the Danish sample.

Table S3: Item fit statistics for the Dutch sample.

Table S4: Item fit statistics for the Norwegian sample.

Appendix S1: Danish sample: Item misfit and DIF by sex.

Appendix S2: Dutch sample: Item misfit and DIF by sex.

Appendix S3: Norwegian sample: Item misfit and DIF by sex.

Hansen TN, Christensen KB, Stahlhut M, Ketelaar M, Klevberg GL, Laursen LB, et al. Use of the Pediatric Evaluation of Disability Inventory ‐ Computer Adaptive Test in Denmark, the Netherlands, and Norway. Dev Med Child Neurol. 2025;67:1482–1493. 10.1111/dmcn.16324

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author upon reasonable request.