To start with, we decided to base the selection of domains represented in WORQ on the Comprehensive ICF Core Set for vocational rehabilitation, because this core set represents the most comprehensive and agreed on list of domains that are relevant to assess functioning in vocational rehabilitation. However, we also realized that the Comprehensive ICF Core Set with its 90 categories would lead to a very long questionnaire with many questions relevant to only a small proportion of clients in vocational rehabilitation and thus it would be totally impractical to use in clinical practice. Therefore, we based our selection on the 13 categories from the Brief ICF Core Set for vocational rehabilitation. This core set contains the least number of categories that are essential to describe the typical spectrum of limitations in functioning and in the environment. As this selection was not sufficient to create a satisfactory functioning profile of the clients, we decided to select additional categories from the Comprehensive ICF Core Set. The selection was based on two additional sources of information – results of a statistical analysis of the Comprehensive ICF Core Set for vocational rehabilitation and the literature.

First we used a statistical approach by conducting an explorative Rasch analysis on the whole Comprehensive ICF Core Set for vocational rehabilitation [35, 36]. We used a data set with 152 participants from a previous study [19].

See additional information: Frame 23.1: The Rasch Model.

Based on this explorative Rasch analysis, 28 ICF categories from Body Functions and Activities and Participation were added to the categories of the Brief ICF Core Set. As the scope of the questionnaire was to assess functioning by providing a functioning profile, no further environmental factors were added. The 28 categories were chosen using the following criteria: 13 categories that fit the Rasch model and showed no local dependency with other categories were directly selected. Seven more categories that fit the Rasch model and where local dependencies with other categories could be eliminated were also added. Since a study based on expert consensus gave higher weight to Activities and Participation within the context of work disability, the rule was that when a category from Functions showed local dependency with a category from Activities and Participation, the Activities and Participation category was selected [17]. Another eight categories were added for further testing if (1) two categories from the same component, either Body Functions or Activities and Participation, showed local dependency; (2) item fit showed no preference for one category; and (3) a meaningful clinical decision based on evidence from the literature could not be made. Finally, in addition to the 28 statistically selected categories, we added another three categories, b280 Sensation of pain, b730 Muscle power functions, and d430 Lifting and carrying objects. Although these categories did not fit the Rasch model, they were included because they were judged to be important for gaining a comprehensive understanding of functioning in vocational rehabilitation based on the literature and clinical expert opinion [37, 38].

This process ended with a pool of 44 ICF categories; 13 belonged to the Brief ICF Core Set for vocational rehabilitation, 28 were identified from statistical testing (Rasch analysis), and 3 were identified from the literature. Four of the categories came from the ICF component Environmental Factors, 15 from Body Functions, and 25 from Activities and Participation.

Based on the selection process described above, it could be presumed that WORQ represents a multidimensional scale. As a consequence it would not be appropriate to compare subjects or groups using item sum scores of this first version of WORQ [39].

When formulating questions, it is essential to consider that a questionnaire is only as good as the questions it contains. Studies have shown that respondents are executing four steps when answering questions. First, they have to understand and interpret the question. Second, they must search their memories for relevant information, and third, integrate all the information. Finally, they must translate their judgment into a response, by selecting one of the response options offered by the question [34, 39].