The World Health Organization Disability Assessment Schedule II (WHO-DAS II) was developed by the WHO's Assessment, Classification and Epidemiology Group within the framework of the WHO/NIH Joint Project on Assessment and Classification of Disablements [1]. It is a 36-item instrument designed to measure disability irrespective of health-related etiology in six domains: understanding and communicating, getting around, self-care, getting along with people, life activities, and participation in society. These domains reflect two dimensions of the International Classification of Functioning, Disability and Health (ICF) model [2]: activity limitations (understanding and communicating; getting around; and self-care) and participation (getting along with others; life activities; and participation in society). A one-dimensional screener with 12 items was also developed for measuring global disability, which is especially suitable for epidemiological studies and routine outcome assessment.

The psychometric properties of the WHO-DAS II have been tested in approximately 28 centres in more than 18 geographically and culturally diverse countries. Previous studies, conducted with patients suffering diverse physical and mental conditions, have focused on the factor structure, internal consistency, item-total correlations, etc, using classical test theory as framework [3–10]. However, this type of analyses rely on omnibus statistics that average over levels of individual variation and offer no means to gauge the quality of individual items or options across different levels of disability. Moreover, traditional analyses are sample dependent, showing certain variations among clinical groups and even within clinical groups.

Item response theory (IRT) methods are powerful tools that provide detailed information about item functioning [11–13]. IRT comprises a group of parametric and non-parametric models expressing the probability of a particular response to a scale item as a function of the latent trait of the individual and of characteristics of the item [14]. Some advantages of IRT models with respect to traditional test methods for health outcome assessment are the next [11, 15]: (1) item statistics are independent of the individuals included in the sample and person statistics are independent of the administered items (invariance); (2) the standard error of the estimate is unique for each latent trait level; whereas the traditional test-based standard error of the measurement is assumed to be the same for all individuals regardless of their underlying trait level; (3) item information and test information vary as a function of the underlying trait level; (4) it is possible to assess differential item functioning (DIF), defined as different probabilities of endorsing an item by individuals from two groups who are equal on the latent trait. Finally, it is important to highlight that IRT and classical test theory methods have not to be conceived as rivals, in fact, as Pollard et al [16] have recently pointed out, the use of both methods is more informative than only using one of the methods.

The present work extends previous studies using IRT methods to assess the item quality of the 12-item WHO-DAS-II. First, a non parametric item analysis [17, 18] was performed to examine the effectiveness of each item and their options in discriminating between changes in the underliying disability level, among patients with a first time diagnosed major depressive episode. Our second specific objective was to evaluate WHO-DAS II items for DIF related to gender.

In the present work we utilized the ERASMAP dataset. The ERASMAP was a cross-sectional observational study carried out in Spain. It was designed to identify the sociodemographic and clinical factors associated with diagnostic delay in primary care patients with a first time diagnosed major depressive episode. A total of 1210 general practitioners from 874 healthcare centres agreed to participate in the study, which was performed in accordance with the ethical standards laid down in the 1964 Declaration of Helsinki and was approved by the Clinical Research and Ethics Committee of the University Hospital La Princesa (Madrid, Spain).

In the present work, we employed a non parametric IRT method to examine the items effectiveness, response category functioning and differential item functioning related to gender of the 12-item WHO-DAS II.

Overall, the results obtained in the present research indicate that all WHO-DAS II items performed well at discriminating varying levels of disability. The inspection of the ICCs and the OCCs indicated that all items assessed well the entire continuum of disability. We could see that certain items, for instance items 8 and 9 (bathing and dressing, respectively), ask about activities in which patients clearly experience fewer problems, whereas in others, for instance items 5 and 12 (emotionally affected by the health problem and work, respectively), patients report more problems or difficulties. Finally, the weights assigned to the individual item options are appropriate for measuring the underlying trait due to the absence of overlap between adjacent OCCs.

Following Santor et al criteria [25], we can consider the items as "good" or "very good" because there was some range of severity in which the majority of options were more likely to be endorsed than any other, the OCCs increased rapidly with changes in overall severity, the region in which each option was most likely to be endorsed were ordered, left to right, in accordance with their option scores (weights), and options spanned the full continuum of severity from -3 to +3. In addition, we did not find "easy" (the majority of options on an item are endorsed at low levels of severity) or "hard" items (the majority of options on an item are endorsed at high levels of severity). Finally, men and women who were at the same point on the disability continuum did not respond differently to items on the WHO-DAS II, that is, group mean differences between men and women in disability cannot be attributable to gender item bias.

Our results partially support those obtained by Rehm and collaborators [26] with a previous version of the instrument. Using data from two field trials, these authors examined the 12-item WHO-DAS II (screener version that possessed different items) by means of confirmatory factor analysis, non-parametric (Mokken scale analysis) and parametric (Birnbaum's two-parametric model) methods of IRT. The non-parametric analysis indicated that the scale could be considered of medium scalability, however, the parametric analysis indicated that the ICCs did intersect, suggesting the need of developing a new version of the instrument that would be IRT compatible.

We have to acknowledge the following limitation in the present work. Although the 12-item WHO-DAS II is sufficiently unidimensional for IRT analysis, we found covariation in a previous study [8] within some pairs of items (1 and 7, 8 and 9, 10 and 11; the confirmatory factor analysis indicated that the three error covariances were statistically significant, ranging from a minimum of 0.42 to a maximum of 0.83), which supposes a violation of the local independence criterion required by IRT. However, having in mind that IRT methods are quite robust to minor violations related to the local independence assumption, especially when a scale consists of 10 or more ítems [20], we think that there are not enough reasons to judge the results reported above as unreliable or non valid.

IRT methods enable researchers to examine important scale properties that can not be addressed with traditional analyses [27]. In this work, using non parametric IRT analyses, we found that WHO-DAS II items and options discriminate well among different latent levels of disablement and that it is a nonbiased instrument with respect to gender. Future studies should try to extend our findings using a parametric IRT model in order to confirm that the instrument is IRT compatible and allows cross-cultural comparability.