This cross-sectional, descriptive study was conducted in Saudi Arabia across various healthcare levels, including primary, secondary, and tertiary centers. Participants were recruited from outpatient clinics, rehabilitation units, and community health programs. A total of 486 individuals with disabilities participated. Inclusion criteria were: (1) self-identified or clinically documented physical, sensory, intellectual, or mental disability, and (2) ability to provide informed consent and complete the questionnaire independently. Exclusion criteria included acute illness or severe cognitive impairment that could hinder questionnaire completion [13]. Data collection for this study was conducted between September 2024 and January 2025, ensuring coverage within an appropriate timeframe for cross-sectional analysis. In addition to administering the WHODAS 2.0 questionnaire, demographic and social information was also collected from participants, including gender, age, education level, marital status, main work status, residence (urban/rural), and living situation. These variables were analyzed to examine group differences in disability scores and are presented in Table 1.

The WHODAS 2.0, an internationally recognized measure of health and social disability developed by the WHO [14, 15]. It balances the depth of information provided about an individual’s functionality and disability with the ease of administration. The tool exists in two variants: one with 36 items that offers detailed evaluation, and the other with 12 items intended for rapid assessment. Respondents answer each item using a 5-point Likert scale from “none” to “extreme or cannot do”, and scoring can be done through simple summation or more intricate methods based on item-response theory, as WHO suggests. WHODAS 2.0 has a broad scope that covers clinical evaluations, health surveys, and research studies to assess disability, monitor treatment results, and policy evaluation. Because it can be administered in multiple languages via self-report, interviewer-led, and proxy methods, WHODAS 2.0 is flexible and enhances both clinical and research [16, 17, 18, 19].

The translation of WHODAS 2.0 into Arabic followed a standardized forward-backward translation process in accordance with the WHO guidelines for cross-cultural adaptation of instruments [20]. Initially, two independent bilingual translators performed forward translations from English to Arabic. The translations were reconciled into a single version through expert consensus. This preliminary Arabic version was back-translated into English by three independent translators who were unfamiliar with the original scale. Discrepancies were discussed and resolved by a committee consisting of language and subject-matter experts. The final version was pilot-tested on a small sample to ensure clarity, cultural relevance, and conceptual equivalence to the original instrument.

Rasch analysis was conducted using WINSTEPS software (version 5.5.3), developed by John M. Linacre (Winsteps.com, Beaverton, OR, USA), to evaluate the psychometric properties of the Arabic version of WHODAS 2.0, including item fit, scale functioning, and measurement reliability [21, 22, 23]. The Rasch rating scale model was adopted to assess item fit, personal and item reliability, separation indices, scale dimensionality, and category functioning. Item and person fit statistics were evaluated using infit and outfit mean squares (MNSQ) and standardized z-values (ZSTD). The acceptable MNSQ values ranged between 0.5 and 1.5. Reliability and separation indices were obtained to assess the internal consistency and discrimination ability of the items and participants. Unidimensionality was tested through principal component analysis (PCA) of residuals, with the variance explained by the Rasch dimension expected to exceed 50% and the first contrast eigenvalue ideally below 2.0. Additionally, mean person measure and mean item measure were calculated to determine the relative positioning of item difficulty and participant ability along the logit scale, providing insights into scale targeting. DIF was also examined to detect potential item bias. Finally, category probability curves were reviewed to ensure that response options functioned in an ordered and interpretable manner. Items demonstrating a poor fit or disordered thresholds were flagged for further review.

Statistical analyses were performed using IBM SPSS Statistics (version 26), developed by IBM Corporation (Armonk, NY, USA). Descriptive statistics, such as means, standard deviations, percentages, and frequencies, were employed to summarize the participants’ demographic characteristics and WHODAS 2.0 scores, providing an overview of the sample distribution and scale responses. Normality of the data was assessed using the Shapiro–Wilk test. Due to violations of normality assumptions in variables, non-parametric tests were employed. The Mann–Whitney U test was used to examine differences in WHODAS scores between dichotomous variables, such as gender (male vs. female), residence (rural vs. urban), and living situation (alone vs. with family). The Kruskal–Wallis test assessed differences across multi-category variables, including age group, education level, marital status, and main work status. A significance threshold of p 0.05 was applied to determine statistical significance across all inferential analyses.

The demographic summary shows that the majority of participants were female (68.7%), had completed school or pre-university education (39.3%), and were married (43.2%) (Table 1). Most respondents were housewives (39.5%), indicating the predominance of non-working individuals. The distribution of urban (50.6%) and rural (49.4%) residences was nearly equal. Additionally, a vast majority (94.7%) lived with their families. These findings reflect a sample largely composed of married, non-working women with a basic education living in family households across both urban and rural areas. The age distribution revealed that the largest group of participants was 40–65 years (38.5%), followed by 24–39 years (26.3%). Participants over 65 years of age comprised 17.1% of the sample, while those under 18 accounted for 13.6%. The smallest group was 18–24 years of age (4.5%). This indicates that the sample consisted primarily of middle-aged and older adults, with fewer young adults and adolescents.

The Kruskal–Wallis test revealed several significant differences in WHODAS 2.0 scores across demographic groups. The ANOVA results showed significant differences in WHODAS scores by age (p 0.001), education level (p 0.001), marital status (p 0.001), and main work status (p 0.001), suggesting that disability levels varied significantly across these categories. Mann–Whitney U test showed no noteworthy differences in the WHODAS scores between females and males (p = 0.621) or between rural and urban residents (p = 0.686). However, a significant difference was found based on the living situation, with individuals living alone reporting lower WHODAS scores than those living with family (p = 0.023). These findings suggest that education, age, employment status, marital status, and living arrangements were important factors associated with functional disability in this sample.

Table 2 presents key Rasch model statistics evaluating the psychometric properties of WHODAS 2.0. The person reliability was 0.80, indicating good consistency in measuring differences in disability levels among respondents. Person separation was 2.00, suggesting the instrument can distinguish among approximately three levels of functional ability. Item reliability was exceptionally high at 0.99, with an item separation of 9.83, confirming a well-dispersed and discriminative item set. The mean person measure was –0.49 logits, indicating that respondents were slightly more able than the average item difficulty. The mean item measure was 0.00 logits, aligning with Rasch model expectations. Infit and outfit mean square values for persons (1.03 and 1.10) and items (1.01 and 1.10) were within the acceptable range of 0.5 to 1.5, indicating a good model fit with no substantial misfitting responses or items. These findings reflect the strong psychometric performance of the Arabic version of WHODAS 2.0, demonstrating excellent reliability and construct validity, and supporting its suitability for assessing functional disability in Arabic-speaking populations.

Table 3 depicts the findings of the PCA of the standardized residuals from the Rasch model, expressed in eigenvalue units. The total raw variance in the observations is 25.73, representing 100% of the variance in the dataset. Of these, 53.4% were explained by the Rasch measures, which is above the commonly accepted 50% threshold, supporting the assumption of unidimensionality. This explained variance is further divided into 36.6% by person measures and 16.8% by item measures, closely aligned with the expected values. The remaining 46.6% of the variance was unexplained. The largest portion of this residual variance is found in the first contrast, with an eigenvalue of 1.82 and accounting for 11.6% of the unexplained variance, which is lower than the expected value of 25%. Although this may indicate a potential minor secondary dimension, it does not exceed the critical eigenvalue threshold of 2.0. The second through fifth contrasts all have eigenvalues below 2.0, and contribute progressively smaller proportions to the unexplained variance (ranging from 3.9% to 7.8%), suggesting that these contrasts reflect random noise rather than meaningful multidimensionality. Overall, this table supports the conclusion that WHODAS 2.0 demonstrates a predominantly unidimensional structure in this sample.

Table 4 presents item-level Rasch statistics for WHODAS 2.0. Most items exhibited good fit, with infit and outfit MNSQ values within the satisfactory range (0.5–1.5). World Health Organization Disability Assessment Schedule 5 (WHODAS5) showed significant misfit. Exact match percentages were close to expectations, confirming model-data fit. Overall, the scale demonstrates strong psychometric properties, though one item may require review to ensure cultural appropriateness and construct alignment.

Fig. 1 displays the category probability curve for item No. 5 (WHODAS5), illustrating the probability of endorsing each response category across varying levels of person ability relative to the item’s difficulty. Ideally, each category should have a distinct peak indicating the point at which it is the most likely response. However, this figure shows disordered thresholds—only the lowest (red: “None”) and highest (cyan: “Extreme”) categories dominate, while the middle categories (blue, pink, green) overlap and never peak distinctly. This suggests poor category functioning, meaning that the respondents are unlikely to choose intermediate options. This pattern indicates that WHODAS5 may not effectively differentiate among moderate levels of disability, possibly due to ambiguous wording, cultural misinterpretation, or difficulty understanding the rating scale. Consequently, this item may need to be revised to enhance the scale’s measurement accuracy.

Fig. 1.

Category probability curve for World Health Organization Disability Assessment Schedule 5 (WHODAS5), showing disordered thresholds and poor distinction among middle response categories. The lowest and highest categories dominate, while intermediate options fail to peak. This suggests response scale dysfunction and limited measurement precision, indicating the item may require revision or removal.

DIF analysis showed no significant DIF across the examined classification groups, indicating that all World Health Organization Disability Assessment Schedule 2.0 (WHODAS 2.0) items function equivalently across subgroups. There were zero DIF classifications and no flagged pairwise comparisons, confirming that the items did not show evidence of measurement bias. The largest item calibration change was only 0.005 logits in the first iteration, reducing to –0.001 logits by the third iteration with zero active cells, suggesting model stability and convergence. These findings support the measurement invariance of WHODAS 2.0, and affirm its validity and fairness for comparing disability scores across different demographic or clinical groups.

Fig. 2 illustrates the Wright Map for WHODAS 2.0, which visually aligns the distribution of person abilities (left side) with item difficulties (right side) along the same logit scale. Each “#” symbol on the left represents a cluster of respondents at a particular ability level, while item labels on the right reflect their relative difficulty. The majority of participants were positioned between –1 and +1 logits, indicating a moderate level of self-reported disability. WHODAS10 and WHODAS11 emerged as the most difficult items, requiring higher levels of functional limitation to be endorsed, whereas WHODAS1, WHODAS2, and WHODAS7 were the easiest items for participants to affirm. The average person ability was slightly lower than the average item difficulty, suggesting that the items were somewhat more challenging than the sample’s functional level. While the item distribution shows good overall targeting, WHODAS5—although of average difficulty—may warrant revision due to its misfit as identified in previous Rasch outputs.

Fig. 2.

Wright Map displaying the alignment of person ability (left) and item difficulty (right) on the same logit scale for WHODAS 2.0. Each “#” symbol represents the frequency of respondents at a given ability level. Items above 0 logits (e.g., WHODAS10, WHODAS11) are more difficult for participants, while those below (e.g., WHODAS1, WHODAS2, WHODAS7) are easier. The map demonstrates good item-person targeting, with most individuals’ abilities distributed across the range of item difficulties. Symbols (S, T, M, +, .) are auto-generated WINSTEPS indicators denoting summary statistics and mean locations.