Comparative psychometric evaluation of the Arabic version of four patient-reported outcome measures for sleep assessment: a construct validity study using Rasch analysis

This study evaluated the psychometric properties of four Arabic-language patient-reported outcome measures (PROMs) for sleep assessment: the Insomnia Severity Index (ISI), Epworth Sleepiness Scale (ESS), Pittsburgh Sleep Quality Index (PSQI), and PROMIS Sleep-Related Impairment (SRI) item bank. A sample of 314 healthy Arabic-speaking adults (mean age, SD: 29.1, 12.0 years; 82.2% female) with subclinical sleep symptoms on average was recruited from healthcare and community settings in Saudi Arabia. Of these, 165 participants (52.5%) were recruited from healthcare facilities, while the remaining 149 (47.5%) were recruited from community settings via an online platform. Participants completed all sleep PROMs with sociodemographic data. Rasch analysis assessed the functioning of the rating scale, model fit, dimensionality, differential item functioning (DIF), item-person targeting, and reliability. The ISI showed one misfitting item, low person reliability (0.68), and poor targeting, limiting its use for mild sleep impairment. The ESS demonstrated ordered categories and acceptable fit but reduced reliability (0.73) and poor targeting. The PSQI demonstrated acceptable item fit; however, it also showed very low reliability (0.48) and poor targeting. The original 16-item PROMIS-SRI was shortened to a 13-item version by removing three items due to poor Rasch model fit. The revised 13-item PROMIS-SRI exhibited robust psychometric properties, including well-ordered categories and thresholds, proper fit to the model, negligible multidimensionality, no substantial DIF, acceptable targeting, and high reliability (0.90). Among the four tested PROMs, the 13-item questionnaire from the PROMIS SRI item bank was the most psychometrically sound tool for assessing sleep-related impairment in Arabic-speaking populations. Although the ISI, ESS, and PSQI are commonly used to quantify sleep disturbance, our analysis highlighted psychometric limitations within this subclinical sample, suggesting caution in their application, particularly for detecting mild sleep disturbances.