Identification of sources of DIF using covariates in patient-reported outcome measures: a simulation study comparing two approaches based on Rasch family models

• DOI: 10.3389/fpsyg.2023.1191107
• Publication Date: 2023-08-10T09:17:31Z
• Authors: Yseulys Dubuy; Jean-Benoit Hardouin; Myriam Blanchin; Véronique Sébille
• Keywords: Differential item functioning; Identification

When analyzing patient-reported outcome (PRO) data, sources of differential item functioning (DIF) can be multiple and there may more than one covariate interest. Hence, it could great interest to disentangle their effects. Yet, in the literature on PRO measures, are many studies where DIF detection is applied separately independently for each under examination. With such an approach, covariates investigation not introduced together analysis, preventing from simultaneously studying potential effects questionnaire items. One issue, among others, that lead false-positive when correlated. To overcome this we developed two new algorithms (namely ROSALI-DIF FORWARD BACKWARD). Our aim was obtain iterative item-by-item method based Rasch family models enable adjust group comparisons presence binary covariates. Both were evaluated through a simulation study various conditions aiming representative health research contexts. The performance assessed using: (i) rates false correct DIF, (ii) size form recovery, (iii) bias latent variable level estimation. We compared another approach likelihood penalization. For both algorithms, rate close 5%. influenced DIF. Rates higher with increasing size. Besides, algorithm fairly identified homogeneous differences threshold parameters, but had difficulties identifying non-homogeneous differences. Over all, performed better penalized approach. Integrating several during process allow assessment understanding This provides valuable insights regarding different approaches undertaken fulfill aim.