Detection of differential item functioning for the Nine-Questions Depression Rating Scale for Thai North dialect / Suttipong Kawilapat ... [et al.]

• Main Authors: Kawilapat, Suttipong, Maneeton, Benchlak, Maneeton, Narong, Prasitwattanaseree, Sukon, Kongsuk, Thoranin, Arunpongpaisal, Suwanna, Leejongpermpool, Jintana, Sukhawaha, Supattra, Traisathit, Patrinee
• Format: Conference or Workshop Item
• Language: English
• Published: 2021
• Subjects: Psychological tests and testing; Anxiety
• Online Access: https://ir.uitm.edu.my/id/eprint/56198/1/56198.pdf; https://ir.uitm.edu.my/id/eprint/56198/
• Institution: Universiti Teknologi Mara

A different probability of response to an item among respondents with the same underlying trait,differential item functioning (DIF), may compromised comparisons across subgroups and lead to misleading results. This cross-sectional study to detect the DIF among 1,475 participants who participated in a validity study of the Nine-Questions Depression Rating Scale (9Q) for Thai North dialect. The 9Q was consisted of 9 polytomous items assessed for the severity and frequency of depression symptoms according to DSM-5 within 2 weeks. The detection of DIF was considered between gender, age, underlying disease, and income using item response theory (IRT)-based and ordinal logistic regression (OLR)-based approach. The IRT difficulty parameters for each item and threshold parameters for each possible category were estimated based on the generalized partial credit model. Item 1 has the highest discrimination parameter following by item 9 (2.533 and 2.387, respectively). The IRT-based approach seemed to be proper than OLR-based approach in the examination of DIF and scoring due to the different of discrimination parameters. The DIF were presented between gender (item 5 and 8), age (item 1, 2, 4, and 7), and underlying disease (item 2, 3, 5, 6, and 7). Scoring accounted for the discrimination and threshold parameters based on IRT approach might be useful to reduce the bias of depression assessment.