Differential item functioning within the Dutch and Flemish PROMIS pain behavior and interference item banks: An empirical comparison of IRT and CFA approaches
Year: 2018
Type: dissertation
Abstract: This study investigated the application of two major approaches in testing for differential item functioning (DIF), or measurement invariance (MI), on Dutch-Flemish PROMIS datasets on Pain Behavior and Pain Interference item banks, developed to be used in computerized adaptive testing (CAT).The Pain Behavior item bank consists of 39 items, and the Pain Interference item bank of 40 items.Both are measured on a six-point Likert scale, although participants who had no pain at baseline (category 1) in the Pain Behavior item bank were excluded from the analysis, which effectively resulted in a five-point scale.We applied item response theory (IRT) and confirmatory factor analysis (CFA) models on responses of approximately 6600 participants from the general population, chronic pain patients, and rheumatoid arthritis patient groups, with the goal of determining whether items exhibited different measurement properties across groups.Through the IRT approach, we found four items, out of a total of 79, showing uniform DIF with minimal impact on theta scores in three comparisons.Through the CFA approach, all items were found to be measurement invariant across groups using ΔCFI being equal to or less than 0.002 as a measure of goodness of fit.As items flagged for DIF had minimal impact on theta scores, it is likely there may not be a significant impact on theta estimates in CAT if items were used without added calibration.We have found that the IRT and CFA approaches show results with a high degree of similarity.
Author: Okicki, Tihana
Institutions: Leiden University, Institute of Psychology
Language: English
Field: Decision Sciences
Domain: Social Sciences
Open Access status: green
OpenAlex ID: W7037345144
