A reconceptualization of IRT calibration with DIF items in a PROMIS Fatigue measure
- UNCG Author/Contributor (non-UNCG co-authors, if there are any, appear on document)
- Jia Ma (Creator)
- Institution
- The University of North Carolina at Greensboro (UNCG )
- Web Site: http://library.uncg.edu/
- Advisor
- Richard Luecht
Abstract: Differential item functioning (DIF) is a statistical procedure intended for examining and evaluating test fairness. After DIF items are detected, there are three methods to deal with DIF items, which are to ignore DIF items, remove DIF items, and create two new items from the original DIF items the related demographic variable, named demographic-specific items. In PRO research, current research and practice only focus on the first two methods. The present study evaluated and compared the performance of the three methods by applying IRT calibration. This study used real word data from MY-Health database with a subset of 1808 cancer patients to provide concrete evidence of the evaluation of the three calibration approaches. Wald test and Welch test were applied for DIF detection, then followed by using GRM and PCM for conducting IRT calibration. The comparison among the three calibration approaches suggested that demographic-specific group approach had the best performance in item fit and person fit; it demonstrates great advantage with improving measurement precision, and at the same time, content validity of the test is still promising, which had a positive impact on clinical studies. The removed DIF item approach was less favorable; it caused new misfit items and made larger standard errors than the other two approaches. The challenge of this study was to deal with the measurement equivalence issue in an existing instrument and patient sample, and it was not aimed at modifying the existing instrument.
A reconceptualization of IRT calibration with DIF items in a PROMIS Fatigue measure
PDF (Portable Document Format)
1162 KB
Created on 5/1/2022
Views: 45
Additional Information
- Publication
- Dissertation
- Language: English
- Date: 2022
- Keywords
- Educational tests & measurements
- Subjects
- Item response theory
- Outcome assessment (Medical care)