Purpose: Sociodemographic characteristics may influence responses on self-reported measures. Differential item functioning (DIF) is when individuals expected to have the same ability level on a construct of interest have a different probability of endorsing an item on an item response theory (IRT) scale due to population characteristics. The goal of this study was to identify DIF for items in an outcome instrument by sociodemographic factors and, one controlling for DIF, assess true differences in function by those same factors.
Methods: The Work Disability Functional Assessment Battery 2.0 (WD-FAB 2.0) is an IRT-based self-reported measure of activity limitations relevant to work. Two samples from WD-FAB developed were used: 3793 SSA disability claimants randomly drawn from a pool of 16,500 claimants and a general sample if 2100 working age adults. We used a two-step IRT-based DIF method for three pairs of respondent characteristics: age, gender, and race/ethnicity, and calculated the weighted absolute difference between item characteristic curves. Independent two-group T-tests assessed differences in scores across groups.
Results: Seventeen items displayed DIF. Men had higher scores than women on two physical and two mental function scales. Older respondents had lower physical and higher mental function scores. The lower education group had lower mental function scores.
Conclusion: DIF impacts function measurement and is important when assessing psychometric characteristics of instruments. Self-report measures should include diverse samples to conduct similar analyses. WD-FAB 2.0 scores are now reflections of function with reduced bias related to gender, race/ethnicity, or age.
Keywords: Differential item functioning; Functional assessment; Psychometrics; Work disability.