Purpose
In order to test the difference between group means, the construct measured must have the same meaning for all groups under investigation. This study examined the measurement invariance of responses to the patient-reported outcomes measurement information system (PROMIS) pain behavior (PB) item bank in two samples: the PROMIS calibration sample (Wave 1, N = 426) and a sample recruited from the American Chronic Pain Association (ACPA, N = 750). The ACPA data were collected to increase the number of participants with higher levels of pain.
Methods
Multi-group confirmatory factor analysis (MG-CFA) and two item response theory (IRT)-based differential item functioning (DIF) approaches were employed to evaluate the existence of measurement invariance.
Results
MG-CFA results supported metric invariance of the PROMIS–PB, indicating unstandardized factor loadings with equal across samples. DIF analyses revealed that impact of 6 DIF items was negligible.
Conclusions
Based on the results of both MG-CFA and IRT-based DIF approaches, we recommend retaining the original parameter estimates obtained from the combined samples based on the results of MG-CFA.