Rasch measurement in rheumatoid arthritis: deriving psychometrically optimal measures from the Rasch Everyday Activity Limitation item bank.

ABSTRACT

Objective: Recently we developed the Rasch Everyday Activity Limitations (REAL) generic item bank for measuring physical function. In this study we evaluate the REAL item bank in 209 RA patients and demonstrate how computerized adaptive testing (CAT) and Optimal Test Assembly methods can be used to derive measures from the REAL item bank with superior measurement performance compared with the HAQ Disability Index (HAQ-DI). Methods: Structural validity of the item bank was assessed using confirmatory factor analysis. The validity of the REAL score metric in RA was evaluated by examining differential item functioning against the general population calibration sample. Besides the REAL-CAT, a 6-item short form (REAL-6) was developed using Optimal Test Assembly that was optimized with respect to common disability levels in RA, content and reading ease. Measurement precision of the different instruments was examined using item response theory methods. Construct validity was evaluated by testing hypothesized correlations with external measures. Results: Good model-data fit was observed for a one-dimensional model and only two items showed differential item functioning of substantial magnitude. The REAL-CAT had superior measurement precision compared with HAQ-DI and REAL-6. REAL-6 outperformed HAQ-DI across all but the very lowest level of physical function. All three instruments demonstrated good construct validity (>75% of hypotheses affirmed) and only HAQ-DI had a ceiling effect (23.9%). Conclusion: This study supports the validity of the REAL item bank and illustrates the potential of CAT and OTA applications based on the REAL item bank for assessing physical function in RA.