Rasch analysis of the Back Pain Attitudes Questionnaire (Back-PAQ)

Context As psychosocial factors have been recognised as significant predictors of the recovery trajectory from chronic back pain, the 34-item Back Pain Attitudes Questionnaire (Back-PAQ) was developed based on themes obtained from patient interviews, but previous psychometric analyses with a general population sample revealed uncertainty around the factor structure of the instrument. Objectives To provide more detailed information about the psychometric properties of the Back-PAQ when used with participants from the general population and also to test the internal validity of the tool for use with General Practitioners (GPs). Methods After applying partial-credit Rasch analysis with a sample of participants from the general population (n = 600), a replication analysis was conducted with a sample of GPs (n = 184). This approach permitted examination of sample-specific personal factors for differential item functioning. Subtests were used to differentiate between local dependency due to underlying dimensionality from local dependency due to method effects. Results A unidimensional fit to the Rasch model was achieved after 14 misfitting items had been deleted. The final 20-item solution also fit with a sample of 184 GPs. In both cases, the Back-PAQ-20 demonstrated good reliability (PSI >= 0.80), with no evidence of differential item functioning by personal factors. Conclusion The ordinal-to-interval conversion algorithms presented here further enhance the precision of the scale and permit analysis of Back-PAQ-20 scores using parametric statistics. The present study provided evidence for valid and reliable assessment of the back pain recovery beliefs of both users as well as providers of health services.