Objective: To test for and estimate variation among rheumatologists in their prescribing of prednisone and second-line agents for the treatment of rheumatoid arthritis (RA), after taking into account the characteristics of their patients.
Methods: Multiple logistic regression incorporating random effects for rheumatologists, with adjustment for patient characteristics.
Results: Values for the likelihood-ratio statistic provided strong evidence of such variation. Random-effect variance estimates showed that the variation is of great magnitude.
Conclusion: Even after patient characteristics have been taken into account, the data show that the rheumatologist may strongly influence the use of prednisone and second-line agents by a patient.