Can severity be predicted by treatment variables in rheumatoid arthritis administrative data bases?

OBJECTIVE: Administrative data bases provide rapid access to data regarding treatment and morbidity of rheumatoid arthritis (RA). A serious limitation of administrative data bases is the lack of information regarding RA severity, as in the case of lymphoma, where RA severity may contribute to the cause of the adverse outcome. We examined whether treatment variables could predict RA severity.
METHODS: We studied 7541 patients with RA who were participating in a longitudinal study of RA outcomes. Disease severity was determined by the Patient Activity Scale (PAS), which represents on a 0 to 10 scale the mean of 0-10 standardized values of pain (by visual analog scale), patient global severity, and the Health Assessment Questionnaire. We tested the ability of disease modifying antirheumatic drugs (DMARD) and biologic treatment variables and the lifetime number of these treatments to predict severity status. The receiver-operating characteristic (ROC) area under the curve (AUC) was used to describe the association between severity and treatment variables.
RESULTS: There was little difference in PAS scores between various treatments and treatment groups, including scores of the 18.3% of patients receiving no DMARD or biologic therapy. The ROC AUC to distinguish PAS scores above and below the median was 0.64 (60.5% correctly classified) and was 0.70 (67.2% correctly classified) in distinguishing first compared to fourth quartiles PAS scores.
CONCLUSION: Treatment variables do not accurately or usefully identify severity status. As a corollary, there is little difference in severity between patients receiving different treatment regimens, and actual measures of severity rather than treatment surrogates are required to assess RA severity.