Background/Purpose:  Discordance between patient and physician assessment of rheumatoid arthritis (RA) disease activity strongly associates with pain scores. Patient-reported outcomes measurement information system (PROMIS) 29 is recommended as the preferred battery of measures to collect in musculoskeletal pain studies.  We evaluated the discordance between Routine Assessment of Patient Index 3 (RAPID3) and Clinical Disease Activity Index (CDAI) and investigated the associations between RA activity measures, discordance and patient-reported outcome measures using PROMIS29.

Methods: For RA subjects enrolled in the University of Pittsburgh Rheumatoid Arthritis Comparative Effectiveness Registry (RACER), a cross-sectional analysis was performed for all RACER patients who completed PROMIS29 short form and had RAPID3 and CDAI recorded.  Association between CDAI and RAPID3 was evaluated by the percentage of agreement, Pearson¡¯s correlation coefficient, and kappa statistics.   For subjects in remission/low disease activity by CDAI who had RAPID3 scores (0-10) consistent with near remission or low disease activity (¡Â 2) versus moderate or high disease activity (>2), predictors were determined using logistic regression.

Results: For the 363 subjects analyzed, age was 64.2 +/- 13.8 (mean +/- SD) years with disease duration of 13.9 +/- 13.1 years. CDAI was moderately correlated with RAPID3 (r=0.60, p<0.0001). PROMIS domains were correlated with RAPID3 and CDAI with various degrees (Figure). RAPID3 had strongest correlation with pain intensity (r = 0.89, p<0.001) and pain interference (r=0.84, p<0.001). Overall agreement between disease severity categories of CDAI and RAPID3 were fair (percent agreement = 40.1, kappa=0.21) (Table).  In subjects with remission to low disease activity by CDAI, pain intensity (standardized regression coefficient beta = 1.33, p<0.001) and physical function (beta = -0.21, p<0.001) among PROMIS29 domains predicted discrepancies between CDAI and RAPID (Akaike information criterion (AIC) =110.2). When pain intensity and physical function were removed from the model, pain interference (beta = 0.18, p<0.001 and impaired social roles (beta = -0.11, p=0.001) predicted discrepancies (AIC = 168.9).

Conclusion:   Discordance between CDAI and RAPID3 was frequent, especially among subjects with remission/low disease activity by CDAI. These subjects were predicted by PROMIS29 measures associated with pain intensity.  Our data suggest that PROMIS29 helps to understand the discordance between patient and physician assessment of disease activity.