RESUMO
OBJECTIVE: Health information technology has enabled efficient measurement of patient-reported outcomes (PROs). The National Institutes of Health Patient-Reported Outcomes Measurement Information System (PROMIS) is becoming more widely adopted for research and routine care, and some PROMIS instruments might be substituted for lengthier, legacy PRO instruments. METHODS: Four PROMIS computer-adaptive testing (CAT) instruments (pain interference, physical function, sleep disturbance, and fatigue) and the Routine Assessment of Patient Index Data 3 (RAPID3), along with pain intensity and patient global assessment score, were administered to participants in the ArthritisPower registry. The RAPID3 was predicted using different combinations of these variables to create a new score (CAT-PROMIS RAPID3). Kappa statistics and Bland-Altman 95% limits of agreement were used to measure agreement between the observed versus predicted RAPID3. RESULTS: A total of 6,154 eligible patients contributed 11,275 observations. The mean ± SD age was 52.7 ± 10.5 years, and 93% of patients were women. The median assessment times ranged from 29 seconds (PROMIS sleep disturbance) to 116 seconds (RAPID3). As single pairwise comparisons, the PROMIS CATs examined were modestly correlated (r approximately 0.4-0.7) to one other and RAPID3. Together with the pain intensity and patient global assessment, the PROMIS instruments explained a high fraction of total variance (R2 = 0.97) of the RAPID3 score. In the model with the highest agreement (κ = 0.93) between the observed RAPID3 and the CAT-PROMIS predicted RAPID3, Bland-Altman 95% limits of agreement showed minimal residual differences and no systematic biases. CONCLUSION: There was excellent agreement between the observed RAPID3 and predicted RAPID3 scores estimated using several PROMIS instruments. The Multidimensional Health Assessment Questionnaire and patient global assessment components of RAPID3 may be unnecessary if PROMIS scores are available.
