Background/Purpose: Health information technology has enabled efficient measurement of PROs using Computer Adaptive Testing (CAT) methods, which have been shown to minimize missing data and reduce participant burden. The bridge between PROMIS and traditional patient-reported PROs (e.g. RAPID3) is not clear, and it might be possible that some instruments in PROMIS might proxy for measuring more lengthy PROs.

Methods:  Four PROMIS CAT instruments: Pain Interference, Physical Function, Sleep Disturbance, Fatigue, and a visual analog scale (Pain Intensity);and the Routine Assessment of Patient Index Data 3 (RAPID3) were administered to participants in the PCORI-funded ArthritisPower patient registry via a mobile application (App) on their smartphone or computer. The RAPID3 was evaluated in terms of multiple correlations (Pearson) with PROMIS and total variance explained using mixed models. Both the RAPID3 score and category was predicted using the selected 4 PROMIS instruments and pain intensity to form a new predicted score (“CAT-PROMIS RAPID3”). Kappa statistics and Bland-Altman 95% limits of agreement (LOA) were used to measure agreement between the measured RAPID3 and the predicted CAT-PROMIS RAPID3.

Results:  At the time of analysis, 2,376 patients had answered all five PROMIS instruments and the RAPID3 and contributed 3906 observations. Mean +-SD age was 51.6 +/- 10.5 years; 90% were women. Most (81%) had high RAPID3 scores. The mean assessment time for each of the PROMIS instruments ranged from a low of 16 seconds (sleep disturbance) to a high of 105 seconds (RAPID3).As single pairwise comparisons, the 4 PROMIS CATs examined were only modestly correlated (r ~ 0.2-0.4) to one other and the RAPID3. However, in aggregate, the PROMIS instruments explained a very high fraction of total variance (R²=80%) of the RAPID3. The measured RAPID3 was very highly correlated with the CAT-PROMIS predicted RAPID3 (r=0.97). Overall agreement (table) between categories of the RAPID3 vs. CAT-PROMIS RAPID3 also was very high (kappa = 0.81). Bland-Altman 95% limits of agreement shows minimal residual differences and no systematic biases.

Conclusion: There was excellent agreement between the measured RAPID3 and a predicted RAPID3 score estimated using several PROMIS instruments. Few if any questions in the RAPID3 may need to be measured if PROMIS scores are available.