PR1 DEVELOPMENT OF A CROSS-WALK OF THE ASSESSMENT OF SPONDYLOARTHRITIS INTERNATIONAL SOCIETY HEALTH INDEX (ASAS HI) AND ANKYLOSING SPONDYLITIS QUALITY OF LIFE (ASQOL) SCORES USING DATA FROM PATIENTS WITH ANKYLOSING SPONDYLITIS AND NON-RADIOGRAPHIC AXIAL SPONDYLOARTHRITIS

Abstract

To conduct a cross-walk analysis of the Assessment of SpondyloArthritis International Society Health Index (ASAS HI) to the Ankylosing Spondylitis Quality of Life (ASQoL) in patients with physician confirmed ankylosing spondylitis (AS) or non-radiographic axial spondyloarthritis (nr-axSpA) in the United States. Data from a cross-sectional survey conducted with rheumatologists and their consulting patients in the United States were analyzed. Data were collected from Jun-Aug 2018 via physician-completed patient record forms and patient self-completed forms. To ensure eligibility to participate, physicians provided confirmed diagnoses of AS and nr-axSpA plus basic demographic and clinical information. Quality of Life measures were reported by the patient using the validated ASAS HI and ASQoL questionnaires. Model performance was assessed by comparing root-mean squared error (RMSE) from 10-fold cross validation, to determine the best mapping from ASAS HI to ASQoL, and also ASQoL to ASAS HI. RMSE was calculated overall, and for lower, middle and upper thirds of the predicted scale. Data from 63 rheumatologists, 283 AS patients, and 274 nr-axSpA patients were included in this analysis. Predicting ASAS HI using ASQoL values, the best model was non-parametric local-linear regression, with overall RMSE 1.851. Predicting ASQoL using ASAS HI values, the best model also used non-parametric local-linear regression, with overall RMSE 2.254. In predicting ASAS HI and ASQoL, models performed better in predicting lower values in the predicted scale (RMSE 1.597, 1.871, 2.871 across thirds for ASAS HI, and 1.719, 2.577, 3.140 for ASQoL). The results demonstrate the appropriateness of cross-walking the ASAS HI to the ASQoL and scoring algorithm approaches. This approach enables comparisons across studies using ASAS HI or ASQoL.