AB1005 Multiple Modeling Methods of Administrative Data Yield Consistent Results but Limited Ability To Predict Rheumatoid Arthritis Disease Activity in US Veterans. (15th July 2016)

• Record Type: Journal Article
• Authors: Cannon, G.W.; Teng, C.-C.; Accortt, N.; Collier, D.; Trivedi, M.; Sauer, B.C.
• Is Part Of: Annals of the rheumatic diseases. Volume 75(2016)Supplement 2
• Page Start: 1245
• Page End: 1245
• Publication Date: 2016-07-15
• Subjects: Rheumatism -- Periodicals; 616.723005
• DOI: 10.1136/annrheumdis-2016-eular.1519
• Languages: English
• ISSNs: 0003-4967

Abstract

Abstract : Background: Epidemiologic research would be significantly enhanced if accurate and reliable methods were available to assess rheumatoid arthritis (RA) disease activity using administrative data. The VA rheumatoid arthritis (VARA) registry is a prospective observational cohort that collects RA disease activity including the 28 joint count disease activity score (DAS28) and clinical disease activity index (CDAI). VA administrative data are available and can be employed to develop models to estimate RA disease activity measured by DAS28 and CDAI. Objectives: 1. Use administrative data to develop models that predict disease activity. 2. Assess the accuracy of the model under different conditions including defining administrative variables as continuous (CON) and/or categorical (CAT) and stratifying by different disease durations. 3. Describe the accuracy of the models for predicting disease activity for both DAS28 and CDAI. Methods: US Veterans with RA enrolled in the VARA registry without cancer, organ transplantation, or other autoimmune diseases were included in the model if they were enrolled in VA with one year of administrative data available before the first DAS28. We identified 1, 275 administrative data elements with suspected association with disease activity for prediction modeling. The least absolute shrinkage and selection operator (LASSO) was used for variable selection and model development. A series of LASSO models were developed that varied the input variables. Variables were required to have ≥1% prevalence in the cohort to be included in model selection. Sub-analysis was also performed in patients with ≤2 years and ≤5 years of disease duration. Results: There were 1, 582 US Veterans with RA who fulfilled enrollment criteria. The average age was 62.7±11.0 years, 91% male, with a disease duration 13.7±12.1 years with 78% and 74% seropositive for rheumatoid factor and anti-CCP antibodies respectively. The adjusted r-square for the six models tested ranged from 0.20–0.24 for DAS28 and 0.17 to 0.24 for CDAI. When stratified by disease duration, the r-square associations for patients with and without ≤2 and ≤5 years of RA duration were 0.29–0.31 versus 0.22–0.27 for DAS28 and CDAI respectively. Conclusions: The multiple models tested yielded very similar results with limited association of predicted disease activity with actual measured disease activity by DAS28 and CDAI in VARA. The use of CON versus CAT variables, multiple definitions for disease activity, and limitation of disease duration to early RA did drastically affect prediction accuracy. Future research should investigate additional strategies to collect components of disease activity measures directly from medical records. Disclosure of Interest: G. Cannon Grant/research support from: Amgen, C.-C. Teng: None declared, N. Accortt Shareholder of: Amgen, Employee of: Amgen, D. Collier Shareholder of: Amgen, Employee of: Amgen, M. Trivedi Shareholder of: Amgen, Employee of: Amgen, B. Sauer Grant/research support from: Amgen