Protocol for the development and validation of a Rheumatoid Arthritis PredIction moDel using primary care health records (RAPID)

Abstract

AbstractBackgroundRheumatoid Arthritis (RA) is a chronic rheumatological condition which causes inflammation of both the joint lining and extra-articular sites. It affects around 1% of the UK population and, if not properly treated, can lead joint damage, disability, and significant socioeconomic burden. The risk of long-term damage is reduced if treatment is started in an early disease stage with treatment in the first 3 months being associated with significantly improved clinical outcomes. However, treatment is often delayed due to long referral waits and challenges in identifying early RA in primary care. We plan to use large primary care datasets to develop and validate an RA risk prediction model for use in primary care, with the aim to provide an additional mechanism for early diagnosis and referral for treatment.MethodsWe identified candidate predictors from literature review, expert clinical opinion, and patient research partner input. Using coded primary care data held in Clinical Practice Research Datalink (CPRD) Aurum, we will use a time to event Cox proportional hazards model to develop a 1-year risk prediction model for RA. This will be validated first in CPRD GOLD and then independently in the Secure Anonymised Information Linkage dataset. We will also conduct a sensitivity analysis for the same model at 2–5-year risk, with a secondary outcome of RA and initiation of a disease modifying drug, and with the addition of laboratory test results as candidate predictors.DiscussionThe resulting risk prediction model may provide an additional mechanism to distinguish early RA in primary care and reduce treatment delays through earlier referral.