OP0067 Development and validation of a model to facilitate recognition of arthritis by general practitioners

Background Early initiation of treatment of rheumatoid arthritis is strongly associated with an improved outcome, but requires the early identification of arthritis. Physical examination of joints is crucial to this end, but is difficult for general practitioners (GPs) who have little experience. Another difficulty is that GPs see many persons with musculoskeletal symptoms per year but only few patients with clinical arthritis. To promote early recognition of arthritis, the Early Arthritis Recognition Clinic (EARC) was initiated in Leiden, the Netherlands in 2010. GPs were instructed to refer to this clinic in case of doubt on the presence of arthritis (and not to wait and see, or to perform additional laboratory tests). At this clinic, patients filled out a form with questions on their symptoms and were seen by a rheumatologist in a short visit that comprised a full joint examination. As reported previously, this clinic importantly improved the early identification of arthritis and RA (1), but this approach may not be easily implemented in other centres or regions. Objectives To assess if a combination of symptoms and signs that are easy to assess can differentiate patients with and without clinical arthritis at joint examination. Methods 1,288 patients in whom GPs doubted on the presence of arthritis visited the EARC between 2010 and 2015. Reported symptoms and signs were studied with the presence of synovitis (joint examination by experienced rheumatologist) as outcome. Multivariable logistic regression was used. A model was derived in 644 patients, and validated in the second set of 644 patients. Results 41% of the patients who visited the EARC had arthritis at examination. Age (per year OR 1.02; 95% CI 1.01–1.03), male sex (OR 1.8; 95% CI 1.3–2.7), symptom duration (4–12 weeks OR 3.83; 95% CI 2.22–6.60), morning stiffness >60 min (OR 1.7; 95% CI 0.9–2.9), difficulty with making a fist (OR 1.6; 95% CI 0.97–2.5), number of tender joints (1–3 tender joints OR 9.7; 95% CI 1.1–81.8) and self-reported swollen joints (OR 3.5; 95% CI 1.8–7.0) were associated with the presence of arthritis in multivariable analysis. The AUC was 0.75 (SE 0.02) in the derivation set and 0.71 (SE 0.02) in the validation set. To facilitate application in practice, a simplified model was made. This consisted of 7 variables and the total score ranged between 0–7. The AUC was 0.73 (SE 0.02). Depending on the cut-off, the PPV of the simplified model ranged between 41 and 74% and the NPV between 100 and 62%. With a cut-off of 4, the NPV was 86%, PPV 49%, specificity 35%, and sensitivity 91%. Conclusions A set of clinical characteristics that can be easily assessed by GPs had a reasonable discriminative ability for clinical arthritis, and can be applied by GPs in case of doubt on the presence of arthritis. This model requires further validation in general practices, but may lead to a tool that could assist GPs in their decision making regarding referral or ordering additional tests for patients with suspected early arthritis. References van Nies JA, et al. Ann Rheum Dis. 2013;72(8):1295–301. Disclosure of Interest None declared