Development of a Web-based Screening Tool for Early Rheumatoid Arthritis-ERASE: the E-triage RA Study in Early Arthritis

PURPOSE: The lack of time devoted to MSK during medical training means that primary care physicians have limited expertise in identifying early RA. Patients with MSK disorders increasingly turn to the web in attempts at self-diagnosis but validated screening tools for disorders such as RA that would prompt a visit to the physician and early referral to a rheumatologist are not available. Our aim was to develop and validate a web-based screening tool (ERASE) for early RA. METHODS: Items for ERASE were identified following a Delphi consensus of Alberta rheumatologists who were asked to provide questions exactly as they would be elicited during a patient interview. ERASE was then administered in paper format to consecutive new patients attending an outpatient rheumatology clinic before evaluation by the rheumatologist who then rated the probability of RA on a 0-10 (NRS). A score of at least 5 on the NRS for RA constituted the dependent variable in univariate regression analysis. Significant items were then tested in multivariate linear, hierarchical, and stepwise logistic regression analyses to identify those questions independently associated with the rheumatologists diagnosis of RA. A scoring tool was developed with weighting for the items associated with RA based on the beta coefficients from logistic regression analysis and this tool was tested for discrimination in ROC analysis. RESULTS: The Delphi consensus (15 rheumatologists participating) resulted in 7 items thought to have both positive and negative predictive value for RA. The latter were symptoms typically associated with fibromyalgia (jaw pain, irritable bowel syndrome, chronic fatigue, daily headaches) (JICH). We tested these items in 124 consecutive newly referred patients of whom 21 were rated as having RA. Logistic regression identified the following items as being significantly positively or negatively associated with RA: self-reported distribution of joint involvement (Odds Ratio (OR) for MCP = 1367, for wrist = 73), duration of symptoms (weeks/months versus years, OR = 34), presence of any two of jaw pain, irritable bowel syndrome, chronic fatigue, daily headaches, (JICH) (OR = 0.02)). The item related to morning stiffness was not significant in multivariate regression when pattern of joint involvement was included first in the stepwise regression. ERASE was further developed as a screening index based on a weighting of these items ((4 for any MCP involvement) + (2 for any wrist involvement) + (2 for duration weeks/months) + ( -3 for any 2/4 of JICH)). ROC analysis based on the ERASE scoring tool revealed an AUC of 0.93 and a sensitivity of 95%, specificity of 76%, using a cut-off score of 2.5. CONCLUSIONS: Our preliminary data shows desirable properties for a simple screening tool for RA (high sensitivity and very good specificity) based on only 3 items: self-reported pattern of joint involvement, symptom duration, and the absence of symptoms typically associated with fibromyalgia.