Yale School of Medicine New Haven, Connecticut, United States
Disclosure(s): No financial relationships with ineligible companies to disclose
Background/Purpose: Lack of timely rheumatology referral of chronic back pain (CBP) patients with suspected axial spondyloarthritis (axSpA) is a major contributor to delayed diagnosis. We previously developed and validated an axSpA online self-screening tool (A-Tool) for patients with CBP to expedite self-referrals and evaluation [1].The original A-Tool consists of 3 pre-screening questions and 8 screening questions pertaining to high yield clinical features of axSpA without lab or imaging data. Here, we present a modified version of A-tool, that incorporates diagnostic testing and estimates post-test probability for patients.
Methods: We modified the A-tool based on the results of our prospective validation study [1]. “Unexplained joint swelling” and “Heel pain” questions were removed, while “Gradual onset” was added. We then ascribed weighted scores to each modified A-Tool item using the known positive likelihood ratios of axSpA features from literature: greater weight was put on prior diagnoses of psoriasis, inflammatory bowel disease, and uveitis (Table 1). A logarithmic model converted scores into post-test probabilities (Figure 1A). Pre-test and maximum post-test probabilities in the model were set to 5% and 45%, respectively, based on prior literature [2]. We added an optional section for diagnostic test results: CRP, HLA B-27 and imaging sacroiliitis. When diagnostic data are included, the maximum post-test probability can increase to 100%. For analysis, we used the original A-Tool cohort to evaluate Receiver Operator Characteristic (ROC) curves, sensitivity, and specificity.
Results: There were 86 patients in the original A-Tool cohort; 29 had axSpA. All patients had CRP, HLA-B27 testing; 78 had axial MRI. ROC curves were generated for the original A-Tool, symptom-only modified A-Tool, and full modified A-Tool: area under the curve (AUC) were 0.56, 0.63, and 0.93, respectively (Figure 1B). At a post-test probability threshold of 80%, the modified A-Tool yielded a sensitivity of 86%, specificity of 95%, and positive predictive value of 89%.
Conclusion: We modified a previously validated online symptom-based axSpA self-referral tool, A-Tool, to include diagnostic testing and calculate post-test probabilities. The modified version had high sensitivity and specificity in an existing cohort. We aim to distribute the modified A-Tool as an publicly available online calculator for patients and primary care providers to facilitate rheumatology referrals (Figure 2). Future validation studies are needed to evaluate the modified A-Tool’s test characteristics on a larger cohort.
References: [1] Danve A. Arthritis Rheumatol. 2022; 74 (suppl 9):1514 [2] Danve A, Deodhar A. Clin. Rheumatol. 2015;34(6):987-993.
