Clinical spectrum and diagnostic outcomes of patients with suspected inflammatory rheumatic disease in the emergency department: a retrospective study.

Little is known about patients presenting to the emergency department (ED) with rheumatological complaints. We aimed to characterize rheumatological patients in the ED and analyze the role of the rheumatologist in their management. This retrospective single-center cohort study analyzed 340 patients presenting to the emergency department (ED) between 2019 and 2023 with a suspected new inflammatory rheumatic disease (IRD). Clinical and process data were evaluated. In addition, multivariable logistic regression and a neural network model were employed to identify factors that are predictive of a subsequent diagnosis of IRD. Arthritis and arthralgia (55%) were the most common symptoms among patients with susIRD, but a wide spectrum of symptoms were observed, including myalgias (23%) and B-Symptoms (20%). Most patients presented as urgencies rather than true emergencies. Symptom duration prior to ED presentation was longer than 7 days in 69% of patients. However, 63% required in-patient management. Overall,y 62% of patients with susIRD were eventually diagnosed with an IRD. Immunosuppressive or anti-inflammatory medication was started in 92% of these. The most common diagnoses were polymyalgia rheumatica (7.9%), reactive arthritis/para infectious arthritis (7.3%) and rheumatoid arthritis (6.3%). Infections (10%) constituted the most common non-rheumatological diagnosis. A logistic regression model and a neural network could predict the presence of an IRD with accuracies of 74% and 60% respectively. Polyarthritis/Polyarthralgia and swollen joints were the strongest positive predictors for IRD. When consulting rheumatologists suspected an IRD, this impression was confirmed by the final diagnosis in 87% of patients. Integrating the rheumatologist's impression into the models increased the accuracy of the logistic regression to 85% and the neural network to 77%. Most rheumatological patients present to the ED with arthritis/arthralgias and urgent rather than life-threatening conditions. Whilst our models could predict IRD with moderate accuracy, consulting rheumatologists demonstrated high diagnostic precision, indicating the need for rheumatological expertise in the ED.