Adult Still's disease (AOSD) is a systemic, auto-inflammatory disease that was first reported in the early 1970s. Most patients with AOSD present with high fever, transient rash, arthralgia or arthritis, and sore throat. The clinical features of AOSD are so similar to Sepsis that it is also known as 'Allergic Subepsis.' It is often difficult to differentiate between the two, especially in the early stages when fever is the initial clinical manifestation.

Studies have shown that ferritin, interleukin 18, and some hematological markers can be used as differential markers between AOSD and Sepsis. However, the lack of specificity makes the differentiation between the two complicated and delays the diagnosis and treatment. Therefore, it is essential to identify AOSD and Sepsis quickly and accurately for early diagnosis and treatment of the disease. To identify AOSD and Sepsis more accurately, this study compared three machine learning methods, namely the Random Forest algorithm (RF), Gradient Boosting Machine (GBM), and Logistic Regression algorithm (LR), to combine the clinical features and hematological indicators of AOSD and Sepsis and to construct a differential diagnosis model for the two.

To facilitate clinical use and generalization, we selected seven medical indicators, namely arthralgia, ferritin times lymphocyte count, white blood cell count, ferritin times platelet count, and α1-acid glycoprotein divided by creatine kinase, to build a differential diagnosis model for AOSD and Sepsis, and named it: AIADSS (AI-assisted discrimination of Still's disease and Sepsis).