A machine learning model for mortality prediction in patients with severe fever with thrombocytopenia syndrome: a prospective, multicenter cohort study.

Severe fever with thrombocytopenia syndrome (SFTS) is an emerging infectious disease that imposes a considerable medical burden. In this study, we enrolled 1,606 SFTS patients, developed and validated machine learning models for mortality prediction, and ultimately constructed a model consisting of six variables. The prediction model, UNION-SFTS, constructed using the multilayer perceptron (MLP) algorithm, achieved the best performance with an area under the curve (AUC) of 0.917, an accuracy of 0.905, and a precision of 0.795 on the internal validation set. Additionally, the model achieved an AUC of 0.883 on the prospective validation set and AUCs of 1.000, 0.927 and 0.905 on the three external validation sets, respectively. We developed a user-friendly web-based calculator for clinical use, available at http://175.178.66.58/english/. By utilizing the UNION-SFTS model, clinicians can promptly predict and monitor the disease severity and mortality risk of SFTS patients, enabling early intervention in severe cases and ultimately reduces patient mortality.