Why carry out this study?

Firstly, it is urgent to precisely predict the risk of severe fever with thrombocytopenia syndrome virus (SFTSV) developing into encephalitis and its mortality at the early stage of the illness, since there is a close association with encephalitis and high mortality rate (i.e., 12–50%).

Secondly, use of artificial intelligence is spreading rapidly in many areas due to its excellent performance. The effectiveness of artificial intelligence in SFTS-assisted diagnosis is unclear.

Our models contribute to the early accurate prognosis of SFTS, even with limited medical resources in underdeveloped areas.

What was learned from the study?

The reservoir computing with boosted topology (RC-BT) model that predicts the risk of developing encephalitis in patients with SFTS in our study contains nine clinical parameters at the admission and shows an area under curve of 0.899 (95% CI 0.882–0.916), a sensitivity of 0.855 (95% confidence interval [CI] 0.824–0.886), and a specificity of 0.859 (95% CI 0.831–0.887).

The RC–BT model for the prediction of the fatality of patients with SFTS in our study includes seven parameters and demonstrates an area under curve of 0.917 (95% CI 0.902–0.932), a sensitivity of 0.913 (95% CI 0.902–0.924), and a specificity of 0.884 (95% CI 0.851–0.917).