A multicenter study on developing a prognostic model for severe fever with thrombocytopenia syndrome using machine...

Area of Science:

• Infectious Diseases
• Machine Learning
• Medical Informatics

Background:

• Severe Fever with Thrombocytopenia Syndrome (SFTS) is a serious viral illness caused by the SFTS virus (SFTSV).
• Accurate prognosis is vital for tailoring SFTS prevention and treatment strategies.
• Existing machine learning prognostic models for SFTS require further development and clinical validation.

Purpose of the Study:

• To develop and validate an interpretable machine learning (ML) prognostic model for SFTS.
• To improve the understanding of SFTS disease progression through ML.

Main Methods:

• A multicenter retrospective study analyzed data from 292 patients for training/internal validation and 104 patients for external validation.
• The Boruta algorithm identified 12 candidate predictors from 24 clinical features, with 10 ML models developed.
• Model performance was evaluated using AUC, accuracy, recall, F1 score, and SHAP for feature importance.

Main Results:

• The XGBoost model exhibited the best discriminatory performance.
• A final interpretable XGBoost model with 7 key features achieved high predictive accuracy (Internal AUC: 0.911, External AUC: 0.891).
• A clinical decision support tool was developed using the Streamlit framework.

Conclusions:

• An interpretable XGBoost-based prognostic model for SFTS demonstrates high predictive accuracy.
• The model's 7 key features provide valuable indicators for early SFTS prognosis.
• The developed clinical tool supports healthcare professionals in managing SFTS.