A conformal regressor for predicting negative conversion time of Omicron patients

Area of Science:

• Infectious Diseases
• Computational Biology
• Epidemiology

Background:

• The Omicron variant of COVID-19 continues to spread globally and within China.
• Optimized prevention and control measures are in place, but predicting individual recovery remains challenging.
• Accurate estimation of viral negative conversion time is crucial for patient management and public health.

Purpose of the Study:

• To develop and validate a predictive model for estimating the negative conversion time of COVID-19 Omicron infections.
• To leverage clinical and symptom data for accurate prediction of viral clearance.
• To provide patients with a tool for self-estimating their recovery duration.

Main Methods:

• Retrospective study of Omicron-infected patients in Shandong Province (first half of 2022).
• Utilized eXtreme Gradient Boosting (XGBoost) model incorporating clinical diagnosis, signs, Traditional Chinese Medicine symptoms, and drug use.
• Implemented Conformal Prediction (CP) framework with XGBoost for controllable error rate probability interval estimation.

Main Results:

• The proposed XGBoost-CP model achieved a mean absolute error of 3.54 days for predicting negative conversion time.
• The model demonstrated the shortest interval prediction results, indicating high accuracy and reliability.
• The prediction intervals offer controllable error rates, enhancing decision-making information.

Conclusions:

• The developed model accurately predicts COVID-19 negative conversion time for Omicron variant infections.
• This tool empowers individuals to better understand their disease course and self-estimate recovery.
• The findings support the integration of predictive modeling in managing infectious diseases and patient care.