Machine learning model for predicting hepatitis C seroconversion in methadone maintenance patients in China.

Area of Science:

• Infectious Diseases
• Public Health
• Data Science

Background:

• Hepatitis C virus (HCV) poses a significant public health challenge, especially for individuals with opioid addiction undergoing methadone maintenance treatment (MMT).
• While MMT programs reduce HIV transmission, HCV infection risk remains elevated in these settings.
• Accurate prediction of HCV seroconversion is vital for effective patient management and harm reduction.

Purpose of the Study:

• To develop and validate a predictive model for HCV seroconversion among individuals in MMT programs.
• To identify key factors contributing to HCV risk in this population.
• To create a tool for early identification of high-risk individuals.

Main Methods:

• Data from 1547 individuals in Shanghai and 283 in Mianyang, China, undergoing MMT were analyzed.
• Thirteen predictive factors were incorporated into four machine learning models.
• The eXtreme Gradient Boosting (XGBoost) model was developed and validated using internal and external cohorts.

Main Results:

• The XGBoost model demonstrated superior performance in predicting HCV seroconversion.
• Achieved high C-indices: 0.793 (training), 0.744 (internal validation), and 0.756 (external validation).
• A publicly accessible web tool was created based on the validated model.

Conclusions:

• The developed XGBoost model accurately predicts individuals in MMT programs at high risk of HCV seroconversion.
• This tool can support targeted interventions and improve health outcomes for opioid-addicted populations.
• The findings highlight the potential of machine learning in public health surveillance and risk prediction.