A Data-Driven Approach to Assessing Hepatitis B Mother-to-Child Transmission Risk Prediction Model: Machine Learning...

Area of Science:

• Virology
• Data Science
• Public Health

Background:

• Hepatitis B virus (HBV) poses a significant risk of mother-to-child transmission (MTCT).
• Accurate risk assessment is crucial for timely clinical decisions and effective prevention of HBV MTCT.
• Data mining offers powerful tools for identifying key diagnostic predictors.

Purpose of the Study:

• To develop a robust predictive model for HBV MTCT using decision tree algorithms (ID3 and CART).
• To identify key clinical and paraclinical predictors, specifically hepatitis B e antigen (HBeAg) status and peripheral blood mononuclear cell (PBMC) concentration.
• To assess model reliability and generalizability through cross-validation for enhanced clinical applicability.

Main Methods:

• Utilized decision tree algorithms (ID3 and CART) on a dataset of 60 HBsAg-positive pregnant women.
• Analyzed clinical and paraclinical parameters, focusing on HBeAg status and PBMC concentration.
• Validated predictive models using various training-test split ratios for robustness and generalizability.

Main Results:

• HBeAg-positive status was significantly associated with HBV MTCT risk (χ²=21.16, P=.007).
• Among HBeAg-negative women, PBMC concentration stratified MTCT risk (≥8 × 10⁶ cells/mL low risk, <8 × 10⁶ cells/mL negligible risk).
• HBeAg status and PBMC concentration consistently emerged as the most influential predictors across all models and validation ratios.

Conclusions:

• Decision tree models effectively stratify HBV MTCT risk using clinical and paraclinical markers.
• HBeAg status and PBMC concentration are critical predictors for HBV MTCT risk stratification.
• Findings support the development of targeted prevention strategies for HBV MTCT, with potential for future studies on treated populations.