Machine-learning methodologies to predict disease progression in chronic hepatitis B in Africa

Area of Science:

• Hepatology
• Infectious Diseases
• Machine Learning in Medicine

Background:

• Determinants of chronic hepatitis B (HBV) infection progression in African populations remain understudied.
• Understanding disease progression is vital for timely intervention and management of HBV.

Purpose of the Study:

• To develop predictive algorithms for chronic HBV disease progression in Ethiopian patients.
• To identify key determinants of disease progression using machine learning.

Main Methods:

• Longitudinal data from Ethiopian patients with chronic HBV infection (without baseline liver fibrosis) were analyzed.
• Machine learning models, including random forest, were employed to establish predictive algorithms.
• Disease progression was defined by increased liver stiffness or treatment initiation.

Main Results:

• Four point four percent (24/551) of patients experienced disease progression over a median follow-up of 69 months.
• A random forest model utilizing standard hematology and biochemistry tests achieved the highest predictive accuracy (AUROC 0.82–0.88).
• These laboratory tests demonstrated strong predictive properties for disease progression.

Conclusions:

• Combined metrics from simple, readily available laboratory tests possess significant predictive value for HBV disease progression.
• Further validation in larger, diverse HBV cohorts is recommended to refine these predictive models.
• This approach offers a practical tool for monitoring HBV patients, particularly in resource-limited settings.