Assessment of Diabetes Risk in Patients With Hepatitis B: A Machine Learning Approach Integrating 11 Inflammatory and...

Area of Science:

• Medical Informatics
• Immunology
• Endocrinology

Background:

• Chronic hepatitis B virus (HBV) infection elevates diabetes risk, yet early detection is difficult due to asymptomatic progression.
• Diabetes complications like cardiovascular disease, renal failure, and neuropathy affect 30-50% of patients, underscoring the need for timely risk assessment.

Purpose of the Study:

• To develop and validate a machine learning model for early diabetes risk assessment in adults with chronic HBV infection.
• To integrate novel inflammation-immune composite indicators with traditional parameters for improved predictive accuracy.

Main Methods:

• A machine learning model was developed using data from 14,287 HBV-positive adults (2003-2018 NHANES).
• Eleven inflammation-immune indicators, demographic, and biochemical data were analyzed using LASSO selection and seven algorithms.
• Model performance was assessed via ROC curve (AUC), calibration, and decision curve analysis (DCA).

Main Results:

• The Artificial Neural Network (ANN) model demonstrated optimal performance for diabetes risk assessment.
• The ANN achieved an AUC of 0.83 and an accuracy of 82%, indicating strong predictive capability.
• SHAP analysis revealed age and UHR as key predictors, with HRR exhibiting a protective effect.

Conclusions:

• A robust and interpretable machine learning model was developed for diabetes risk assessment in HBV patients.
• The model effectively integrates novel inflammation-immune markers, offering potential for clinical application in screening undiagnosed diabetes.