Optimizing wild bird fecal surveillance sites for HPAI through risk estimation using machine learning algorithms.

Area of Science:

• Veterinary epidemiology
• Wildlife disease ecology
• Biosecurity and risk assessment

Background:

• Wild birds are primary reservoirs for highly pathogenic avian influenza (HPAI), facilitating its spread.
• Current fecal-based surveillance in South Korea lacks quantitative optimization for early HPAI detection in poultry.
• Inefficient surveillance risks HPAI introduction into poultry populations.

Purpose of the Study:

• To develop a quantitative, risk-based strategy for optimizing wild bird fecal sampling sites.
• To enhance the early warning system (EWS) capacity for HPAI detection.
• To strengthen biosecurity measures and prevent HPAI transmission to poultry farms.

Main Methods:

• Utilized machine learning algorithms (logistic regression, GBM, XGBoost, ensemble model) to predict AIV risk.
• Integrated wild bird fecal sampling data with climatic, population, and elevation factors across 110 surveillance grids (2021-2024).
• Evaluated model performance using AUC, sensitivity, specificity, and accuracy.

Main Results:

• An ensemble model demonstrated superior predictive performance for avian influenza virus (AIV) detection.
• Key predictors included climatic factors, sample size, population density, and elevation.
• Significant spatial discrepancies were found between current and proposed risk-based sampling sites.

Conclusions:

• The proposed risk-based framework optimizes fecal sampling site selection for HPAI surveillance.
• Enhanced EWS efficiency is expected, improving the prevention of HPAI spread to poultry.
• This strategy provides a quantitative basis for improving wildlife disease surveillance.