Environment, vector, or host? Using machine learning to untangle the mechanisms driving arbovirus outbreaks

Area of Science:

• Veterinary Epidemiology
• Disease Ecology
• Machine Learning in Public Health

Background:

• Vector-borne diseases like Bluetongue virus (BTV) pose significant economic threats and their outbreaks are shaped by complex environmental and host interactions.
• Understanding the specific drivers for different BTV serotypes is crucial for effective disease management, as relationships can be nonlinear and vary taxonomically.
• BTV is a re-emerging pathogen in Europe, necessitating advanced analytical approaches to predict and mitigate its impact.

Purpose of the Study:

• To investigate the serotype-specific ecological predictors of Bluetongue virus (BTV) outbreak risk in Europe.
• To untangle the complex, nonlinear relationships between environmental/host factors and BTV outbreaks for prevalent serotypes (1, 4, and 8).
• To develop and apply a machine learning pipeline for enhanced prediction and understanding of BTV epidemiology.

Main Methods:

• Utilized a machine learning (ML) pipeline with 23 environmental and host features.
• Analyzed 24,245 BTV outbreaks across 25 European countries from 2000 to 2019.
• Developed predictive models for BTV serotypes 1, 4, and 8, assessing serotype-specific risk profiles.

Main Results:

• Machine learning models achieved high predictive performance (accuracies > 0.87) for all BTV serotypes.
• Identified strong nonlinear relationships between BTV outbreak risk and environmental/host features.
• Demonstrated unique outbreak risk profiles for BTV serotypes 1, 4, and 8, with distinct key drivers (e.g., temperature, midge abundance, goat density) and serotype-specific interactive effects.

Conclusions:

• Ecological predictors and their interactions significantly influence BTV outbreak risk in a serotype-specific manner.
• The developed ML pipeline provides in-depth insights into BTV epidemiology, highlighting the need for tailored intervention strategies.
• Findings can guide policymakers in implementing targeted measures to reduce the economic and social costs associated with BTV outbreaks.