Capacity of a Bayesian model to detect infected herds using disease dynamics and risk factor information from...

Area of Science:

• Veterinary Epidemiology
• Infectious Disease Control
• Statistical Modeling in Animal Health

Background:

• Control programs for non-regulated infectious diseases in farm animals are common.
• Varying definitions of 'freedom from infection' complicate international animal trade.
• Accurate identification of infection-free herds is crucial for safe animal trading when diseases persist.

Purpose of the Study:

• To evaluate a Bayesian Hidden Markov Model (BHMM) for estimating herd-level infection probabilities.
• To compare the BHMM's ability to detect infected herds against using test results solely.
• To assess the model's performance across diverse infection scenarios and test characteristics.

Main Methods:

• Simulated longitudinal data including herd-level risk factors and test results.
• Employed a Bayesian Hidden Markov Model to predict herd infection status.
• Evaluated model performance using simulated herd status as the gold standard.

Main Results:

• The BHMM detected more infected herds than single tests in 85% of converged scenarios.
• The model identified 20-40% of test-undetected infected herds, particularly with low test sensitivity.
• Model convergence issues arose in 39% of scenarios, primarily linked to low herd test sensitivity.

Conclusions:

• The BHMM offers significant benefits for animal disease control programs, especially with endemic infections and low-sensitivity tests.
• The model enhances the identification of infected herds, improving the reliability of 'freedom from infection' classifications for trade.
• While improving detection, the model may increase false positives, though this is context-dependent.