[Paper Review] An Agent-Based Model for Bovine Viral Diarrhea

Jason Bassett, Pascal Blunk+3 more|arXiv (Cornell University)|Dec 14, 2018

Animal Disease Management and Epidemiology8 references4 citations

TL;DR

This paper presents a stochastic, event-driven, agent-based model simulating Bovine Viral Diarrhea (BVD) transmission across German cattle herds, integrating farm-level dynamics, animal movement, and testing/vaccination protocols. The model replicates real-world herd structures and trade networks, enabling simulation of policy interventions—such as testing, vaccination, and the Young Calf Window strategy—demonstrating their impact on reducing persistently infected (PI) cattle prevalence with high fidelity to German agricultural data.

ABSTRACT

We present an exhaustive description of a stochastic, event-driven, hierarchical agent-based model designed to reproduce the infectious state of the cattle disease called Bovine Viral Diarrhea, for which the livestock-trade network is the main route of spreading. For the farm-node dynamics, it takes into account a vast number of breeding, infectious and animal movement mechanisms via a susceptible-infected-recovered type of dynamics with an additional permanently infectious class. The interaction between the farms is described by a supply-demand farm manager mechanism governing the network structure and dynamics. The model includes realistic disease and breeding dynamics and allows to study numerous mitigation strategies of present and past government regulations, including different testing and vaccination scenarios.

Motivation & Objective

• To develop a detailed, data-driven agent-based model of BVD transmission in German cattle herds, incorporating realistic farm, herd, and animal-level dynamics.
• To simulate the impact of various mitigation strategies—such as testing, vaccination, and the Young Calf Window protocol—on reducing the prevalence of persistently infected (PI) cattle.
• To support cost-benefit analysis of BVD eradication policies by modeling intervention scenarios under realistic network and demographic conditions.
• To provide a simulation framework aligned with the ODD protocol for transparency and reproducibility, using real German livestock data from the HI-Tier database.
• To enable prediction of policy outcomes under different timing and frequency of interventions, including national-scale regulatory changes.

Proposed method

• The model uses a hierarchical, event-driven, stochastic agent-based framework with four levels: System, Farm, Herd, and Animal, each with distinct state variables and event triggers.
• Animal-level dynamics include susceptible-infected-recovered (SIR) with a persistent infection (PI) class, maternal antibody protection, pregnancy, and breeding events, all governed by time-distributed parameters.
• Farm-level interactions are governed by a supply-demand mechanism that simulates animal trade and movement, with realistic farm size distributions derived from the HI-Tier database.
• The model incorporates 13 distinct intervention scenarios, including baseline (no control), old and new national regulations, vaccination, and Young Calf Window (YCW) testing at semi-annual or annual frequencies.
• Events are prioritized and scheduled by time, with key durations (e.g., maternal antibody protection, pregnancy, infectious periods) drawn from triangular distributions based on empirical data.
• The simulation runs for 20,000 days (~55 years), with interventions applied after the system reaches equilibrium, enabling comparison of long-term policy effects.

Experimental results

Research questions

• RQ1 How does the inclusion of realistic farm-level trade and movement dynamics affect the predicted spread of BVD in a national cattle population?
• RQ2 What is the relative effectiveness of different testing and vaccination strategies—such as the Young Calf Window protocol—in reducing the prevalence of persistently infected (PI) cattle?
• RQ3 How do timing and frequency of interventions (e.g., semi-annual vs. annual testing) influence the long-term reduction of PI prevalence?
• RQ4 What is the impact of combining vaccination with targeted testing strategies on BVD eradication efficiency?
• RQ5 How well can the model reproduce historical trends in PI prevalence, particularly around the 2011 regulatory shift, given the lack of nationwide pre-2011 data?

Key findings

• The model successfully reproduces equilibrium dynamics of BVD in German cattle herds using real farm size distributions from the HI-Tier database, ensuring high realism.
• The Young Calf Window (YCW) strategy with semi-annual testing (Scenario 5a) significantly reduces PI prevalence compared to annual testing, demonstrating the importance of testing frequency.
• Combining the new national regulation with vaccination and semi-annual YCW testing (Scenario 6a) leads to the most effective reduction in PI prevalence across all scenarios.
• The baseline scenario (no intervention) stabilizes at a fixed point after ~10,000 days, indicating system equilibrium, which is used as the starting point for all intervention strategies.
• The model’s simulation timeline maps the 10,000th day to the 2011 enforcement of the old national regulation, though this is speculative due to lack of pre-2011 nationwide records.
• The model enables cost-benefit analysis of policies by isolating the impact of each intervention strategy, with Scenario 12 and 13 showing that combining YCW and vaccination yields the most favorable outcomes.