SimInf

SimInf for large stochastic models

I've recently been working with SimInf, a software package for simulating large-scale disease outbreaks. It links many smaller populations, each with their own random infection patterns. The software is well-designed and performs very efficiently, far exceeding my expectations.

Often, disease modelling tools work fine for small examples or teaching, but they quickly struggle when scaled up or used in more realistic, complex situations. SimInf stands out as an exception, handling larger scenarios reliably.

SimInf is developed by Stefan Widgren at Sweden's National Veterinary Institute. At Wageningen University's Infectious Disease Epidemiology group, (former) PhD students You Chang and Mariken de Wit have successfully used it to model Bovine Tuberculosis and the Usutu virus.

I began using SimInf for the Erraze Pandemic Preparedness project. After creating an initial framework for modelling wildlife and livestock diseases across the Netherlands -work presented at the SVEPM 2024 conference -, we adapted it to the 2023 bluetongue virus outbreak in the Netherlands. 

At the national scale, the simulation involves about 18,000 grid cells, 130 million events (we are blowing midges around in wind plumes across the Netherlands), 3.7 million cattle, 1.1 million sheep, and roughly 7.5 billion midges. Compiling the model took around one minute, and running a full epidemic scenario for four months took less than five minutes.