OpenABM-Covid19

OpenABM-Covid19 simulates the spread of COVID-19 through a city. Each simulated person belongs to a household and a workplace and also meets random contacts, and the virus passes along these links.¹ The model follows each infected person through the course of the disease.¹ Its distinctive feature is contact tracing: it can trace and quarantine the contacts of a case instantly.¹ The team used it to judge how a phone contact-tracing app should be configured.¹

It was built by the Pathogen Dynamics group at the Big Data Institute, University of Oxford, with IBM UK and the data science company Faculty.¹ The model paper, by Hinch and colleagues, appeared in PLOS Computational Biology in July 2021.² By default it is set up with UK demographics and calibrated to the UK epidemic, and it can be re-parameterised for other countries.² It simulates a million people at a few seconds per simulated day, fast enough to run many variations and to fit the model to data statistically.² A companion model, adaptER-covid19, estimates the economic effects alongside the epidemic.¹

The core is written in C and needs a compiler and the GNU Scientific Library; it is best run from Python, and it also has an R interface.¹ Inputs are a table of parameters and a table of household make-ups to sample from, and the repository documents both, along with every output file and the model's test suite.¹ Example notebooks show a lockdown triggered by case numbers followed by app-based tracing.¹

The last tagged release is version 1.0 from April 2021.³ It remains a clear, well-documented example of an epidemic model built around social networks. The code is GPL-3.0.¹ For a maintained alternative in Python, see Covasim and Starsim.