MetaWards: A flexible metapopulation framework for modelling disease spread

Abstract

Understanding how disease spreads through populations is important when designing and implementing control measures. MetaWards implements a stochastic metapopulation model of disease transmission that enables geographical modelling of disease spread that can scale all the way from modelling local transmission up to full national-or international-scale outbreaks. It is built in Python and has a flexible plugin architecture to support complex scenario modelling. This enables the code to be adapted to model new situations and new control measures as they arise, e.g. emergence of new variants of disease, enaction of different types of movement restrictions, availability of different types of vaccines etc. It implements a userdefinable compartmental transmission model, such as an SIR model, that can be extended multi-dimensionally via multiple demographics or sub-populations, and multiple geographical regions. Models can be constructed from the various sources of movement and demographic data that are available, and are accelerated via Cython (Behnel et al., 2020), OpenMP, Scoop (Hold & Gagnon, 2019) and MPI4Py (Dalcin & Fang, 2021) to scale efficiently from running on personal laptops to large supercomputers. Python, R and command line interfaces and a complete set of tutorials empower researchers to adapt their models to a variety of scenarios.