Agent-Based Modeling in Epidemiology

Magdalena Cerda and Katherine Keyes

ABMs consist of simulations that follow prescribed rules about the actions of agents that interact with each other within a user-defined environment. They have been well-developed for simulating infectious disease pathogenesis and control and are increasingly used to provide insight into the mechanisms and potential control strategies for so-called “non-infectious” health outcomes too, particularly those with significant social dynamics such as violence, substance use, and obesity.  ABMs provide a useful adjunct to existing data sources, because they can estimate the consequences that changes in one aspect of the system can have on other system components, in ways that are difficult to estimate using existing data sources. Further, ABMs can be used to project what ‘could’ happen to health outcomes after the implementation of policies and intervention programs, given assumptions about existing regulations, socio-demographic characteristics, and dynamics in the existing environment, among other features. Finally, by pacing interconnected networks of agents in a spatial context with its own heterogeneous set of features, ABMs make it possible to project what the expected effect of policies and interventions would be, under different assumptions about local conditions and the spatial distribution of existing prevention and intervention efforts. However, the strong assumptions required for identification of effects in ABMs render some research questions not suitable for these types of models.

In this playlist, we provide a conceptual overview of papers that have addressed the definition and uses of agent-based models within epidemiology, as well as some key empirical demonstrations of agent-based modeling in the epidemiological literature. 


Leave a Reply