Agent-based models of disease transmission in a small population
Agent-based models of infectious disease propagation have become a popular industry in the last few years. However, in order to properly disentangle the effects of social processes and spatial constraints from physiological mechanisms of disease spread, we need modelling of disease transmission at the scale of individuals and their day-to-day activities, including the effects of real-world spatial geometries. This presents a number of challenges that have yet to be thoroughly addressed in many agent-based models.
In the first place we have to deal with the semantic use of space, rather than simply its geometry: the reasons why people go to particular places and the manner of their use have to be included in the model. In addition, the way in which generic descriptions of places are interpreted by agents as references to specific locations requires attention (e.g. an instruction to “go home” has different meanings for different agents).
Furthermore, model time stepping needs both to include sufficient resolution to deal with individual physical interactions and event-based scheduling to govern the changes in movement patterns that occur as a result of routine daily activity.
Finally, the way in which social-network structures condition spatial patterns of movement and spatial aggregation need to be simulated, mediated by these interpretations of spatial position.
This project uses data gathered in a small primary school to investigate how we can begin to understand the interaction between social use of space and the transmission of disease. A school is an ideal place to attempt this kind of investigation as it is both a very closed environment during the school day, and has a highly controlled set of activities that agents can undertake.
Records of absences from school for medical reasons are available for five years – these show various structures, including class-to class transmission in some cases, but not in others.
Using a laser scanner a complete geometry of the school layout has been obtained in 3 dimensions.
In the initial model implementation this has been simplified to leave just walls and desks.
Virtual children and teachers navigate the building, avoiding obstacles and each other, whilst following the daily timetable of school events.
Infected agents pass on disease based on proximity and contact time – shown here marked with red spheres. Walls are outlined in blue and school desk in green. The grid show 1m squares for reference.
As the day progresses, time-table events are communicated to teachers (a virtual “school bell”). Teachers can then modify this basic schedule by adding their own local class-based variations to allow for different in-class activities. Individual children may then further add variations in behaviour to allow for their own requirements (e.g. a request to go the toilet). As the day progresses these activities lead to structured patterns of contacts between children that mediate the spread of disease from individual to individual.
We are grateful to the head teacher, school governors and staff who gave permission for and facilitated the data collection exercise. We would also like to thank E.G. Bithell for assistance during 2006/7 and subsequent discussions.
- Bithell, M. And Brasington,J. “Small Scale Modelling of infectious disease transmission” IGU conference on Applied GIS and Spatial Modelling, Leeds 2013