A Model for the Spread of Infectious Diseases in a Region

Elizabeth Hunter, Brian Mac Namee, John D. KelleherInternational Journal of Environmental Research and Public Health 2020 · DOI

TL;DR

This paper scales up an existing agent-based model of infectious disease spread from a single town to a regional network of towns. They demonstrate that a town's "centrality" within this network, particularly the number of people commuting into it, significantly influences its susceptibility to an outbreak, offering insights for public health interventions.

What Problem Does This Paper Address?

Understanding how infectious diseases spread is crucial for public health preparedness, especially in our increasingly interconnected world. While models exist for simulating disease spread within a single town, these often overlook the critical role of inter-town travel and connectivity. This paper addresses the gap by developing and validating a model that simulates disease spread across a network of towns, allowing for a more realistic assessment of how a town's position within a region impacts its vulnerability to outbreaks.

Methods

The authors adapt an existing agent-based model (ABM) for infectious disease spread, originally developed for a single town, to a regional scale. This involves several key methodological shifts:

  • Scaling Up the Environment: The detailed environmental representation of a single town was simplified to a single point per "small area" (a census unit) to manage computational complexity.
  • Constant Society: Instead of recreating the population for each simulation run, a constant synthetic population was used to improve comparability and reduce variability.
  • Gravity Model for Transportation: To better represent inter-town travel over larger distances, a gravity model was implemented, where movement probability is proportional to population density and inversely proportional to distance.
  • Network Analysis: The core innovation is the analysis of "centrality" within the town network. They investigate three types of centrality:
  • Total Degree Centrality: The overall number of connections (commuters) into and out of a town.
  • In-Degree Centrality: The number of commuters entering a town.
  • Out-Degree Centrality: The number of commuters leaving a town.
  • Closeness Centrality: A measure of how close a town is to all other towns in the network, considering geographic distance.

The model was validated by comparing simulation results of isolated towns to the scaled-up regional model and by analyzing outbreak patterns under different starting conditions and centrality measures.

Datasets

The study utilizes publicly available data for its model:

  • Irish Census Data (CSO): Used to create the synthetic population, including age, sex, household size, and economic status at the small area level.
  • Irish Vaccination Data: Used to determine the percentage of each age group vaccinated for measles.
  • CSO Place of Work, School or College - Census of Anonymity Records (POWSCAR): This dataset was crucial for modeling commuting patterns between electoral divisions (one level above small areas).

Key Insights and Ideas

  • Commuters Matter (A Lot): The most striking finding is that the number of people commuting into a town (in-degree centrality) has a significantly greater impact on whether an outbreak will spread to that town than the number of people commuting out (out-degree centrality). This suggests that focusing interventions on controlling incoming traffic might be more effective.
  • Centrality is Key: A town's position within the regional network, specifically its connectivity, strongly influences its susceptibility to outbreaks. Highly connected towns are more likely to experience outbreaks.
  • Outbreak Origin Matters: Where an outbreak starts within the network significantly affects its spread. Outbreaks originating in highly central towns can have a different impact than those starting in more isolated locations.
  • Distance and Centrality Interact: While closeness centrality (geographic proximity) is important, its effect can be modulated by a town's degree centrality. A close town with high degree centrality is a prime candidate for rapid spread.

Key Results

  • The scaled-up regional model showed comparable results to the original town-isolated model, validating the scaling assumptions.
  • In-degree centrality demonstrated a moderate to strong positive correlation with the likelihood of an outbreak in a town, across various outbreak starting locations.
  • Out-degree centrality showed a much weaker correlation with outbreak likelihood, suggesting that outgoing commuters have less influence on a town's susceptibility compared to incoming ones.
  • When an outbreak starts in a highly central town (e.g., Manorhamilton), the centrality of other towns becomes less influential on the outbreak's progression. Conversely, when an outbreak starts in a less central town, the centrality of other towns plays a larger role.
  • Distance plays a role: outbreaks are more likely to spread to nearby towns, especially if those nearby towns also have high degree centrality.

Limitations and Future Work

  • Model Simplification: While necessary for computational feasibility, the reduction in environmental fidelity (e.g., representing small areas as single points) might abstract away some finer-grained spatial dynamics.
  • Disease Specificity: The model is currently parameterized for measles. While adaptable, further work could explore modeling diseases with different transmission dynamics (e.g., COVID-19's varying severity).
  • Intervention Simulation: The authors suggest future work could involve simulating specific interventions (e.g., school closures, travel restrictions) to assess their effectiveness based on the model's findings.
  • Broader Geographic Scope: The current model focuses on a single county. Extending it to model multiple counties or an entire country would provide a more comprehensive understanding of regional disease spread.
  • Data Limitations: While using open data is a strength, the model's realism is still dependent on the quality and granularity of available data.

Summary updated

Abstract

In understanding the dynamics of the spread of an infectious disease, it is important to understand how a town's place in a network of towns within a region will impact how the disease spreads to that town and from that town. In this article, we take a model for the spread of an infectious disease in a single town and scale it up to simulate a region containing multiple towns. The model is validated by looking at how adding additional towns and commuters influences the outbreak in a single town. We then look at how the centrality of a town within a network influences the outbreak. Our main finding is that the commuters coming into a town have a greater effect on whether an outbreak will spread to a town than the commuters going out. The findings on centrality of a town and how it influences an outbreak could potentially be used to help influence future policy and intervention strategies such as school closure policies.
BibTeX
@article{hunter2020model,
  author = {Elizabeth Hunter and Brian Mac Namee and John D. Kelleher},
  title = {{A Model for the Spread of Infectious Diseases in a Region}},
  journal = {International Journal of Environmental Research and Public Health},
  year = {2020},
  doi = {10.3390/ijerph17093119},
}
Figure