Contact Graph Epidemic Modelling of COVID-19 for Transmission and Intervention Strategies

Abby Leung, Xiaoye Ding, Shenyang Huang, Reihaneh RabbanyarXiv 2020 · arXiv

Introduction

Classical epidemiological models, such as the Susceptible-Exposed-Infected-Recovered (SEIR) framework, typically rely on the assumption of "homogeneous mixing." This assumption posits that any individual in a population is equally likely to interact with any other. While this simplification makes differential equations easier to solve, it fails to capture the nuances of human behavior. In reality, humans move in clusters, frequent specific "hubs" like schools or offices, and maintain varying numbers of social connections.

When public health officials use simplified models to plan non-pharmaceutical interventions (NPIs), they often struggle to simulate the structural impact of specific choices, such as closing high-traffic venues. This paper introduces the Contact Graph Epidemic Model (CGEM), a stochastic simulation framework designed to bridge the gap between theoretical populations and the complex, graph-like structure of real-world human contact.

Methodology: The CGEM Framework

CGEM operates on a contact graph where nodes represent individuals and edges represent potential transmission paths. Unlike traditional models that use continuous differential equations, CGEM employs an algorithmic approach to state transitions.

Graph-based SEIR transitions

At each discrete time step, an infected node has a specific probability ($\phi$) of infecting its susceptible neighbors. To ensure the model remains grounded in empirical data, the authors calibrate CGEM by fitting a standard SEIR model to real-world infection data from Montreal to determine the transmission rate ($\beta$), which is then mathematically mapped to the edge-level probability ($\phi$).

Simulating Interventions

CGEM allows for the granular simulation of various NPIs by modifying the graph structure or transmission probabilities:

  • Quarantine: Removing edges from a subset of identified infected or exposed nodes.
  • Social Distancing: Randomly removing a percentage of all edges across the network.
  • Mask Mandates: Reducing the transmission probability ($\phi$) on edges based on masking compliance.
  • Closures (Hub Removal): Strategically removing "hubs"—nodes with the highest degree of connectivity—to simulate the closure of schools or workplaces.

Datasets and Network Structure

The study evaluates CGEM across several network types to determine how underlying structure influences disease progression. These include synthetic networks (Erdos-Renyi, Regular, and Barabasi-Albert) and real-world proxies, most notably a novel Montreal Wifi dataset. This dataset uses public Wifi hotspot connections as a proxy for physical proximity; if two users connected to the same hub simultaneously, an edge was established between them.

ER vs. Montreal Wifi Networks

As shown in the comparison above, the Erdos-Renyi (ER) graph—which mirrors the "well-mixed" assumption—displays a uniform, dense distribution of connections. In contrast, the Montreal Wifi network reveals a highly clustered and sparse structure, which significantly alters the dynamics of viral spread.

Key Insights and Results

The "ER" Overestimation

The research demonstrates that assuming an ER graph structure (the default for many standard models) can overestimate the infection peak by more than 3x compared to realistic networks like the Montreal Wifi graph. Because real networks are more fragmented, the virus can become "trapped" in specific communities rather than spreading uniformly.

The Power of Hubs

In real-world networks, a small fraction of "hubs" are responsible for the vast majority of transmission. The study found that removing just the top 1% of these hubs is incredibly effective in realistic networks, whereas it has almost no impact in the "well-mixed" models used by many agencies.

Reopening and the Second Wave

One of the most critical findings concerns the scale of resurgence after NPIs are lifted. Traditional models often predict a manageable "bump" upon reopening. However, CGEM simulations show that reopening hubs (like schools) can trigger a second wave with a peak higher than the initial wave. This occurs because the underlying network structure allows the virus to persist in clusters, leading to an explosive resurgence once high-connectivity hubs are reintroduced.

Conclusion

The Contact Graph Epidemic Model provides a more accurate lens for hospital capacity planning and NPI evaluation. By accounting for the non-random nature of human contact, CGEM reveals that interventions targeting high-connectivity hubs are far more effective than previously thought, while also warning that the risks of reopening such hubs may be significantly underestimated by traditional homogeneous models.

Summary updated

Abstract

The coronavirus disease 2019 (COVID-19) pandemic has quickly become a global public health crisis unseen in recent years. It is known that the structure of the human contact network plays an important role in the spread of transmissible diseases. In this work, we study a structure aware model of COVID-19 CGEM. This model becomes similar to the classical compartment-based models in epidemiology if we assume the contact network is a Erdos-Renyi (ER) graph, i.e. everyone comes into contact with everyone else with the same probability. In contrast, CGEM is more expressive and allows for plugging in the actual contact networks, or more realistic proxies for it. Moreover, CGEM enables more precise modelling of enforcing and releasing different non-pharmaceutical intervention (NPI) strategies. Through a set of extensive experiments, we demonstrate significant differences between the epidemic curves when assuming different underlying structures. More specifically we demonstrate that the compartment-based models are overestimating the spread of the infection by a factor of 3, and under some realistic assumptions on the compliance factor, underestimating the effectiveness of some of NPIs, mischaracterizing others (e.g. predicting a later peak), and underestimating the scale of the second peak after reopening.
BibTeX
@misc{leung2020contact,
  author = {Abby Leung and Xiaoye Ding and Shenyang Huang and Reihaneh Rabbany},
  title = {{Contact Graph Epidemic Modelling of COVID-19 for Transmission and Intervention Strategies}},
  year = {2020},
  eprint = {2010.03081},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2010.03081},
}
Figure