Infection Curve Flattening via Targeted Interventions and Self-Isolation

Mohammadreza Doostmohammadian, Houman Zarrabi, Azam Doustmohammadian, Hamid R. Rabiee

Abstract

This paper investigates the relationship between social network topology and the efficacy of epidemic intervention strategies. By simulating disease spread on Watts-Strogatz small-world networks, the study evaluates how structural properties like clustering and path length influence the "flattening" of the infection curve. The research demonstrates that targeted self-isolation based on path-based centrality metrics—specifically closeness and betweenness—is significantly more effective at reducing peak infection rates than traditional degree-based strategies. These findings suggest that identifying "bridge" individuals in small-world social structures is critical for effective public health responses.

Introduction

The "flatten the curve" strategy has become a fundamental principle of public health, aimed at spreading infections over a longer duration to prevent healthcare systems from exceeding their capacity. While much epidemiological research has focused on scale-free networks, where a few highly connected hubs dominate transmission, real-world human social networks often exhibit "small-world" characteristics. These networks are defined by high local clustering and short average path lengths, often referred to as the "six degrees of separation."

This study addresses how these specific topological features—clustering and small-worldness—affect the rate of disease progression. Furthermore, it seeks to identify which mathematical centrality metrics are most effective at selecting individuals for self-isolation to achieve the most significant reduction in the epidemic peak.

Methodology

Network Modeling

The researchers utilized the Watts-Strogatz (WS) model to generate synthetic networks for simulation. By adjusting the rewiring probability ($\beta$), the researchers could transition the network from a regular lattice (characterized by high clustering and long paths) to a random graph (low clustering and short paths). This allowed for controlled experimentation across varying degrees of "small-worldness." Simulations were conducted on networks with parameters such as $n=500, k=6$ and $n=150, k=8$.

Infection Simulation and Curve Fitting

The study employed a distance-based infection model. In this deterministic framework, an infection originates at a seed node and spreads to all immediate neighbors in a single time step, continuing iteratively. The resulting infection curve represents a histogram of shortest-path distances from the source. To quantify the "flatness" of these curves, the authors fitted the data to a Gamma distribution, using the scale parameter ($b$) as a mathematical proxy for curve flatness.

Targeted Intervention Strategies

The researchers simulated "self-isolation" by removing a specific percentage of nodes (approximately 14-15%) from the network before the infection spread. To determine which nodes to remove, they compared seven distinct centrality metrics:

  • Degree Centrality (number of direct connections)
  • Betweenness Centrality (frequency of appearing on shortest paths)
  • Closeness Centrality (average distance to all other nodes)
  • Katz Centrality
  • PageRank
  • Expected Force
  • Eigenvector Centrality

Key Insights

The Role of Clustering and Small-Worldness

The simulations revealed that high local clustering can act as a natural barrier to viral spread. In networks where connections between clusters are sparse, a virus may spread rapidly within a local group but struggle to transition to other clusters, naturally slowing the overall progression. Conversely, "small-worldness"—the presence of long-range shortcuts—acts as an accelerator. These shortcuts allow the virus to bypass local clusters and reach distant parts of the network rapidly, resulting in sharp, high-peak outbreaks.

The Superiority of Path-Based Metrics

A primary finding of the study is that in small-world networks, "where you sit" in the network is more critical for intervention than "who you know." Metrics that account for the shortest paths across the entire network (Closeness and Betweenness) were the most effective at identifying "bridge" individuals. Isolating these individuals effectively breaks the shortcuts that characterize small-world networks, leading to a more flattened infection curve.

Results

  • Quantitative Impact: In a network with low rewiring ($\beta=0.025$), isolating 15% of nodes based on Closeness Centrality reduced the peak number of simultaneous infections from 2104 (in the control scenario) to 870—a reduction of nearly 60%.
  • Metric Performance: Path-based measures consistently outperformed degree-based measures. For instance, at $\beta=0.025$, isolation based on Degree Centrality only reduced the peak to 1066, proving significantly less effective than Closeness-based isolation.
  • Structural Correlation: The results confirmed a strong correlation between network topology and the Gamma distribution scale parameter ($b$). Higher clustering ($C$) and lower small-worldness ($S_2$) consistently led to higher $b$ values, indicating a flatter and more manageable infection curve.

Discussion and Limitations

While providing valuable insights into targeted interventions, the study acknowledges several limitations. The model assumes a deterministic 100% infection probability between neighbors, whereas real-world transmission is probabilistic and influenced by biological factors. Additionally, the simulations utilized a static topology, while real-world social networks are dynamic, with connections forming and dissolving over time.

Future research directions include moving toward link isolation (modeling social distancing rather than total quarantine) and incorporating probabilistic distance distributions to better reflect the inherent uncertainty of real-world epidemics.

Summary updated

BibTeX
@inproceedings{doostmohammadian2023infection,
  author = {Mohammadreza Doostmohammadian and Houman Zarrabi and Azam Doustmohammadian and Hamid R. Rabiee},
  title = {{Infection Curve Flattening via Targeted Interventions and Self-Isolation}},
  booktitle = {Soc. Netw. Anal. Min},
  year = {2023},
  url = {https://openreview.net/forum?id=6X1Z1S7yX3},
}
Figure