Infection curve flattening via targeted interventions and self-isolation

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

TL;DR

This paper explores how the structural properties of "small-world" networks—specifically clustering and path length—influence the success of flattening the infection curve. By simulating targeted self-isolation on Watts-Strogatz networks, the authors find that path-based centrality measures (like closeness and betweenness) are significantly more effective at reducing epidemic peaks than simple degree-based strategies.

What Problem Does This Paper Address?

The "flatten the curve" strategy is a cornerstone of public health, aiming to spread out infections over time to prevent healthcare systems from being overwhelmed. While previous research has focused heavily on "scale-free" networks (where a few highly connected hubs dominate), real human social networks often behave as "small-world" networks. These are characterized by high local clustering (your friends are likely friends with each other) and short path lengths (the "six degrees of separation" phenomenon).

The authors set out to determine how these specific topological features—clustering and small-worldness—affect the rate of disease spread and, crucially, which mathematical metrics are best for identifying the most important individuals to isolate to achieve a flatter curve.

Methods

The researchers employed a multi-step simulation approach:

  • Network Modeling: They used the Watts-Strogatz (WS) model to generate networks. By tuning the "rewiring probability" ($\beta$), they could transition a network from a regular lattice (high clustering, long paths) to a random graph (low clustering, short paths), capturing various degrees of "small-worldness."
  • Infection Simulation: They used a distance-based model where an infection spreads from a seed node to all its neighbors in one time step, then to their neighbors, and so on. The "infection curve" is essentially the histogram of shortest-path distances from the source.
  • Targeted Intervention: They simulated "self-isolation" by removing a percentage of nodes (14-15%) from the network.
  • Centrality Metrics: To choose which nodes to isolate, they compared seven different metrics: Degree, Betweenness, Closeness, Katz, PageRank, Expected Force, and Eigenvector centrality.
  • Curve Fitting: They fitted a Gamma distribution to the resulting infection data, using the scale parameter ($b$) as a mathematical proxy for "flatness."

Datasets

The study primarily relies on synthetic networks generated via the Watts-Strogatz model to maintain strict control over network parameters ($n=500, k=6$ and $n=150, k=8$). While they acknowledge the use of MATLAB codes from the Koblenz Network Collection (KONECT), no specific real-world social dataset was used for the primary simulations.

Key Insights and Ideas

  • The Clustering Barrier: High clustering can act as a natural barrier. If connections between clusters are sparse, the virus may spread quickly within a group but struggle to jump to the next, naturally slowing the overall progression and flattening the curve.
  • Small-Worldness as an Accelerator: "Small-worldness" (the presence of long-range shortcuts) is the enemy of a flat curve. These shortcuts allow the virus to bypass local clusters and reach distant parts of the network rapidly, leading to sharp, high-peak outbreaks.
  • Path-Based Superiority: In small-world networks, "who you know" (Degree) is less important for intervention than "where you sit" in the network. Metrics that account for the shortest paths between all nodes (Closeness and Betweenness) were the most effective at identifying the "bridge" individuals whose isolation most drastically flattens the curve.

Key Results

  • Quantitative Flattening: In a network with low rewiring ($\beta=0.025$), isolating 15% of nodes based on Closeness reduced the peak number of simultaneous infections from 2104 (no isolation) to 870—a reduction of nearly 60%.
  • Metric Comparison: Path-based measures (Closeness, Betweenness) consistently outperformed degree-based measures (Degree, PageRank, Katz). For example, at $\beta=0.025$, Degree-based isolation only reduced the peak to 1066, significantly less effective than Closeness.
  • Structural Correlation: The results confirmed that higher clustering ($C$) and lower small-worldness ($S_2$) lead to higher $b$ values in the Gamma distribution, indicating a flatter, more manageable infection curve.

Limitations and Future Work

  • Deterministic Spread: The model assumes a 100% infection probability between neighbors at every step. In reality, transmission is probabilistic and depends on various biological and environmental factors.
  • Static Topology: The simulations assume the network structure is fixed (aside from the initial isolation). Real-world networks are dynamic, with links forming and breaking over time.
  • Future Directions: The authors suggest moving toward link isolation (modeling social distancing rather than total quarantine) and incorporating probabilistic distance distributions to better mirror the uncertainty of real-world epidemics.

Summary updated

Abstract

Understanding the impact of network clustering and small-world properties on epidemic spread can be crucial in developing effective strategies for managing and controlling infectious diseases. Particularly in this work, we study the impact of these network features on targeted intervention (e.g., self-isolation and quarantine). The targeted individuals for self-isolation are based on centrality measures and node influence metrics. Compared to our previous works on scale-free networks, small-world networks are considered in this paper. Small-world networks resemble real-world social and human networks. In this type of network, most nodes are not directly connected but can be reached through a few intermediaries (known as the small-worldness property). Real social networks, such as friendship networks, also exhibit this small-worldness property, where most people are connected through a relatively small number of intermediaries. We particularly study the epidemic curve flattening by centrality-based interventions/isolation over small-world networks. Our results show that high clustering, while having low small-worldness (higher shortest path characteristics) implies flatter infection curves. In reality, a flatter infection curve implies that the number of new cases of a disease is spread out over a longer period of time, rather than a sharp and sudden increase in cases (a peak in epidemic). In turn, this reduces the strain on healthcare resources and helps to relieve the healthcare services.
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