Monotonic and Non-Monotonic Epidemiological Models on Networks

Contact networks can significantly change the course of epidemics, affecting the rate of new infections and the mean size of an outbreak. Despite this dependence, some characteristics of epidemics are not contingent on the contact network and are probably predictable based only on the pathogen. Here we consider SIR-like pathogens and give an elementary proof that for any network increasing the probability of transmission increases the mean outbreak size. We also introduce a simple model, termed 2FleeSIR, in which susceptibles protect themselves by avoiding contacts with infectees. The 2FleeSIR model is non-monotonic: for some networks, increasing transmissibility actually decreases the final extent. The dynamics of 2FleeSIR are fundamentally different from SIR because 2FleeSIR exhibits no outbreak transition in densely-connected networks. We show that in non-monotonic epidemics, public health officials might be able to intervene in a fundamentally new way to change the network so as to control the effect of unexpectedly-high virulence. However, interventions that decrease transmissibility might actually cause more people to become infected.

💡 Research Summary

The paper investigates how the structure of contact networks influences epidemic dynamics, focusing on two contrasting classes of models: the classic SIR‑like framework and a newly introduced “2FleeSIR” model that incorporates behavioral avoidance.
First, the authors provide an elementary yet rigorous proof that for any finite network, increasing the transmission probability (β) cannot reduce the expected final outbreak size. By representing an epidemic as a collection of possible infection paths on a graph, they show that each path’s activation probability is monotone in β, and therefore the expectation of the total number of infected nodes is also monotone. This result holds regardless of the network’s topology—regular lattices, Erdős‑Rényi random graphs, scale‑free networks, or any arbitrary adjacency structure. The theorem formalizes the intuitive notion that a more transmissible pathogen will always generate a larger average epidemic, even though the precise size depends heavily on the underlying contact pattern.
The second contribution is the definition of the 2FleeSIR model. In this variant, whenever a susceptible node becomes aware that a neighbor is infected, it instantly severs that edge (i.e., “flees”). The process is dynamic: each infection event may trigger a cascade of edge removals, continuously reshaping the network as the epidemic unfolds. The authors derive a set of stochastic update rules that capture this co‑evolution of disease spread and network topology.
Through analytical arguments and extensive simulations on several canonical network families, they discover a striking non‑monotonic phenomenon: in densely connected graphs (high average degree), raising β beyond a certain “escape threshold” actually reduces the expected final size. The intuition is that a highly transmissible pathogen quickly creates many infected nodes, which in turn cause a massive, simultaneous withdrawal of susceptible contacts. The resulting fragmentation of the network prevents the disease from reaching large portions of the population. Consequently, the classic epidemic phase transition—where a small increase in β triggers a sudden jump from minor outbreaks to widespread epidemics—disappears in the limit of very dense connectivity.
The paper formalizes the escape threshold as a function of the network’s mean degree and the initial number of infected seeds. When the mean degree is large, the threshold is low, so even modest transmissibility can trigger the avoidance cascade. Conversely, in sparse networks the avoidance effect is weak, and the model behaves much like the standard SIR, preserving monotonicity.
From a public‑health perspective, the findings overturn the conventional wisdom that lowering transmissibility is always beneficial. In a non‑monotonic setting, interventions that merely reduce β (e.g., mask mandates, partial vaccination) might inadvertently suppress the avoidance response, leading to a larger outbreak. Conversely, policies that deliberately increase perceived risk while simultaneously encouraging rapid contact reduction—such as targeted isolation of high‑risk groups or temporary closure of densely used venues—could exploit the avoidance mechanism to curb spread. The authors argue that controlling the network’s dynamic response, rather than only the pathogen’s biological parameters, opens a new strategic dimension for epidemic control.
The manuscript is organized as follows: an introduction outlining the motivation and prior work; a formal section proving monotonicity for SIR‑like models; a definition and mathematical description of 2FleeSIR; a results section presenting analytical derivations, simulation data, and the characterization of the escape threshold; a discussion of policy implications and potential pitfalls of traditional interventions; and finally a conclusion that highlights limitations (e.g., assumptions about instantaneous awareness and edge removal) and suggests avenues for future research, such as incorporating delayed behavioral responses or heterogeneous avoidance thresholds.
In summary, the paper makes three principal contributions: (1) a general proof of monotonicity for classic SIR on arbitrary networks; (2) the introduction of a behavior‑driven, non‑monotonic epidemic model that fundamentally alters outbreak dynamics in dense contact structures; and (3) a conceptual framework for designing public‑health interventions that leverage network‑level behavioral feedback rather than solely focusing on pathogen transmissibility. These insights advance both theoretical epidemiology and practical disease‑control strategies.