Vaccination intervention on epidemic dynamics in networks

Vaccination is an important measure available for preventing or reducing the spread of infectious diseases. In this paper, an epidemic model including susceptible, infected, and imperfectly vaccinated compartments is studied on Watts-Strogatz small-world, Barab'asi-Albert scale-free, and random scale-free networks. The epidemic threshold and prevalence are analyzed. For small-world networks, the effective vaccination intervention is suggested and its influence on the threshold and prevalence is analyzed. For scale-free networks, the threshold is found to be strongly dependent both on the effective vaccination rate and on the connectivity distribution. Moreover, so long as vaccination is effective, it can linearly decrease the epidemic prevalence in small-world networks, whereas for scale-free networks it acts exponentially. These results can help in adopting pragmatic treatment upon diseases in structured populations.

💡 Research Summary

The paper introduces a vaccination‑augmented SIS model (S‑I‑V‑S) that incorporates imperfect vaccination into the classic susceptible–infected–susceptible framework. Individuals can be in one of three states: susceptible (S), infected (I), or vaccinated (V). Vaccination occurs at a per‑capita rate ϕ, moving S → V, while waning immunity returns V → S at rate φ. Because vaccines are not perfectly protective, vaccinated nodes can still be infected, but at a reduced transmission rate δα, where δ (0 < δ ≪ 1) quantifies vaccine inefficacy. The dynamics are governed by five parameters: transmission α, recovery β, vaccination ϕ, waning φ, and inefficacy δ. Two dimensionless ratios are defined: λ = α/β (effective transmissibility) and η = ϕ/φ (vaccination‑to‑waning ratio).

The authors analyze this model on three canonical network topologies:

1. Watts‑Strogatz (WS) small‑world networks – characterized by a narrow degree distribution with average degree ⟨k⟩ = 2K. Using a homogeneous mean‑field (MF) approach, they derive coupled differential equations for the densities of S, I, and V. Setting time derivatives to zero yields an implicit equation for the steady‑state infected density ρ. The epidemic threshold is found to be

   λ_c = (η + 1)/(δ η + 1) · 1/⟨k⟩.

   This expression shows that increasing vaccination effort (higher η) or improving vaccine efficacy (smaller δ) raises the threshold, making outbreaks less likely. For λ > λ_c a unique endemic equilibrium (EE) exists; for λ < λ_c the disease dies out (disease‑free equilibrium, DFE). Moreover, when the product β φ (1 − δ) exceeds 4, a bistable region emerges: two endemic equilibria coexist with the DFE between a lower “persistence” threshold λ_b and the invasion threshold λ_c. This bistability produces hysteresis—whether the disease persists depends on initial conditions, a phenomenon absent in the classic SIS model.

2. Barabási‑Albert (BA) scale‑free networks – and more generally random scale‑free (SF) networks with degree distribution P(k) ∼ k^−γ (γ≈3). Because of the heterogeneous degree distribution, the authors employ a heterogeneous mean‑field (HMF) treatment, tracking the infection probability ρ_k for nodes of degree k. The epidemic threshold becomes

   λ_c ≈ (η + 1)/(δ η + 1) · ⟨k⟩/⟨k²⟩.

   In SF networks with γ ≤ 3, ⟨k²⟩ diverges as network size grows, causing λ_c → 0. Consequently, even infinitesimal transmissibility can sustain an epidemic, regardless of vaccination, unless η and δ are sufficiently large to offset the divergence. When vaccination is effective, the steady‑state prevalence decays exponentially with η: ρ ∼ exp(−C η), where C > 0 depends on network parameters. This contrasts with the linear decay ρ ∝ (λ − λ_c) observed in WS networks.

3. Numerical simulations – The authors validate the analytical predictions with extensive Monte‑Carlo simulations on networks of size N ≈ 10⁴–10⁵. In WS networks, simulation results reproduce the predicted λ_c, the linear increase of ρ above the threshold, and the bistable region for appropriate δ and η values. In SF networks, simulations confirm the near‑zero threshold and the exponential suppression of prevalence as vaccination effort rises. Even with a modest inefficacy δ = 0.001, increasing η from 1 to 10 reduces ρ by orders of magnitude.

Key insights and implications

• Network topology matters: Uniform networks (WS) exhibit a finite epidemic threshold that can be shifted by vaccination, while heterogeneous SF networks have vanishing thresholds, making them intrinsically more vulnerable.
• Vaccination efficacy vs. coverage: Both η (coverage) and δ (efficacy) appear together in the threshold expression. High coverage can compensate for moderate inefficacy, but in highly heterogeneous networks the product η·(1 − δ) must be large to achieve any meaningful control.
• Bistability and hysteresis: In WS networks, certain parameter regimes produce two stable endemic states separated by an unstable branch. Public‑health policies must therefore consider initial disease prevalence; aggressive early vaccination can push the system into the low‑prevalence basin.
• Exponential vs. linear suppression: The exponential reduction of prevalence in SF networks suggests that modest improvements in vaccine coverage or efficacy can have disproportionately large effects, a useful insight for resource‑limited settings.
• Model limitations: The study assumes static networks, homogeneous mixing within degree classes, and a simple “leaky” vaccine model. Real populations exhibit time‑varying contacts, behavioral responses, and multi‑strain dynamics, which are left for future work.

Conclusion

The paper provides a comprehensive analytical and computational investigation of how imperfect vaccination interacts with network structure to shape epidemic thresholds and steady‑state prevalence. It demonstrates that vaccination strategies cannot be designed in isolation from the underlying contact network. In homogeneous small‑world societies, vaccination raises the threshold linearly and may induce bistability, whereas in scale‑free societies, vaccination must overcome the intrinsic vulnerability of hub nodes, leading to an exponential suppression of disease. These findings offer quantitative guidance for tailoring vaccination campaigns to the topology of real‑world contact networks.