A MEAN FIELD GAME ANALYSIS OF SIR DYNAMICS WITH VACCINATION

Josu Doncel, Nicolas Gast, Bruno GaujalProbability 2020 · DOI

TL;DR

This paper models the decision of individuals to vaccinate against a disease within an SIR (Susceptible-Infected-Recovered) framework using a mean-field game (MFG) approach. It reveals that individual rational vaccination decisions lead to a suboptimal outcome compared to a centralized, socially optimal strategy, suggesting that subsidies are necessary to encourage optimal vaccination behavior.

What Problem Does This Paper Address?

The paper tackles the challenge of understanding and optimizing vaccination strategies in a population, specifically when individuals make their own decisions about when and how much to vaccinate. While the SIR model is a well-established epidemiological tool, incorporating individual decision-making under uncertainty and strategic interaction requires more sophisticated modeling. The authors frame this as a mean-field game, where each individual's decision is influenced by the aggregate behavior of the population, but their individual actions don't significantly alter the overall population dynamics. This allows them to analyze the "selfish" vaccination behavior of individuals and compare it to a centralized "socially optimal" strategy that minimizes the total cost of the epidemic for the entire population. The core question is: how do individual rational choices in vaccination align with, or diverge from, what's best for the collective good, and what are the implications for public health policy?

Methods

The paper employs a finite state mean-field game (FMFG) framework. This involves:

  • SIR Model Dynamics: A standard SIR model is used to describe the spread of the disease, with the addition of a vaccination control action for each player.
  • Mean-Field Game Formulation: The problem is cast as a continuous-time FMFG with a finite state space (Susceptible, Infected, Recovered/Vaccinated). Each player aims to minimize their individual cost, which includes vaccination costs and infection costs.
  • Bellman Equation: The optimal strategy for an individual player is determined by solving a Bellman optimality equation, which captures the trade-off between immediate costs and future expected costs.
  • Mean Field Equilibrium (MFE): An MFE is defined as a fixed point of the best-response correspondence, where each player's strategy is a best response to the population's average strategy. The paper proves the existence and uniqueness of a pure MFE.
  • Centralized Control Problem: For comparison, the authors also analyze the problem of minimizing the total population cost (social optimum) using optimal control theory, specifically Pontryagin's Maximum Principle.
  • Numerical Simulations: The paper uses numerical experiments to compare the MFE and social optimum strategies, analyze the convergence of finite-player games to the MFE, and study the efficiency of the equilibria.

What's novel here is the application of FMFG theory to this specific epidemiological problem, allowing for a rigorous analysis of individual versus collective optimal vaccination strategies and providing a framework for understanding the economic incentives behind vaccination.

Datasets

Not discussed in the paper. The paper uses simulated data based on parameters from a previous study on the H1N1 epidemic.

Key Insights and Ideas

  • Individual Rationality vs. Collective Good: The most striking insight is that individuals acting rationally to minimize their own costs (selfish vaccination) lead to a suboptimal outcome for the population as a whole. This means that the "jump time" (when individuals stop vaccinating) for the mean-field equilibrium is always earlier than for the socially optimal strategy.
  • The Need for Subsidies: Because individual vaccination decisions are less than socially optimal, the paper strongly suggests that vaccination should be subsidized. This subsidy, effectively reducing the cost of vaccination, can incentivize individuals to vaccinate more, aligning their behavior with the collective good and reducing the overall cost of the epidemic.
  • Bang-Bang Strategies: Both the individual rational strategy (MFE) and the socially optimal strategy are characterized as "bang-bang" strategies. This means individuals (or the central planner) will either vaccinate at the maximum rate or not at all, with a single switch-over point in time.
  • MFG as a Powerful Tool: The paper demonstrates the utility of mean-field game theory in analyzing complex strategic interactions in epidemiological models, offering a more tractable approach than analyzing finite-player games directly.

Key Results

  • Unique Mean Field Equilibrium: The paper proves the existence of a unique, pure, mean-field equilibrium (MFE) for the vaccination game. This equilibrium is a bang-bang strategy.
  • MFE vs. Social Optimum Jump Times: The vaccination period for the MFE is always shorter than that for the social optimum. Specifically, the jump time for the MFE ($T_{eq}$) is always less than or equal to the jump time for the social optimum ($T^*$), with equality only in trivial cases.
  • Convergence of Finite-Player Games: Numerical experiments show that the cost and switching curves of N-player games (both Nash equilibrium and social optimum) converge to their mean-field counterparts as N increases. The convergence rate for the cost is found to be $1/N$.
  • Efficiency: The N-player equilibria are shown to be reasonably efficient, with costs close to the global optimum, especially for larger N.

Limitations and Future Work

  • Model Simplifications: The model assumes a homogeneous population and a simplified SIR structure. Future work could explore heterogeneous populations, more complex disease dynamics, or different vaccination strategies.
  • Parameter Sensitivity: While numerical experiments are presented, a more extensive sensitivity analysis across a wider range of parameters could provide deeper insights.
  • Analytical Tractability of N-Player Games: The paper notes that analyzing the N-player Nash equilibrium directly is computationally challenging. Future work could focus on developing more efficient analytical or numerical methods for these finite-player games.
  • Generalization of MFG: The authors suggest that the framework used is generalizable to other problems beyond vaccination, indicating potential for broader applications of their methodology.

Summary updated

Abstract

We analyze a mean field game model of SIR dynamics (Susceptible, Infected, and Recovered) where players choose when to vaccinate. We show that this game admits a unique mean field equilibrium (MFE) that consists in vaccinating at a maximal rate until a given time and then not vaccinating. The vaccination strategy that minimizes the total cost has the same structure as the MFE. We prove that the vaccination period of the MFE is always smaller than the one minimizing the total cost. This implies that, to encourage optimal vaccination behavior, vaccination should always be subsidized. Finally, we provide numerical experiments to study the convergence of the equilibrium when the system is composed by a finite number of agents ( $N$ ) to the MFE. These experiments show that the convergence rate of the cost is $1/N$ and the convergence of the switching curve is monotone.
BibTeX
@inproceedings{doncel2020mean,
  author = {Josu Doncel and Nicolas Gast and Bruno Gaujal},
  title = {{A MEAN FIELD GAME ANALYSIS OF SIR DYNAMICS WITH VACCINATION}},
  booktitle = {Probability},
  year = {2020},
  doi = {10.1017/s0269964820000522},
}
Figure