This paper presents a law of large numbers result, as the size of the population tends to infinity, of SIR stochastic epidemic models, for a population distributed over $L$ distinct patches (with migrations between them) and $K$ distinct groups (possibly age groups). The limit is a set of Volterra-type integral equations, and the result shows the effects of both spatial and population heterogeneity. The novelty of the model is that the infectivity of an infected individual is infection age dependent. More precisely, to each infected individual is attached a random infection-age dependent infectivity function, such that the various random functions attached to distinct individuals are i.i.d.
The proof involves a novel construction of a sequence of i.i.d. processes to invoke the law of large numbers for processes in $D$, by using the solution of a MacKean-Vlasov type Poisson-driven stochastic equation (as in the propagation of chaos theory). We also establish an identity using the Feynman-Kac formula for an adjoint backward ODE. The advantage of this approach is that it assumes much weaker conditions on the random infectivity functions than our earlier work for the homogeneous model in [20], where standard tightness criteria for convergence of stochastic processes were employed. To illustrate this new approach, we first explain the new proof under the weak assumptions for the homogeneous model, and then describe the multipatch-multigroup model and prove the law of large numbers for that model.
Stochastic heterogeneous SIR model with infection-age dependent infectivity on large random graphs
Rice University(莱斯大学);Aix–Marseille Université, CNRS, Centrale Marseille, I2M, UMR 7373(艾克斯-马赛大学,法国国家科学研究中心,中央马赛学院,I2M,UMR 7373);Université Paris-Saclay, Université d’Evry Val d’Essonne, CNRS, LaMME, UMR 8071(巴黎萨克雷大学,埃夫里-瓦勒-德-埃松纳大学,法国国家科学研究中心,LaMME,UMR 8071)