An epidemic in a dynamic population with importation of infectives (Q525295)

The authors study a Markovian SIR epidemic model with demography and importation of infectives. Here, SIR is short for (susceptible \(\to\) infective \(\to\) recovered). The population is modeled by an immigration-death process \(N^{(n)}(t)\) with constant immigration rate \(\mu n\) and death rate \(\mu N^{(n)}(t)\). Here, \(n\) is a parameter, the target population size, i.e., \(N^{(n)}(t)\) fluctuates around \(n\) as \(t\) varies. Each individual in the population is either susceptible, infected or recovered; the corresponding decomposition of the process \(N^{(n)}(t)\) is \(N^{(n)}(t) = S^{(n)}(t)+I^{(n)}(t)+R^{(n)}(t)\). Each immigrant is either susceptible or infected with the fraction of infected immigrants being \(\kappa_n\). While infectious, any given individual infects any given susceptible individual independent of everything else at rate \(n^{-1} \lambda_n\), where \(\lambda_n\) is another model parameter. Each infectious individual recovers at rate \(\gamma_n\). The process \(((S^{(n)}(t),I^{(n)}(t),R^{(n)}(t)))_{t \geq 0}\) thus becomes a continuous-time Markov chain on \(\mathbb{Z}_+^3\). The paper deals with the case where the average population size \(n\) tends to infinity such that the total importation rate \(\mu n \kappa_n\) of infectives tends to a strictly positive constant \(\mu \kappa\) and the infection and recovery rates satisfy \(\lambda_n/\gamma_n \to R_0>1\) and \(\lambda_n/\log n \to \infty\) as \(n \to \infty\). The aim of the authors is to approximate for large \(n\) the model or, more precisely, the process \(\bar S^{(n)}(t) = n^{-1} S^{(n)}(t)\) (the process describing approximately the fraction of susceptible individuals at time \(t\)), by a limiting stochastic process \(S = (S(t))_{t \geq 0}\). The limiting process \(S\) is a Markovian regenerative process which grows deterministically apart from one downward jump per regenerative cycle. However, the authors argue that \(\bar S^{(n)}\) cannot converge to \(S\) in law in the Skorokhod space of right-continuous functions with left limits as the limiting process \(S\) has discontinuous paths almost surely, whereas the process \(\bar S^{(n)}\) has jumps of size \(n^{-1}\), which becomes small as \(n \to \infty\). To overcome this technical difficulty, the authors sandwich \(\bar S^{(n)}\) between two processes \(\bar S_-^{(n)}(t)\) and \(\bar S_+^{(n)}(t)\) which coincide with \(\bar S^{(n)}(t)\) most of the time but sandwich \(\bar S^{(n)}\) during major outbreaks (when \(\bar S^{(n)}(t)\) drops significantly within short periods of time). The main result of the paper is a functional limit theorem saying that \(\bar S_-^{(n)}(t)\) and \(\bar S_+^{(n)}(t)\) converge in law (in the Skorokhod space) to the process \(S\).