Heterogeneous population dynamics and scaling laws near epidemic outbreaks (Q327588): Difference between revisions

The authors use a \(SIS\) (susceptible-infective-susceptible) model with standard incidence as a starting point to discuss the influence of heterogeneity upon the global behavior of deterministic and stochastic disease propagation models and upon the possibility of determining warning signs for impending critical transitions. The paper is essentially self-contained, a primer on the asymptotic behavior of epidemic models in general and of the homogeneous (i.e., without heterogeneity) \(SIS\) model in particular being provided in Sections~1 and 2. The heterogeneous model of concern is introduced in Section~3, in the form \[ \begin{aligned} S^\prime(t, \omega)&=-\beta (\omega)\frac{J(t)}{T(t)+J(t)}S(t,\omega)-\eta (\omega)S(t,\omega)+\gamma (\omega)I(t,\omega),\quad S(0,\omega)=S_0(\omega)\\ I^\prime(t, \omega)&=\beta (\omega)\frac{J(t)}{T(t)+J(t)}S(t,\omega)+\eta (\omega)S(t,\omega)-\gamma (\omega)I(t,\omega),\quad I(0,\omega)=I_0(\omega),\end{aligned} \] where \[ T(t)=\int _\Omega q(\omega)S(t,\omega)d\omega,\quad J(t)=\int _\Omega q(\omega)I(t,\omega)d\omega. \] In the above, \(\omega\) represents the heterogeneity state which is assigned to an individual (a trait which is relevant to the spread of the disease). Due to the fact that the heterogeneous model satisfies a constant population assumption, order reduction can be performed, one being consequently left with a Cauchy problem for a single heterogeneous equation. The basic reproductive number \(R_0\), defined as the the expected number of secondary infections caused by a single infective individual during its entire period of infectiousness when introduced in a totally susceptible population, is shown to retain its threshold stability properties from the homogeneous counterpart of the model of concern. The authors then proceed to extend their model by laying out the motivation behind the introduction of additive stochastic noise (which requires also the use of a cutoff function) and of a slow variable into the system. Having in mind that a transcritical bifurcation occurs when \(R_0\) passes through 1, the authors recall several techniques for obtaining warning signs for homogeneous fast-slow systems, concluding that those analytic findings are not applicable in their heterogeneous situation. They then proceed to discuss this matter further from a numerical viewpoint, considering the variance of \(I(t)=\int _\Omega I(t,\omega)d\omega\) as an early warning sign, the case of discrete and continuous heterogeneity states being investigated. The situation in which the transmission rate \(\beta (t,\omega)\) is nonseparable is investigated into greater details. The authors' conclusion is that more focus on quantitative scaling laws is required for heterogeneous model validation in respect to actual data.