Approximation Methods for Analyzing Multiscale Stochastic Vector-borne Epidemic Models

TL;DR

This paper develops and evaluates mathematical approximation methods for complex stochastic vector-borne disease models, enabling better analysis of their dynamics by reducing model complexity through scaling limits and threshold analysis.

Contribution

It introduces novel mathematical techniques for analyzing high-dimensional stochastic VBD models by applying scaling limits and model simplifications based on disease transmission features.

Findings

Model dynamics depend on R_0, initial infectives, and scaling type.

Different scalings produce different model approximations.

Fast vector dynamics can simplify the model by averaging out vector effects.

Abstract

Stochastic epidemic models, generally more realistic than deterministic counterparts, have often been seen too complex for rigorous mathematical analysis because of level of details it requires to comprehensively capture the dynamics of diseases. This problem further becomes intense when complexity of diseasees increases as in the case of vector-borne diseases (VBD). The VBDs are human illnesses caused by pathogens transmitted among humans by intermediate species, which are primarily arthropods. In this study, a stochastic VBD model is developed and novel mathematical methods are described and evaluated to systematically analyze the model and understand its complex dynamics. The VBD model incorporates some relevant features of the VBD transmission process including demographical, ecological and social mechanisms. The analysis is based on dimensional reductions and model simplications via scaling limit theorems. The results suggest that the dynamics of the stochastic VBD depends on a threshold quantity R_0, the initial size of infectives, and the type of scaling in terms of host population size. The quantity R_0 for deterministic counterpart of the model, interpreted as threshold condition for infection persistence as is mentioned in the literature for many infectious disease models, can be computed. Different scalings yield different approximations of the model, and in particular, if vectors have much faster dynamics, the effect of the vector dynamics on the host population averages out, which largely reduces the dimension of the model.