Modeling vector borne diseases

Abstract

Models of vector borne diseases are traditionally based on detailed knowledge of the pathogen, host, vector, and transmission process. Widely-used compartmental models, for example, contain several differential equations and parameters for each species of vector and host. Alternative approaches take advantage of big data and machine learning to predict disease incidence, but provide little insight into the causes that need to be addressed to prevent and control epidemics. Where disease drivers have been identified, the literature is focused on correlations of small sets of variables with disease incidence. While correlated data points can help predict disease, they are unreliable in a changing environment, as is produced for example by climate change and technological development. This project aims to examine the causal factors of vector borne diseases in Brazil. We will compile data on known correlates and analyze their causal relationships with disease incidence in municipalities and states, utilizing approaches from complex networks. With these results, we will build a machine learning model for prediction of epidemics. Vector borne diseases such as Yellow Fever, Zika virus, and Dengue pose a public health concern in Brazil. This study aims to increase understanding of disease drivers, which will help better tune existing models, select relevant data sources for more accurate predictions, and inform public health decisions. (AU)