Statistical and mathematical methods to improve models of infectious disease transmission in and between human and animal populations
4 October 2017
Universiteit Hasselt, Campus Diepenbeek, theatre H5 - Agoralaan - 3590 Diepenbeek
Yimer Wasihun Kifle
Prof Niel Hens, Prof Philippe Beutels, Prof Christel Faes
PhD defence Yimer Wasihun Kifle - Faculty of Medicine and Health Sciences and UHasselt
In the one-health manner, public health experts have strong interest in human-animal health research. Several pathogens (e.g., avian influenza and Ebola) circulate in the animal population. Direct or indirect human-animal interactions enable these pathogens to pass from the animal to the human population. However, empirical data on human-animal interactions is rarely used to characterize the transmission risk of zoonotic infections. Even if demographical changes and technological innovations could change social interactions over calendar time, the impact of macroscopic changes in mixing patterns over time has not yet been investigated. Since improving existing statistical and mathematical methods in the ecology of infectious diseases has paramount importance to characterize the risk of disease transmission, this dissertation focuses on three major improvements. (1) It relates empirical evidence on human social contact patterns with human-animal interactions via animal ownership and touching. (2) It investigates the macroscopic changes in mixing patterns over time relevant to the spread of respiratory infections. (3) It develops models for predicting the spatiotemporal distribution of Culicoides midges that can transmit vector-borne diseases.
We find that a larger proportion of participants own/touch pets than poultry and livestock. Larger households are more likely to own an animal and, unsurprisingly, that animal owners are more likely to touch animals. We observe a significant effect of age on animal ownership and touching. Animal owners have more social contacts than non-animal owners during weekend. Assuming that animal ownership and/or touching are at-risk events, we show that human-animal interactions involving young children (0-9 years) and adults (25-54 years) have the highest potential to cause a major zoonotic outbreak, but the overall probability is very low in Flanders.
We find that the number of contacts seems to increases over time for young adults aged 20 to 25 years and adults aged 40 to 45 years while there is no evidence for changes in other age categories. Our main finding here is that the macroscopic changes in contact patterns over time are negligible. Note however that our study is based on two social contact surveys conducted in Flanders, Belgium anno 2006 and anno 2011 (5-years apart) and thus limited to two time points with a small-time interval between them.
We developed univariate and multivariate additive and coupled spatiotemporal models using stochastic partial differential equation (SDPE) approach. Our methods are illustrated with Culicoides data from Belgium. Models accounting for space-time interaction (coupled) outperform models not accounting for (additive) in terms of prediction accuracy measures. The Bayesian spatial predictions show that the highest prevalence of Culicoides species is found in the Northeastern and central parts of Belgium during summer.