Hawthorne L. Beyer
Division of Ecology and Evolutionary Biology, University of Glasgow, Glasgow G12 8QQ, UNITED KINGDOM.
Epidemiological models are frequently used to estimate basic parameters, evaluate alternative control strategies, and set levels for control measures such as vaccination, culling, or quarantine. However, inferences drawn from these models are sensitive to the assumptions upon which they are based. While many simple models provide qualitative insights into disease dynamics and control, they may not fully capture the mechanisms driving transmission dynamics and, therefore, may not be reasonable approximations of reality. This thesis examines how the predictions made by simple models are influenced by assumptions regarding the dispersion of the transition periods, alternative infection states, and transmission heterogeneity resulting from population structuring. More realistic models of rabies transmission dynamics among domestic dogs in Serengeti District (Tanzania) are developed and applied to the problem of assessing vaccination efficacy, and designing pulsed vaccination campaigns.
Several themes emerge from the discussion of the models. First, the characteristics of outbreaks can be strongly influenced by the dispersion of the incubation and infectious period distributions, which has important implications for parameter estimation, such as the estimation of the basic reproductive number, R0. Similarly, alternative infection states, such as long incubation times, can substantially alter outbreak characteristics.
Second, we find that simple SEIR models fail to accurately capture important aspects of rabies disease outbreaks among domestic dog populations in northern Tanzania, and therefore may be a poor basis for assigning control targets in this system. More complex models that included the role of human intervention in limiting outbreak severity, or that included population structure, were able to reproduce the observed outbreak size distribution. We argue that there is greater support for the structured population model, and discuss the implications of the three models on the evaluation of vaccination efficacy.
Third, at a more regional scale, we build metapopulation models of rabies transmission among domestic dog sub-populations. We use a Bayesian framework to evaluate competing hypotheses about mechanisms driving transmission, and sources of reinfection external to the dog population. The distance between sub-populations, and the size of the sub-populations receiving and transmitting infection are identified as important components of transmission dynamics. We also find evidence for a relatively high rate of re-infection of these populations from neighbouring inhabited districts, or from other species distributed throughout the study area, rather than from adjacent wildlife protected areas. We use the highest ranked models to quantify the efficacy of vaccination campaigns that took place between 2002-2007. This work demonstrates how a coarse, proximate sentinel of rabies infection is useful for making inferences about spatial disease dynamics and the efficacy of control measures.
Finally, we use these metapopulation models to evaluate alternative strategies of pulse vaccination in order to maximize the reduction in the occurrence of rabies. The strategies vary in both the way in which vaccine doses are allocated to subpopulations, and in the trade-off between the frequency and intensity of vaccination pulses. The most effective allocation strategy was based on a measure of the importance of sub-populations to disease dynamics, and it had 30-50% higher efficacy than the other strategies investigated. This work demonstrates the strong potential for the role of metapopulation models in optimizing disease control strategies.