Microbial risk models designed to inform water treatment policy decisions.

Chemical risk assessments often focus on measuring exposure as if individuals were subject only to exogenous environmental sources of risk. They also presume that the infection outcomes are independent. For infectious diseases, exposure might not only depend on exogenous sources of microbes, but also on the infection status of other individuals in the population. For example, waterborne agents such as Cryptosporidium parvum and Escherichia coli : O157:H7 might be transmitted from contaminated water to humans through drinking water, from interpersonal contact, or from infected individuals to the environment and back to other susceptible individuals. These multiple pathways and the dependency of exposure on the prevalence of infection in a population suggest that epidemiological models are required to complement standard risk assessments in order to quantify the risk of infection. This dissertation presents models of infection transmission systems that are being developed for the U.S. Environmental Protection Agency as part of a project to quantify the risk of microbial infection. The first part of the dissertation presents de terministic infection transmission models for both homogeneous and heterogeneous populations. The models include infection transmission routes from direct exposure to contamination and from both secondary transmissions (human-human and human-environment-human loops). They are designed to help inform water treatment system decisions. This work shows that the best policy depends on the values of model parameters, including the secondary transmission rate, probability of infection, rate of contamination by exogenous environmental sources and the HIV prevalence. Here it is shown that assessments of water treatment benefits can be misleading if secondary transmission is not properly included in the risk assessment. In addition, a threshold parameter that says whether or not a waterborne infection can remain endemic in a population is derived. The second part of the dissertation develops analogous stochastic infection transmission models including all transmission routes presented in the deterministic models. Since some key secondary transmission parameters influencing water treatment policy are typically unknown, methods to infer these parameters are proposed. We use Bayesian methods and approximations to stationary distributions of prevalence. Data from both simulations and New York City, indicate that otherwise unobservable parameters can be inferred with these techniques. Most importantly, secondary transmission parameters can be inferred from endemic data, without the outbreak data needed by some other approaches. We also simulate the stochastic infection transmissions in a heterogeneous population and study effects of contact patterns on the endemic prevalence. Mixing patterns are shown to affect endemic levels, and indicate areas for further research.Ph.D.Applied SciencesBiological SciencesEngineering, Sanitary and MunicipalHealth and Environmental SciencesMicrobiologyOperations researchPublic healthUniversity of Michigan, Horace H. Rackham School of Graduate Studieshttp://deepblue.lib.umich.edu/bitstream/2027.42/123286/2/3068970.pd