Assessment of risk from long term exposure to waterborne pathogens

Disease due to waterborne pathogens, whether in outbreak or endemic form, c ontinues to be a problem in both the developing and the developed world. Co ntrol of waterborne disease requires accurate assessment of the pathogen do se-response relation and of likely patterns of exposure. Heretofore, risk a ssessment of pathogen exposure has been done on the basis of several standa rd biologically plausible dose-response models. In this paper, the problem of estimating the long-term risk from waterborne pathogens is put into a ri gorous mathematical and statistical framework. The implications of the biol ogic assumptions embedded in the dose-response models (e.g., heterogeneity in susceptibility) are fully considered, as are the likely patterns of long -term exposure (e.g., temporal correlations within individuals and heteroge neity of mean exposures). Two types of long-term risk are described, risk p er person-time and risk per individual where the latter is the risk of infe ction at least once. The effects on these risks of heterogeneity in individ uals' susceptibilities and mean exposures and of temporal correlations of e xposures are described, both theoretically and empirically using a sample o f experimental data sets. Because different models with equal plausibility may give very different results in the low-dose range but fit the experimen tal data equally well, we apply the model uncertainty algorithm of Buckland et al. (1997) on example data sets. Finally, the computational aspects of the general problem, which are often challenging, are discussed along with the conditions under which simplifying approximations may be utilized.