Conclusion
Modelling host-pathogen systems can be considered a “wicked problem” :
its success is dependent on multidisciplinary thinking between
statisticians, epidemiologists, and ecologists; and there is a balance
between accepting over-simplified solutions and being overwhelmed by
overly complex ones . Moreover, any solution involves balancing
conflicting and fluid temporal and spatial ecological scales . In
addition to space and time, the environmental processes that describe
host movement—such as climate or seasons—are an often a disregarded,
yet essential, third dimension required to model disease systems .
A deeper forensic approach is required to better understand and
parameterise complex host-pathogen systems, and the Bayesian toolkit
provides a good starting place for this. Overall, future studies of
host-pathogen systems require a better representation of ecological
hierarchy and scale. Scale needs to be examined in terms of applying
suitable Bayesian methods (statistical hierarchy) and by paying
attention to the complexity of the system that is being analysed
(ecological hierarchy). The join-up between these scales has rarely been
studied within ecological systems, and has never been completed for a
single system, yet is essential to providing a whole-system model.
Our survey of the wildlife disease literature shows that the current
application of Bayesian networks to disease problems is limited: in
particular, there is a paucity of hierarchical analyses that infer truly
latent parameters or individual heterogeneities across ecological
scales. Bayesian methods are now being used in several wildlife disease
systems but usually only to tackle standard hypotheses at a single level
of the ecological hierarchy or, at most, spanning two levels of the
ecological or statistical hierarchies.
By developing Bayesian hierarchical modelling methods and integrating
them with real-world empirical data that is not exclusively serological,
the potential exists for ecologists to create whole-system models that
can provide unique insights into the epidemiology of wildlife disease
networks. And the first step towards the whole-system model is to
develop a Bayesian hierarchical model that spans the state-space nature
of each level of the host-pathogen ecological hierarchy.