Ostfeld, R.S., Glass, G.E., and Keesing, F.

Spatial epidemiology: an emerging (or re-emerging) discipline

Reviewed 09/29/11

Spatiotemporal patterns are at the base of many ecological studies. These patterns can be modeled based on incidence, spread, or prediction. They can also be analyzed through a mechanism based framework or one that looks more at large scale biogeographical processes.  The use of spatial modeling in epidemiology has many facets itself. There are many different kinds of diseases and likewise many modes of transmission, which can affect how a pathogen is examined. Diseases that are passed via ingestion of some kind are fundamentally different from those that can jump hosts at small spatial scales.
Models can be either spatially explicit or implicit. Implicit models look at space from relativistic perspective, the where itself does not matter as long as there is a where. Metapopulation analysis is one example as is wave spread of emerging diseases in previously uninfected and non-resistant populations.
Explicit analysis on the other hand tries to combine the spatial and temporal dynamics of the vector (if one exists), the reservoir hosts (if we are only interested in a terminal or primary infected populations such as humans or economically important species), and pathogen incidence itself in focal host or hosts. Forward looking models are often used to map areas and time-scales of high and low ecological risk. When modeling human diseases unfortunately though it does not seem as if taking a purely ecological perspective is the best however. Fundamental to the definition of ecological risk is the concept of exposure in the absence of preventative measures; which many human populations take steps to avoid. This can include such simple measures as bed netting for mosquitoes in areas with traditionally high malaria rates to vaccines for entire populations.
The authors of this paper argue in many subtle and not as subtle ways for the use of mechanisms in linking underlying spatiotemporal distributions with disease incidence. It is not enough to correlate a high percentage of wetlands with mosquito abundance if no information is available on how frequently mosquitoes in the area are infected with the virus in general. This is just one example given in a table displaying similar lines of thinking.
There are of course many difficulties that can arise from this type of modeling perspective. For one it is possible to have multiple vectors or hosts/reservoirs, each with their own unique spatial patterns. One example is given of how the life stages of a vector can drastically affect how tick-borne encephalitis is transmitted, and those specific requirements are difficult to synthesize using spatial models alone. Landscape level conditions can have profound impacts beyond just the inclusion of density or abundance spatial models. Patchy habitats and the interspace between those "ideal" spots can alter how disease incidence cycles and moves at the regional or global perspective. Emerging diseases are also difficult to model as they can be either completely new to a population through introduction or new in terms of their accelerating degree of infection through some sort of ecological release.