The analysis of infectious disease has seen much development of time-based modeling and prediction. However the development of a spatial toolkit for analysis and prediction has seen only limited advance. Spatial prediction of disease spread is a fundamental public health necessity. The spatial questions: Where will an outbreak start? Where will the outbreak go next? When will the outbreak stop in a particular area? These are basic and very relevant questions in the real time surveillance context.
Bayesian spatial models can play an important role in this development. Not only do they provide recursive learning/updating potential, but they also provide flexible formulation of correlated model components and the ability to incorporate neighborhood effects within lagged terms. Different model formulations will be discussed and their relevance to surveillance considered.
Added to a model-based approach, there is a need to consider metrics that can add useful information about the progression of outbreaks. Different metrics based on predictive measures can be formulated and they can include neighborhood signaling as well as single region flagging. I will review currently available metrics and their applications and also suggest new metrics that address spatial behavior. A variety of examples will be presented.