Spatio-Temporal Disease Surveillance

There are two principal impediments in statistical process control methods for the detection of bio-terrorism events: firstly, these methods aggregate over space by examining total counts and thus ignore the spatial dimension of the task and secondly they fail to adjust for the usual (seasonal) behaviour of diseases (e.g., Steiner et al., 2011 where the focus is early detection of the start of influenza outbreak). If disease outbreaks were expected to be spread relatively uniformly in space, then the former reason is unimportant. However, since bioterrorism attacks are likely to be introduced to specifically targeted locations, then the resulting disease instances are likely to cluster in space. Consequently, monitoring total counts is likely to reduce the signal-to-noise ratio of this outbreak by aggregating over regions where there is no outbreak. Exploiting the spatial clustering should be able to provide additional power and efficiency in detecting the outbreak.