Spatial and temporal analyses of anthrax: An exploratory retrospective and prospective examination of outbreaks in Kazakhstan

This thesis is an exploratory examination of the retrospective and prospective spatio-temporal distribution of anthrax outbreaks cattle, sheep, and other domestic animal operations in Kazakhstan across multiple spatial scales during the period 1960--2006. Analyzing the spatial and temporal occurrence of outbreaks using multiple clustering algorithms may provide a better assessment of the distribution of the disease. Retrospective and prospective surveillance were employed to uncover both spatial and temporal patterns of the disease. First, local clustering statistics including Local Moran's I, Getis' Gi*, and Amoeba were used retrospectively to detect outbreak hotspots using both rayon and 5 km-gridded aggregate levels. Results from these analyses show agreement between algorithms indicating problematic areas of anthrax outbreaks in South-central and eastern Kazakhstan. Research has also shown that monitoring a disease presence on a continuous basis, prospectively, may aid in the allocation of resources needed during an outbreak. This thesis also utilizes a prospective CUSUM approach to evaluate inter-annual variability in the presence of outbreaks using multiple baseline rates of disease. In this analysis baseline expectations were derived from three types of z-scores; Local Moran's I (a spatially constrained calculation), standard z-scores (based on mean and standard deviations of outbreaks), and a moving window average (a temporally constrained calculation). Results indicate anthrax persistence across the time period as well marked differences in the detection of clusters based on the baseline rate. This thesis is one of only a few studies to evaluate the use of spatial techniques to better understand anthrax epidemiology and suggests that both retrospective and prospective techniques for cluster detection are useful in better understanding the spatial and temporal patterns of this disease. While caution is needed to determine an appropriate baseline for prospective CUSUM analyses, these results suggest that the technique may prove useful in understanding outbreak periodicity in Kazakhstan and may also be employed in other countries or regions with repeat anthrax outbreaks. Likewise, retrospective analyses were useful for identifying regions of concern for outbreaks, which may help refine surveillance priorities for local public health authorities.