| Many public health practitioners now use administrative data for surveillance to detect disease outbreaks, despite the limited evidence that this approach to surveillance is effective. The aim of this dissertation is to develop and apply a simulation-based method for evaluating outbreak detection through surveillance of administrative data.; The model is mechanistically driven, reflecting what is known about the underlying processes that follow exposure and lead to an outbreak becoming evident in administrative data. The model comprises four related mechanisms and the interaction between these mechanisms: dispersion, infection, disease, and health-care utilization. In a study that uses the simulation model, I evaluate the performance of syndromic surveillance for detection of an outbreak resulting from a release of anthrax. I also compare the performance of outbreak detection through syndromic surveillance to detection through clinical case-finding using blood-culture testing.; The results of the study demonstrate that the sensitivity and timeliness of outbreak detection through syndromic surveillance are influenced strongly by specificity. At a specificity above 0.9, clinical case-finding detected the majority of outbreaks before syndromic surveillance did. Syndromic surveillance did detect some outbreaks before clinical case-finding, however, even when operating at high specificity. I conclude that syndromic surveillance offers some benefit over clinical case-finding for detection of anthrax outbreaks. Further research is required to determine if this benefit is meaningful in practice. |
