Data Analysis and Outbreak Detection

Chen, Hsinchun; Zeng, Daniel; Yan, Ping

The analysis components of a syndromic surveillance system focus on detecting the changes in public health status, which may be indicative of disease outbreaks. At the core of these analysis components is the automated process of detecting aberration or data anomalies in the public health surveillance data, which often have prominent temporal and spatial data elements, by statistical analysis or data mining techniques. These methods are also capable of dealing with various common problems in epidemiological data such as bias, delay, lack of accuracy, and seasonality. These techniques are the focus of this chapter. When processing public health surveillance data streams, it is often necessary to map the collected syndromic data into a small set of syndrome categories to facilitate follow-up analysis and outbreak detection. Section 4.1 discusses related syndrome classification approaches. In Sect. 4.2, we provide a taxonomy of anomaly analysis and outbreak detection methods used for biosurveillance. Sections 4.3–4.6 summarize various specific detection methods spanning from classic statistical methods to data mining approaches, which quantify the possibility of an outbreak conditioned on surveillance data.