Surveillance in public health: the methodological problems in identifyng risk factors

Identification of clusters of health events as well as more general problems in public-health surveillance are being increasingly debated. The authors discuss the differences between the systematic (epidemiologic surveillance) and non-systematic (episodic identification of clusters and/or sentinel events) observations. In this regard the role played by statistical methods appears to be specially worthy of consideration, so as to define the capabilities of a given surveillance system in identifying increasing risks. To take into account the relationship between size of the risk to be evidenced and probability (= power) that the implemented test may be significant, specific tables have been developed displaying the relationship between 1st and 2nd type error, sample size and minimum detectable risk, under different exposure levels.