Outbreak detection through automated surveillance: a review of the determinants of detection

Abstract

Public health agencies and other groups have invested considerable resources in automated surveillance systems over the last decade. These systems generally follow syndromes in pre-diagnostic data drawn from sources such as emergency department visits. A main goal of syndromic surveillance systems is to detect outbreaks rapidly and the number of studies evaluating outbreak detection has increased recently. This paper reviews these studies with the goal of identifying the determinants of outbreak detection in automated syndromic surveillance systems. The review identified 35 studies with 22 studies (63%) relying on naturally occurring outbreaks and 13 studies (37%) relying on simulated outbreaks. In general, the results from these studies suggest that syndromic surveillance systems are capable of detecting some types of disease outbreaks rapidly with high sensitivity. The determinants of detection included characteristics of the system and of the outbreak. Influential system characteristics included representativeness, the outbreak detection algorithm, and the specificity of the algorithm. Important outbreak characteristics included the magnitude and shape of the signal and the timing of the outbreak. Future evaluations should aim to address inconsistencies in the evidence noted in this review and to identify the potential influence of other factors on outbreak detection.