The most fundamental question about surveillance data is whether they contain information that can facilitate detection or characterization of some specific outbreak (or disease).
To facilitate detection and characterization, surveillance data must be both sufficiently timely and indicate either the presence of disease in a population (or individual) or contribute to the analytic power (as in the case of weather or water supply data, which on its own cannot help detect outbreaks, but in combination with clinical data may enable earlier detection). Note that evaluators always ask these questions with respect to a specific disease or type of outbreak (or a group of similar diseases). Evaluators consider data that are both sufficiently early and indicate disease as having informational value.
There are a variety of methods that evaluators may employ to measure informational value. These methods vary in the effort required to conduct the study and the validity of the results.
Handbook of Biosurveillance ISBN 0-12-369378-0
They are often used in combination. Many of the methods that we discuss have been used in published studies, which we will cite as examples. The methods range from surveys of sick individuals to value-of-information calculations, which attempt to measure the information value of surveillance data in the same units (dollars) as the costs of a system to enable direct comparisons.
The types of measurements that are most feasible at present include the following:
• Measures of the improvement in sensitivity, specificity, and timeliness of case detection attributable to the surveillance data
• Signal-to-noise ratios (for surveillance data intended for use in outbreak detection)
• Correlation analysis (for surveillance data intended for use in outbreak detection)
• Measures of the improvement in sensitivity, specificity, and timeliness of outbreak detection attributable to the surveillance data
• Linked record analysis
Was this article helpful?