Our objective when evaluating a biosurveillance detection algorithm is to measure its accuracy (sensitivity and false alarm rate) and time to detection (e.g., from the time of infection in an individual case or from the start date of the outbreak). If the algorithm is capable of inferring other outbreak characteristics, such as size, route of transmission, or infectivity, we are also interested in its accuracy for these parameters and the time during the course of the outbreak when the algorithm is capable of estimating these parameters.
Researchers typically measure these algorithm characteristics in the laboratory using surveillance data collected during outbreaks or synthetic data that resembles real outbreak data. However, it is also important to know how well an algorithm or analytic method works when incorporated into a fielded biosurveillance system. Therefore, evaluators also conduct field testing to determine whether the performance in the field is similar to that found in the laboratory. Field testing also assesses other important factors, such as what amount of training is required for professionals working in the field to use the algorithm effectively.
In this chapter, we discuss laboratory evaluation of algorithms. We discuss methods for field testing of algorithms in Chapter 37.
Was this article helpful?