Evaluators measure the false-alarm rate by running an outbreak detection algorithm on surveillance data that do not contain outbreaks. The false-alarm rate is the proportion of non-outbreak days (or weeks, depending on the organization of the time series) on which the algorithm signals an alarm. In Figure 20.3, the false-alarm rate is 1/20, or 0.05 false alarms per day of monitoring, or one false alarm every 20 days. (The expression one false alarm every 20 days is called the recurrence interval. It is an easy-to-understand way to present the false alarm rate that you may encounter in the literature.)
Was this article helpful?