Measuring Detectability as a Function of Outbreak Size

One question of particular interest to evaluators is how small of an outbreak is the algorithm capable of detecting. We refer to analyses that pursue this question as detecta'ility analyses. A detectability analysis is only possible if the evaluator can quantify the size of the outbreaks under study in some manner (e.g., as the fraction of people in a region who were sickened during an outbreak). She can study the effect of outbreak size

Impact Cidt Outbreak Detection

figure 20.4 Sensitivity versus timeliness for two detection algorithms, computed from a hypothetical experiment.

False Alarms, per rear figure 20.5 An AMOC curve from a hypothetical experiment. The curve shows the tradeoff involved in manipulating the detection threshold to improve either timeliness or false alarm rate.

figure 20.4 Sensitivity versus timeliness for two detection algorithms, computed from a hypothetical experiment.

False Alarms, per rear figure 20.5 An AMOC curve from a hypothetical experiment. The curve shows the tradeoff involved in manipulating the detection threshold to improve either timeliness or false alarm rate.

Percent oI Population Affected figure 20.6 Timeliness versus outbreak size for two algorithms. The curves shows that large outbreaks are detected earlier and it provides insight about the smallest outbreak that is detectable by this algorithm (using a particular type of surveillance data).

Percent oI Population Affected figure 20.6 Timeliness versus outbreak size for two algorithms. The curves shows that large outbreaks are detected earlier and it provides insight about the smallest outbreak that is detectable by this algorithm (using a particular type of surveillance data).

on its detectability graphically by plotting timeliness against outbreak size, while holding the false alarm rate at a constant level. Figure 20.6 is a timeliness versus outbreak size plot that compares two algorithms.

0 0

Post a comment