Fig 1.
The optimal alert threshold of outbreak detection systems (that maximizes outbreak detection accuracy) under different testing rates and noise structures.
Each imperfect test uses the same value for both its sensitivity and specificity (either 85% or 90%). Circular markers represent tests with 0-day turnaround times, and triangular markers represent tests with delayed results. Λ(4) indicates the mean noise incidence is 4 times higher than the mean measles incidence, for example. S1 Table provides the underlying values in a table format at Λ(7) to help distinguish between lines that overlap.
Fig 2.
The accuracy of outbreak detection systems under different testing rates and noise structures, at their respective optimal alert thresholds.
The shaded bands illustrate the 80% central interval, and the solid/dashed lines represent the mean estimate. Each imperfect test uses the same value for both its sensitivity and specificity (either 85% or 90%). Solid lines represent tests with 0-day turnaround times, and dashed lines represent tests with result delays. Λ(4) indicates the mean noise incidence is 4 times higher than the mean measles incidence, for example.
Fig 3.
The detection delay of outbreak detection systems under different testing rates and noise structures, at their respective optimal alert thresholds.
The shaded bands illustrate the 80% central interval, and the solid/dashed lines represent the mean estimate. Each imperfect test uses the same value for both its sensitivity and specificity (either 85% or 90%). Solid lines represent tests with 0-day turnaround times, and dashed lines represent tests with result delays. Λ(4) indicates the mean noise incidence is 4 times higher than the mean measles incidence.
Fig 4.
The proportion of the time series in alert of outbreak detection systems under different testing rates and noise structures, at their respective optimal alert thresholds.
The shaded bands illustrate the 80% central interval, and the solid/dashed lines represent the mean estimate. Each imperfect test uses the same value for both its sensitivity and specificity (either 85% or 90%). Solid lines represent tests with 0-day turnaround times, and dashed lines represent tests with result delays. Λ(4) indicates the mean noise incidence is 4 times higher than the mean measles incidence.
Table 1.
Compartmental model parameters.
Fig 5.
A schematic of the outbreak definition and alert detection system.
A) Measles incidence time series. B) Noise incidence time series. C) Observed time series of test positive cases according to a given testing scenario. The orange bands present in all 3 panels represent regions of the measles time series that meet the outbreak definition criteria. In panel C, the dark blue bands represent regions of the test positive time series that breach the alert threshold (the horizontal dashed line), and constitute an alert.