Generation of an early warning
For a defined case definition, the optimal \(m\) and \(c\) for a desired accuracy target are selected through an iterative process, every time a new time point \(t\) is observed. Briefly:
1. Observed cases up to \(t\) are analyzed for all possible window sizes \(\left(m\right)\) and thresholds \(\left(c\right)\).
2. For each of the \(m\) and \(c\) combinations, the \(Se_{t_{m,c}}\) and \(Sp_{t_{m,c}}\) are estimated for the defined case definition.
3. The \(m'\) and \(c'\) that give the best \(Se_{t_{m',c'}}\) and \(Sp_{t_{m',c'}}\) combination are selected.
4. Based on \(m'\) and \(c'\), EVI is calculated at the most recent time point \(t\) and a decision is made on whether a warning signal is issued or not.
The graphical representation of the EVI calculation is given in Figure 2, while the extensive technical details are described in the Appendix.