Dengue is endemic and with year-round presence in Singapore and it presents a serious
economic and disease burden to the public. Early detection of dengue outbreaks is very important for an
effective disease surveillance system, which targets to detect signals of disease outbreaks earlier before
it bursts into larger scale infections and spread. In the current practice, an outbreak is often defined as
the dengue incidence exceeding a threshold, which is calculated based on the mean and standard
deviations during past few years. However, such a threshold is not robust in determining the outbreaks,
as argued by many authors in the literature. In this study, we propose to determine the threshold by
using the generalized Pareto distribution, which is a useful extreme value model. We show that the
dengue outbreaks by the proposed thresholds are more reasonable than the existing thresholds.
This work is partially supported by the National Research Foundation Singapore under 748 grant No. NRF2017VSG-AT3DCM001-04.