A spatiotemporal dengue fever early warning model accounting for nonlinear associations with hydrological factors: a Bayesian maximum entropy approach

Abstract

Dengue fever is identified as one of the most widespread vector-borne diseases in the world. In the last decade, high incidence of dengue fever has been observed in southern Taiwan on an annual basis. There is an urgent need to develop a dengue fever early warning model for the area. Previous studies showed that dengue fever transmission has complex space–time dynamics highly depending upon the activities of human and vectors and their interactions, as well as hydrological processes. This study developed a 1-week-ahead dengue fever warning system in southern Taiwan considering weekly-based nonlinear temporal lagged associations between dengue cases and hydrological factors across space and time by investigating the disease database during 1998–2011. The proposed model is based on an integration of distributed lag nonlinear model and spatiotemporal dependence structure under an epistemic framework of Bayesian maximum entropy method. This study identified that minimum temperature and maximum 24-h rainfall are the most significant to dengue fever incidences. Their associations to dengue fever risks are presented with respect to both hydrological measurement levels and temporal lags. A nested spatiotemporal covariance of transmissions was used to account for the disease diffusion across space and time. Results show that the proposed approach can provide early warnings of the dengue fever occurrences in both initial and peak epidemic stages in 2012, and obtain the spatiotemporal distribution of dengue fever epidemics. These features can be useful for local governmental agencies to seek more effective strategies for dengue fever prevention and control.