Dengue transmission dynamics prediction by combining metapopulation networks and Kalman filter algorithm

Predicting the specific magnitude and the temporal peak of the epidemic of individual local outbreaks is critical for infectious disease control. Previous studies have indicated that significant differences in spatial transmission and epidemic magnitude of dengue were influenced by multiple factors, such as mosquito population density, climatic conditions, and population movement patterns. However, there is a lack of studies that combine the above factors to explain their complex nonlinear relationships in dengue transmission and generate accurate predictions. Therefore, to study the complex spatial diffusion of dengue, this research combined the above factors and developed a network model for spatiotemporal transmission prediction of dengue fever using metapopulation networks based on human mobility. For improving the prediction accuracy of the epidemic model, the ensemble adjusted Kalman filter (EAKF), a data assimilation algorithm, was used to iteratively assimilate the observed case data and adjust the model and parameters. Our study demonstrated that the metapopulation network-EAKF system provided accurate predictions for city-level dengue transmission trajectories in retrospective forecasts of 12 cities in Guangdong province, China. Specifically, the system accurately predicts local dengue outbreak magnitude and the temporal peak of the epidemic up to 10 wk in advance. In addition, the system predicted the peak time, peak intensity, and total number of dengue cases more accurately than isolated city-specific forecasts. The general metapopulation assimilation framework presented in our study provides a methodological foundation for establishing an accurate system with finer temporal and spatial resolution for retrospectively forecasting the magnitude and temporal peak of dengue fever outbreaks. These forecasts based on the proposed method can be interoperated to better support intervention decisions and inform the public of potential risks of disease transmission.Author summary: Dengue fever is a vector-borne disease transmitted by the Aedes aegypti mosquito and has become a global threat. The increased possibility of individual exposure to disease vectors due to human travel allowed for the rapid spread of dengue virus to new susceptible populations, ultimately resulting in severe febrile illness and substantial disease burden. Therefore, the integration of information on human movement into infectious disease prediction model will contribute to the prevention and control of infectious diseases. This study developed and retrospectively validated a forecasting system for predicting the spatial spread of dengue fever in Guangdong province, China, using population mobility and dengue fever incidence data. Compared with the local forecasting model designed for each individual city, the proposed system generated improved prediction performance with respect to the targets including peak time, peak intensity, and total incidence for retrospective forecasts of dengue fever epidemics in 12 cities of Guangdong province. The proposed method can be potentially adapted to predict other mosquito-borne infectious diseases.