Spatial transmission network construction of influenza-like illness using dynamic Bayesian network and vector-autoregressive moving average model

Abstract Background Although vaccination is one of the main countermeasures against influenza epidemic, it highly essential to make informed prevention decisions guarantee that limited resources are allocated places where they most needed. Hence, fundamental steps for decision making in characterize its spatio-temporal trend, especially on key problem about how transmits among adjacent and much impact place could have neighbors. To solve this while avoiding too additional time-consuming work data collection, study proposed a new concept route as well estimation methods construct transmission network. Methods The influenza-like illness (ILI) Sichuan province 21 cities was collected from 2010 2016. A joint pattern based dynamic Bayesian network (DBN) model vector autoregressive moving average (VARMA) utilized estimate routes, which were applied two stages learning process respectively, namely structure parameter learning. In learning, first-order conditional dependencies approximation algorithm used generate DBN, visualize routes infer impacts others transmission. VARMA adopted strength these impacts. Finally, all estimated put together form final Results results showed period cycle longer Western Chengdu Plain than Northeastern Sichuan, there would be potential bordering provinces or municipalities into province. Furthermore, also pointed out several with relatively high associations, serve clues hot spot areas detection surveillance. Conclusions This framework exploring potentially stable between different measuring specific sizes effects. It help timely reliable prediction trend infectious diseases, further determining possible next epidemic by considering their neighbors’ incidence relationships.