Abstract
Background: Influenza is an acute respiratory infection caused by an influenza virus, and the primary intervention strategy is seasonal vaccine. Due to various influenza strains and their rapid mutation each year, how to recognize the key population and timing of the vaccination becomes essential. Considering the importance of finding possible spreading directions and effects of influenza between cities for department of influenza prevention, the construction of influenza transmission network becomes meaningful.Methods: 21 cities in Sichuan province were divided into different learning communities according to whether they were adjacent to each other or not. In each community, the first-order conditional dependencies approximation algorithm was performed to learn the possible structure of the time-lagged correlations between different time series vectors of the ILI estimated weekly number, and the vector autoregressive moving average models were performed for learning the lag orders and parameters of the time-lagged correlations between different time series vectors in each community.Results: It detected a number of significant time-lagged correlations between cities in Sichuan province using two models, and the lag was from 1 week to 3 weeks. The parameters indicating the suspected propagation relationship were between -0.90 and 0.75, and the proportion of the negative values in parameters increased with time. Furthermore, the spreading routes learning from two models were almost in accordance with the traffic network of Sichuan province.Conclusions: This study proposed an innovative framework for exploring the potentially stable transmission routes between different regions and measuring specific size of the transmission effect. It could be used for the infectious disease key area confirmation by considering their adjacent areas’ incidence and the transmission relationship.
Spatial Transmission Network Construction of Influenza-Like Illness Using Dynamic Bayesian Network and Vector-Autoregressive Moving Average Model
