Spatial analysis and mathematical modeling of human infection with avian influenza A (H7N9) virus in Zhejiang, 2013−2016

Objective To investigate the spatial distribution of the cases of human infection with influenza A (H7N9) virus in Zhejiang province during 2013−2016, establish mathematic models to identify the environmental and social factors associated with disease distribution and provide methodological support for the prevention and control of the disease.

Methods Based on the incidence data of human infection with H7N9 virus, virus surveillance data of environmental samples as well as population and socioeconomic data in Zhejiang during this period, a spatial database was established. Global Moran’s I and Getis-Ord Gi* were applied to explore the spatial distribution of H7N9 virus infection epidemic. Geographically Weighted Regression (GWR) model was constructed to analyze the spatial correlation between the epidemic and population density, poultry density and virus distribution in environment. The results were compared with those of Ordinary Least Squares (OLS) model.

Results From 2013 to 2016, a total of 252 human cases of H7N9 virus infection were reported and the viruses were detected in 846 surveillance sites distributed in 77 counties and districts of 11 prefecture-level cities, . At provincial level, the epidemic spread from central-northern part of Zhejiang to surrounding areas; at city level, the epidemic spread from downtown area with high population density to surrounding counties or districts. Spatial regression analysis suggested that the number of human infection cases was correlated with population density and number of H7N9 virus positive surveillance sites (t＝4.127 and 2.697 respectively, P＝0.000 and 0.009 respectively). GWR model (R²＝0.504) showed better fitness compared with OLS model (R²＝0.257).

Conclusion The H7N9 virus infection epidemic in Zhejiang originated in Huzhou and gradually spread to the surrounding areas of Hangjiahu plain. The virus is distributed wildly in the province now. Live poultry marketing might be the main route for the spread of the infection. The spatiotemporal clustering of human infection with H7N9 virus was very obvious, the epidemic was more likely to occur in area with high population and poultry densities and more virus positive surveillance sites. GWR model is suitable for practical application due to its good predicting effect for the distribution of human infections with H7N9 virus.