Predicting the risk of avian influenza A H7N9 infection in live-poultry markets across Asia.

Marius Gilbert,Zhi Huang,Peter W Gething,Catherine Linard,Nick Golding,Simon I Hay,Hang Zhou,Thomas P Van Boeckel,Jane P Messina,Vincent Martin,Hongjie Yu,Samir Bhatt,Timothy P Robinson,G R William Wint,Danhuai Guo,Sheng Zhou,Xiangming Xiao,Andrew J Tatem,Shengjie Lai,Huihui Jiang ,Jeremy Farrar

doi:10.1038/ncomms5116

Abstract

Two epidemic waves of an avian influenza A (H7N9) virus have so far affected China. Most human cases have been attributable to poultry exposure at live-poultry markets, where most positive isolates were sampled. The potential geographic extent of potential re-emerging epidemics is unknown, as are the factors associated with it. Using newly assembled data sets of the locations of 8,943 live-poultry markets in China and maps of environmental correlates, we develop a statistical model that accurately predicts the risk of H7N9 market infection across Asia. Local density of live-poultry markets is the most important predictor of H7N9 infection risk in markets, underscoring their key role in the spatial epidemiology of H7N9, alongside other poultry, land cover and anthropogenic predictor variables. Identification of areas in Asia with high suitability for H7N9 infection enhances our capacity to target biosurveillance and control, helping to restrict the spread of this important disease.

Highlights

Two epidemic waves of an avian influenza A (H7N9) virus have so far affected China
In addition to outbreak investigations and virus surveillance in humans and animals, predictive models linking the locations of the reported cases in humans and live-poultry markets to environmental risk factors can improve risk-based surveillance and control. This is demonstrated by the precedent of highly pathogenic avian influenza H5N1 virus, in which several epidemic waves in Asia were found to be strongly correlated with the spatial distributions of domestic ducks, human populations and wetlands[10]; these associations were used to map the distribution of this disease[11,12,13]
Evaluation of the environmental space occupied by markets (Fig. 1), as determined by the values of key predictor variables for avian influenza, showed that infected markets were present in a limited area of geographic space, they covered a large portion of the available environmental space in China (Fig. 1)

Summary

Introduction

Two epidemic waves of an avian influenza A (H7N9) virus have so far affected China. Most human cases have been attributable to poultry exposure at live-poultry markets, where most positive isolates were sampled. In addition to outbreak investigations and virus surveillance in humans and animals, predictive models linking the locations of the reported cases in humans and live-poultry markets to environmental risk factors can improve risk-based surveillance and control This is demonstrated by the precedent of highly pathogenic avian influenza H5N1 virus, in which several epidemic waves in Asia were found to be strongly correlated with the spatial distributions of domestic ducks, human populations and wetlands[10]; these associations were used to map the distribution of this disease[11,12,13]. The areas predicted to be most suitable for new H7N9 market infection include specific urban areas of China where the disease has not yet occurred, an extensive area in Bengal, the river deltas of Vietnam, and parts of Indonesia and Philippines

Methods

Results

Conclusion