Silent geographical spread of the H7N9 virus by online knowledge analysis of the live bird trade with a distributed focused crawler

Chen, Chen; Lu, Shan; Du, Pengcheng; Wang, Haiyin; Yu, Weiwen; Song, Huawen; Xu, Jianguo

We designed a tool to determine all possible connections between cities using public open source code from Tim Dwyer and Thomas Jakobsen (http://bl.ocks.org/mbostock/4062045#index.html). The input data comprised a table of connection results with node columns, connected target columns, and their connections. For example, if Shanghai was the node, its connections would be ''Jiangsu'' and the number of links would be ''295'' on IC. All possible connections were constructed, recorded, and entered in a structured query language (SQL) database. The pair of connections A to B and B to A was only considered once in our analysis. A force-directed graph was constructed by the connections between two provinces and cities. We only removed all connections to provinces when we calculated the connection between cities. Thus, information for some cities with connections only to provinces was lost in this graph. When we constructed the graph of the provinces, we only considered four types of connections: (i) node province to target province, (ii) cities in the node province to the target province, (iii) node province to cities in the target province, and (iv) cities in the node province to cities in the target province. These connections were accumulated as the connections between provinces. Because the connection to diseased birds was important in our analysis, we focused exclusively on the provinces and the cities directly connected to first-wave provinces and cities. Thus, at least one component in the pair of connections (node or target) belonged to the first-wave H7N9 outbreak area. Shanghai, Jiangsu, Zhejiang, and Anhui were selected as the first-wave provinces and 11 cities in these provinces were also included. Not all live poultry carried the virus, and too much data from connections would distort our understanding of the transfer of diseased live poultry. The connections from this firstwave were not included and considered in our graph. For example, Guangdong is connected to Zhejiang, and Guangxi is connected to Zhejiang; thus, these three provinces are all included in our graph. Although we had evidence showing that Guangdong had trade information on Guangxi, we did not present the connection between Guangdong and Guangxi in our graph because this connection was not associated with the first-wave trade information. The CC and IC were used together as the connections, and IC was used as the weight for each connected line. All aforementioned filters and selections were carried out by a self-designed Java script and SQL queries before a force-directed graph was drawn. The potential transmission networks of live bird trade in cities with reported H7N9 cases were derived from the connection of cities. Cities with patients but no direct connection to first-wave cities were also connected by seeking the cities that they connected to in the figures. For example, Fuzhou was not directly connected to the first-wave cities but was connected to Nanping and Ningde. We also included the relationships between cities from the database in the figures to show their possible transmission networks.