Social Graph of 8GCHP

The 8th Global Conference on Health Promotion took place in Helsinki on 10-14th June 2013. It was the first of these conferences to feature a Twitter Wall, a projected feed of tweets that included the conference hashtag:  #healthinall. Given the novelty of the initiative to many colleagues at the conference and in public health, I thought I would illustrate the patterns of tweeting and re-tweeting with a network diagram.


 Screen Shot 2013-06-30 at 7.36.15 AM

Click on the image to download a PDF version. 


  • The graph is derived from all tweets (n = 3660) using the  #healthinall hashtag emitted between the 4pm Helsinki time on 9th June and 4pm on 16th June. The dataset is downloaded using the Twitter API and can be made available to anyone interested (given that Twitter only stores tweets for a week or so). I am happy to also share all relevant code and data in a later tutorial post if there are requests in the comments.
  • Each node in the graph represents a user account on Twitter who was active in the network during that week. Each tie between nodes indicates that one user re-tweeted a message from the other; the graph is undirected and for simplicity I do not show here who was the original and who the follower.
  • This visualization includes the largest connected group of nodes in the dataset. A smattering of other accounts was active in the periphery of the group but they never mentioned anyone in the core network, and they are therefore excluded from this.
  • Nodes and labels are sized by eigenvector centrality and coloured by closeness of connections within sub-groups.
  • The data was downloaded using a script written in Python. The graph was constructed using the Python package: NetworkX. The graph was exported from NetworkX and visualized using Gephi


  • The novelty of the Twitter wall engaged a significant minority of the conference participants. It provided an interactive space where main speakers, participants, young professionals, civil society, and, to a lesser extent, outsiders, were able to contribute to the discussions. The network of tweets extended far beyond the confines of Finlandia Hall. Twitter increased engagement in the conference and exposure outside the meeting, but a few of the participants complained that it distracted them from attending to the main presentations.
  • Conference organisers in public health would do well to consider adding Twitter to their arsenal. Examining the role of the different nodes in the graph, it is clear that particularly well-connected Twitter users (notably @WHO itself) were crucial in generating a wider set of re-tweets. Anyone considering using this tool should get key users on board before the conference and ask them to tweet actively on the conference hashtag — this will help massively increase exposure.
  • Social network analysis of public health data is growing more frequent with work ranging from the study of sexual networks to the risk of obesity. The tools to conduct this sort of analysis are becoming rapidly more accessible and I hope that more public health professionals will engage and add these statistical methods to their research routines.