Predicting Epidemics via Social Networks: An Author Spotlight on James H. Fowler and Nicholas A. Christakis

As flu season approaches, we thought it would be interesting to spotlight the authors of a recently published PLoS ONE paper entitled, Social Network Sensors for Early Detection of Contagious Outbreaks. The authors James H. Fowler, a professor in the School of Medicine and Division of Social Sciences at UC San Diego, and Nicholas A. Christakis, a professor in the School of Medicine and in the Faculty of Arts and Sciences at Harvard University, were kind enough to answer a few questions over email about themselves, their research and their experience with PLoS ONE.

First, a bit of detail on your scientific background – when did you become interested in the connection between social networks and health?

We met in 2002, and at the time James was studying political social networks and Nicholas was studying social networks and health.  A mutual friend who heard us ranting realized that we were asking the same kinds of questions.  So he introduced us and we have been working together on these ideas ever since.  It’s a curious thing, that people assemble themselves into elaborate social networks.  Why do we do it?  What does it mean for our lives?  And how might we exploit any insights about social networks to make the world a better place?

You mention in your paper that modeling research for outbreak surveillance has been done before.  Can you explain how your approach differs from previous research and current tracking methods used by the CDC?

Past research has been mainly theoretical, so this is the first effort that we are aware of that shows these ideas actually work in the real world.  Plus, prior methods have generally required that the network be mapped, whereas in this paper, we show that one can exploit the ‘friendship paradox’ to identify sentinels or sensors in a network (the central nodes) without having to map it.  Moreover, the current tracking methods used by the CDC typically lag a few days or weeks — if I want to know how many people have the flu today, I have to wait a few days to find out.  In contrast, our method offers the possibility of early detection.  By watching the sensor group, we can predict what will happen to the whole network two weeks in the future (in the case of flu, at least).

Your study sample consists of college students in a university setting.  How would it scale to a national level?  Could you apply the social graph information in Facebook, for example, to identify individuals who need to be tracked?

The method should, in principle, be scalable.  For example, we could contact 1000 people in New York City, ask them to name their friends, and then follow those friends.  As soon as we notice an uptick in the flu (or anything else that we want to track) in this friend group, it would indicate that an epidemic was on its way for the city.  And one way to use Facebook would be to incorporate passive monitoring of the friend group.  We already know from Google Flu Trends that online searches give information about outbreaks; but by following a sentinel group composed of nominated friends (or of individuals otherwise known to be central in the network), we could get this information even earlier.

Your paper includes a movie illustrating the progression of the flu in a friendship network.  Can you explain it in a bit more detail?

It mainly shows that people at the center of the network tend to get the flu before people on the edges.  Each frame of the movie shows the largest component of the network (714 people) for a specific date, with each line representing a friendship nomination and each node representing a person.  Infected individuals are colored red, friends of infected individuals are colored yellow, and node size is proportional to the number of friends infected.  When you watch the video, you get the sense of flu ‘blooming’ among people nearer the center of the network earlier in the course of the epidemic.

As you both have been long time collaborators, what is your next big research project or where would you like to go from here?

As we discuss in our book Connected, we think social networks are an integral part of human evolution.  We are currently very interested in the question: “How Does the Social Become Biological?”  If, as we argue, our vast social networks bind us together into a human superogranism, then we should be able to identify biological sytems that influence the structure and function of our connections to one another.

This is the second time you’ve published with PLoS ONE, what made you decide to come back?

PLoS has two important advantages: it is fast, and it does a very good job of publicizing new papers because papers are immediately and widely accessible.  When our first paper on the spread of sleep loss and drug use came out earlier this year, we were able to publish our results quickly, and we had many colleagues give us feedback because they had seen the paper online.

Both of Fowler’s and Christakis’ papers Social Network Sensors for Early Detection of Contagious Outbreaks and The Spread of Sleep Loss Influences Drug Use in Adolescent Social Networks are freely available to read, reuse, comment and rate.  You can also see their recent TED talk on the topic by clicking here. And finally, if you’re interested in reading more research on social networks please visit PLoS ONE and search using keywords “social network.”

This entry was posted in Interviews, Media and tagged , , , , , , , , . Bookmark the permalink.

One Response to Predicting Epidemics via Social Networks: An Author Spotlight on James H. Fowler and Nicholas A. Christakis

  1. Pingback: Quick Links | A Blog Around The Clock

Add Comment Register



Leave a Reply

Your email address will not be published. Required fields are marked *

You may use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <strike> <strong>