In a TED conference, Nicholas Christakis presented a conference (1) entitled “How social networks predict epidemics” where he followed 130 students for flu symptoms on the campus of Yale during the 2009 winter. We transpose his idea of a real-friends-network to a social media network of friends such as Facebook. Now the influenza virus doesn’t spread through Internet connection (at least not that one) , yet we assume that having a lot of friends on Facebook is an indicator of an extended social life involving more human to human contacts (²) . We were looking for an easy way to make someone auto-evaluate his condition regarding flu. The idea of a funny looking thermometer on Facebook having 4 levels of fever jumped to our mind because it is easy to use and has a nice look. Each participant can invite one or more of his Facebook friends to use the thermometer application FB. Fever is a common symptom for various winter conditions such as respiratory infections. We started our measurements on december 15th meaning temperature variation are more likely to be correlated with an incoming flu than by other winter diseases already present at the start of our experiment. Results of this study could be follow online with a barometer and a map. The comparison of the Facebook flu tracking thermometer FB with the official flu reports generated weekly by the Health authorities would allow us to evaluate how effective a social media could be . We started the study on december 16th 2010 as this would be some weeks before the seasonal Flu arrival in Europe. The 2010-2011 seasonal flu virus is the A/H1N1 although some other influenza virus are currently present in the northern hemisphere. I search recently the blogosphere and. I found a facebook flu initiative from the university of Auckland dating back from April 2010. We closed the readings around february 15th. 
Over that period, we collected N=431 readings from the facebook applications facebook accounts. Peak for flu events (level3 and level4 ) was observed around Jan 4th (week 1). In order to evaluate these results, we compare them with a reference method. In Belgium, The Scientific Institute of Public Health manages the description of the epidemic in time to design the epidemic curve as a part of a broader clinical and epidemiological surveillance program repeated each year. In this reference method, general practitioners, geographically distributed across Belgium, collect weekly the number of flu syndromes and other acute respiratory infections. They also collect naso-pharyngeal swabs in a part of their patients presenting with flu syndrome. Based on this information, the epidemic curve is drawn and the intensity of the circulation of the virus is measured.
As observed in the reference curve for the 2010-2011 seasonal flu epidemic, the maximum of new FLU cases was observed during week 1. Split per age category of the official epidemic results showed a good match both in time and in shape with the results generated by the Facebook thermometer application. The Facebook apps results did not predict/anticipate the epidemic. It is our current assumption that the Facebook Apps design did not reward the FB user in a proper immidate way failing to assure virality. Virality of the Facebook Thermometer App is the key factor to generate predictive values on Facebook. Failing to make it viral, we were still able to record the Flu cases increase of the 15-64 years category through Facebook in a succesfull way.
See also for reference on last year pandemic
http://www.actualitedumaroc.info/2010/12/pandemies-reseaux-sociaux-pour-sonner.html
