Google Flu Trends Data

I’ve been quiet for a week since the Great Techniques of Integration Debate mostly because I’ve been buried under a mountain of grading and suffering from the usual fatigue that sets in for me at this point in an academic term.

Here is a quick post to make you aware of some freely available (and current) data from Google (http://www.google.org/flutrends/) which you could use, if appropriate, in your classes.  In an article published in Nature in 2009, several employee’s at Google together with an epidemiologist from the CDC devised a method of tracking flu activity around the world by following trends in the search queries about flu-like symptoms.  Their methods and results are here.

The website organizes the information nicely, so that you can sort the data by regions and compare sets over various years.  For example, here is a screenshot of the results from my examination of the current flu trends in California.  The dark blue line is the current trend.  For comparison, the light blue line is the record for 2009-2010, where the flu was apparently incredibly active at this time of year.

At the time of their writing, approximately 90 million people searched for health care related information online per year.  Such a figure has probably only increased since.    They collected those queries which were related to flu-like symptoms and compared it to the usual data tracking of influenza cases in the US.  The fit was quite good. Using this they built a mathematical model which uses this information to estimate, each week, the number of influenza like illness (ILI) physician visits per 100 thousand people that week.  That spike in 2009 should probably then be attributed to the outbreak of the H1N1 virus.  The media hysteria over that outbreak likely fueled many more physician visits regarding flu-like symptoms and many more search queries regarding flu-like symptoms.

You can have students download the data (which is kept current), import it into a spreadsheet software tool like MS Excel, and the create their own charts/visuals from that information, analyze average rates of change, etc.  Their paper itself is not long.  I think it would make a great case study for a mathematical modeling course discussing relatively simple applications of statistics and probability.