04 Dic Google Uveils Fast, Accurate Flu Tracking System
Synapse, CA, By Alison Silvis, December 4, 2008
Google’s latest innovation reaches far beyond the realm of the Internet, computer science and extravagant employee benefits: Google Flu Trends, an influenza tracking system with a lag time of only a single day, up to two weeks faster than the CDC’s current flu monitoring system. In the event of a pandemic, additional warning time would enable early implementation of preparedness measures, and would almost certainly save lives. Dr. Larry Brilliant and his colleagues at Google and the CDC will publish the results of their model in a manuscript accepted by Nature (due to the public health utility of the information, the journal allowed Google release a draft of manuscript before publication). In principle, the model is simple: it tracks how frequently Google users query terms that relate to the flu – such as “flu symptoms” and “muscle aches” – and projects the incidence of influenza in a given geographical region. Privacy is maintained by aggregating the data by region and discarding all identifiers. The projections and disease burden of any state in the U.S. are now publically accessible at www.google.org/flutrends. Google is not the first to analyze search trends as a proxy for disease burden, but it certainly has access to the most comprehensive query dataset: literally billions of search terms were analyzed to test the model’s accuracy against CDC tracking data from 2003-2008. It performed quite well. For example, on January 28, 2008, Google’s model detected an increase in influenza incidence two weeks before the CDC. As with any data collection system, there are potential problems with Google’s innovative program. Tracking a surrogate measure rather than the clinical outcome or disease can lead to false alarm or complaisance. Not everyone who types in “flu symptoms” actually has the flu, and not everyone who has the flu turns to Google for answers. Also, users who are cognizant of this tracking system could artificially inflate estimates by entering search criteria multiple times, whether or not they are not sick. I imagine it would not be difficult to create an algorithm to do this, although I also hope that Google in its infinite wisdom has anticipated this problem. If it is successful, Flu Trends may be the first of entire toolbox of surveillance and tracking systems released by Google. InSTEDD project (Innovation to Support Emergencies, Diseases, and Disasters) aims to create a reporting system for healthcare workers to access by cell phone anywhere in the world. Dr. Brilliant hopes to combat many global crises with Google.org; some of his future projects include comparing blood from African bush hunters and monkeys to look for emerging infectious diseases, and monitoring soil moisture in the Amazon to predict droughts and famines. In a January 2008 interview with Nature, he said, “We are just at the beginning here.”