Google.org announced yesterday that they have expanded Google Flu Trends to 16 additional countries. This system, which is available online, estimates current influenza-like illness (ILI) activity in the U.S. — and now in countries like Bulgaria and Russia — using massive amounts of data generated by Google web search queries for flu-related terms. Details about the development of the predictive model are described in this article that appeared in Nature.
According yesterday’s release, Google Flu Trends had a 0.92 correlation with the Center for Disease Control’s official U.S. flu activity data for the previous season. Since Google Flu Trends data are available almost immediately, unlike official reports that can lag behind by a few weeks, this high correlation suggests that the system can be a useful tool for public health officials.
Here’s what Google Flu Trends is showing for the U.S. right now:
Note that ILI looks to be headed for a pre-December peak, even though we are only 5 days into the official 2009-10 flu season. This is a marked deviation from recent years where the peak happened in January or February. This CDC graph (which could use a little Googlizing from a design perspective) shows peak flu activity for 26 of the last 27 flu seasons:
Any chance that Google is showing an early peak because of worried parents searching for swine flu?
An important aspect of Google Flu Trends is that we filter out terms that may be popular because people hear about them in the news. What we do not use in the models is a term like [swine flu] since people are more likely to type that into Google because they want to know more information about it, given the news headlines, and not because they actually have H1N1 or swine flu.
However, while frantic swine flu and H1N1 web searches are not artificially inflating the current ILI estimate, CDC data do suggest that H1N1 is the driver of the current trends. According to the CDC’s surveillance report for Week 38 ending on September 26, 2009, less than 1% of respiratory specimens that tested positive for influenza were identified as influenza A (H1), A (H3), or influenza B — 53 percent of the positive influenza A specimens were of the A (2009 H1N1) subtype. Since subtyping was either not performed or could not be performed on the remaining samples, the CDC reported that “over 99% of all subtyped influenza A viruses reported to CDC this week were 2009 influenza A (H1N1) viruses.”
When can we expect a peak in flu activity this season? The CDC says it’s unclear:
In past years, seasonal flu activity typically did not reach its peak in the U.S. until January or February, but flu activity has occurred as late as May. However, the 2009 H1N1 virus caused illness, hospitalizations, and deaths in the U.S. during the summer months when influenza is very uncommon. So it is not known when flu activity will increase, when it will be most intense (peak), what viruses will circulate, or how long the season might last.
Check back with Google Flu Trends to be the first to know when relief is on its way. Impress your friends with predictions that stay ahead of the nightly news!
