Online Searches Warn of Flu Spikes

Current methods of tracking the flu all come with a bit of a time lag—which is unfortunate when trying to monitor for potential pandemics like today’s swine flu crisis. There is a faster way: According to a February 2009 report in Nature, Google researchers can track flu incidence in real time by monitoring online search queries. The Google model catches a flu outbreak one to two weeks earlier than the Center for Disease Control’s current reporting methods.


Google’s model (black) uses Internet search queries about the flu to estimate current flu levels a week or two faster than the CDC (red). Reprinted by permission from MacMillan Publishers, LTD, Jeremy Ginsberg, et al., Detecting influenza epidemics using search engine query data, Nature 457:1012- 1014, copyright 2009.“Having that one to two week advantage of knowing that something may be developing can have a significant impact on the public health outcome,” says Kumanan Wilson, MD, an investigator of public health policy at the University of Toronto.


Public health officials in the United States and Canada now depend on sentinel doctor’s offices to regularly report the number of people who walk through the door with “influenza-like illness” (ILI) symptoms. But this approach is slow, prone to human error, and relatively costly, says Gunther Eysenbach, MD, MPH, a senior scientist at the Centre for Global eHealth Innovation and professor at the University of Toronto in Canada.


In 2006, Eysenbach, who first had the idea of using Internet search queries to track the flu, performed a pilot study showing that he could largely eliminate the reporting lag with an automated system. Eysenbach’s strategy tracked how often folks searched for “flu” or “flu symptoms” online, and then noted how many users subsequently clicked on an informational ad about seasonal influenza. The number of users who clicked on the ad closely traced Canada’s seasonal influenza spike, and offered the data a week or so faster than traditional methods. His study was reported in the American Medical Informatics Association Annual Symposium Proceedings.


Google built on the work of Eysenbach and others. The Google researchers started with 50 million of the most common searches. They compared the weekly frequency of each with the up and down of seasonal flu spikes over five years. Those that correlated best (the top 45) were all flu-related.


With those top 45 search queries, Google created a linear model for tracking the flu in real time. Current data can be found at Google Flu Trends (, which, since April, is also tracking flu trends in Mexico.


Internet queries can pick up a flu spike quickly because they immediately register any increased interest in the flu. That is both a strength and a weakness of what Eysenbach calls “infodemiology.” The downside, he says, is that in a pandemic situation, you may be monitoring more panic than actual flu cases. “Our current swine flu data demonstrate that it can be difficult to separate the signal from the noise,” he says.


Before the search query approach can be adopted as an early warning signal on the national or international level, its effectiveness needs to be better proven, says Wilson. But he likes the idea of a freely available, Internet-based system that would likely encourage more transparent reporting by governments and health officials.


Eysenbach is now investigating many other ways of using the Internet to observe and influence people’s health. He wants to interact with those who search online through questionnaires, and he is seeing what he can gather from microblogs such as Twitter.

All submitted comments are reviewed, so it may be a few days before your comment appears on the site.

Post new comment

The content of this field is kept private and will not be shown publicly.
This question is for testing whether you are a human visitor and to prevent automated spam submissions.
Enter the characters shown in the image.