Searching the Web for Flu Outbreaks

Searching the Web for Flu Outbreaks

New York Times, November 28, 2008
 
Two recent studies have shown the promise of using data from search engines to provide early warning of influenza outbreaks — and the pitfalls and limitations, as well. Privacy considerations aside, it is a technology that will need refinement if it is to be used by public health officials for early warning duties. One study, published by the journal Clinical Infectious Diseases, examined the relationship between searches for influenza-related terms on the Yahoo search engine and the actual occurrence of influenza over a four-year period in the United States. The searches spiked one to three weeks in advance of a sharp rise in laboratory cultures testing positive for influenza and up to five weeks in advance of a rise in mortality due to influenza. The other study, published in the journal Nature, found that Google searches for influenza-related terms could actually predict, with high accuracy, spikes in doctor-diagnosed influenza cases that only registered on traditional surveillance networks a week or two later. This study was conducted by Google in collaboration with the Centers for Disease Control and Prevention, giving it substantial credibility. The underlying assumption is that when people start feeling sick, one of the first things many of them do is check the Internet for medical information, even before they see a doctor. Detecting an upsurge in flu cases a week or two earlier than normal could provide substantial health benefits. Doctors and clinics could stock up and dispense vaccines and antiviral medicines in time to help patients, hospitals could prepare for an influx of sick people, and health officials could alert the public. The approach has limitations. It is based on what search terms people choose to use, which is subject to change and will need to be updated frequently. It can sound a false alarm if lots of people start searching for information even though they are not personally sick. It provides no information on what strains are circulating or how virulent they are. And, of course, it raises privacy issues. Google mitigates the privacy concerns by revealing only anonymous, aggregated data, but there are no clear legal or technological safeguards to prevent disclosure of individual search histories should the courts or the government demand it. Still, the potential benefits seem high and the risks containable — for the flu in this country, and for other infectious diseases as well.