Google Flu Trends, a program used to estimate U.S. influenza-like illnesses (ILI) rates from Internet searches, is not as accurate at estimating rates of laboratory-confirmed influenza as the Centers for Disease Control and Prevention (CDC) national surveillance programs, according to a University of Washington study released today.

The study is a response to a Google study that was published in Nature in 2009, lead researcher Jason Ortiz told DOTmed News.

"Google Flu Trends, clearly I think, made a tremendous contribution to public health. What we did is in no way 'gotcha' or to try to undermine their contribution," Ortiz said. "In general, influenza surveillance is a difficult thing to accomplish because there are so many objectives to survey, and it's hard to meet all of those objectives with one system."

In order to study the accuracy of Google Flu Trends, the researchers studied the incidence of flu outbreaks in the United States between 2003 and 2008. They found that Google Flu Trends strayed the most from CDC surveillance for positive flu tests during the 2003-2004 flu season. That year was particularly notable for early and intense influenza activity, and there was a great deal of media attention to influenza during that year, the report said.

"It's possible that during periods of panic or anxiety about influenza infection, people's Internet searches change," Ortiz said. "If that's the case, Google Flu Trends would deviate more from [normal] behavior."

Influenza data were gathered from Google Flu Trends, the CDC Outpatient ILI Surveillance Network (CDC ILI Surveillance) and the U.S. Influenza Virologic Surveillance System (CDC Virologic Surveillance). The study found that Google Flu Trends was 25 percent less accurate at estimating lab-confirmed influenza virus infection.

In fact, studies have shown that ILI is only caused by the influenza virus in 20 to 70 percent of flu season cases, according to the report.

ILI is defined as a non-specific combination of flu-like symptoms, including a fever with a cough or sore throat, lacking any laboratory testing.

Ortiz pointed out that it might not always be necessary to differentiate between influenza and ILI.

"It depends on your system," he said. "In some places [differentiating the two types of illnesses] may not be important."

In order to track ILI trends, Google Flu Trends uses the popularity of certain Google search entries in real time to estimate nationwide rates of ILI activity. But these results can't determine whether a person is actually sick.

The Google study published in Nature found that Google Flu Trends has a "close relationship between how many people search for flu-related topics and how many people actually have flu symptoms," according to Google.

The Internet search engine recognized that not every person who searches for the term "flu" is sick, but it noted that there was a pattern when all flu-related search entries were added together.

Although Google Flu Trends isn't as accurate as CDC surveillance programs, it can churn out immediate results, giving health care providers a jump start at preparing for possible flu outbreaks; traditional systems could take days or weeks to collect and release data, according to the report.

"[The] CDC publishes national and regional data from these surveillance systems on a weekly basis, typically with a one- to two-week reporting lag," according to Google's 2009 study.

Ortiz and his researchers recognize the simplicity and immediacy Google Flu Trends provides, and they don't disapprove of its use, but warn that the data should be "interpreted with caution" and that other surveillance systems are more accurate when it comes to reporting influenza activity.

"I really want to stress that Google Flu Trends is a great contribution to global public health, but it does not replace the need for more infrastructure-intensive, basic public health surveillance, including a component of lab-confirmed influenza," Oritz said.

Google did not return calls or e-mails for comment.