Google Flu Trends had increased accuracy for estimating 2012-13 flu outbreaks after modifications made by scientists at Columbia University's Mailman School of Public Health.
The search giant's flu outbreak metric doubled after its erroneous predictions in 2012 when its forecasts were proven correct in 63 percent of the cities in the U.S. two to three weeks before the outbreak's peak. It hit 73 percent accuracy at the top of the peaks.
The flu metric tool predicted the virus propagation by monitoring the frequency and geographical spread of keywords for flu-related queries.
"Google Flu Trends is not a predicting tool. It's a surveillance tool," Jeffrey Shaman, assistant professor of Environmental Health Sciences at Columbia, said.
"What we're doing is we're using a measuring of influenza incidence and using it together with a mathematical model that describes propagation through a population and then forecasting the flu," he added.
The team multiplied the local monitoring data with CDC data that honed in concrete flu cases. This made the tool's prediction more accurate. However, the tool still had its flaws as it faired poorly in predicting the flu outbreak in Chicago.
"Population density may also be important. It suggests that in a city like New York, we may need to predict at a finer granularity, perhaps at the borough level. In a big city like Los Angeles, we may need to predict influenza at the level of individual neighborhoods," Shaman explained.
He added the tool's false prediction in 2012 was due to the number of people going online searching for flu symptoms and the high degree of media handling flu stories.
Shaman noted the flu metric still has its flaws and has to be used "judiciously and intelligently."
The new system will be used as this year's flu season peaks. The data will be posted on a Web site hosted by Columbia.
To contact the editor, e-mail: