First, Google Flu Trends was touted as a potential epidemic detection tool. Then researchers at Johns Hopkins School of Medicine developed a Twitter screening method for delivering real-time data on flu cases to determine which publicly available tweets were linked to actual infections.
Now, researchers at Los Alamos National Laboratory in New Mexico say that tracking Wikipedia page views can forecast the spread of influenza and dengue fever. In a study published this month in PLOS Computational Biology, researchers show that through use of an algorithm, they can connect relevant Wikipedia searches with information from the Centers for Disease Control and Prevention for real-time disease prediction.
The researchers say their algorithm allows them to overcome challenges such as weak scientific peer review and underdeveloped forecasting capabilities that hamper the reliability of other similar data surveillance methods based on Internet information.
"Using simple statistical techniques, our proof-of-concept experiments suggest that these data are effective for predicting the present, as well as forecasting up to the 28-day limit of our tests," the researchers say. "Our results also suggest that these models can be used even in places with no official data upon which to build models."
Still, not everyone is sold on the idea of using Wikipedia to predict disease outbreaks. Heidi Larson, of the London School of Hygiene and Tropical Medicine, told BBC News that she was "wary" of such use, despite her interest in the method.
"There are different things that drive people to Wikipedia, sometimes a new piece of research can drive people to go online," she said. "For issues like Ebola, I don't think people at the beginning of the outbreak in West Africa would have [been searching], because they wouldn't have had [Ebola] before."
Early last year, it was reported that Google Flu Trends drastically overestimated peak flu levels. The researchers of this latest study acknowledge the failure, saying that was caused, in part, by media activity.