Wikipedia Page-Views Can Predict Disease Outbreaks

Trying to stay ahead of outbreak trends is not an easy task. It usually involves laboratory tests, collecting data, calls to doctors’ surgeries and trying to keep track of the number of people who visit health facilities. This process might be accurate, but it is also slow and expensive, and by the time an outbreak is being announced to the publicly, it had already been in effect for a couple of weeks.

But some scientists in the US have published a new study in PLOS Computational Biology with a promising suggestion. They say that the number of views a disease-related page on Wikipedia has could actually predict an outbreak of a disease nearly a month before official health advice.

The team, from Los Alamos National Laboratory in New Mexico, explained that this was because people check out their symptoms online before seeking professional medical help. So, as part of their study, the researchers tracked the number of page-views on related pages between 2010 and 2013.

Wikipedia is an open-data site, which means that hourly traffic data on all pages is publicly available. But because this data doesn’t indicate which country a Wikipedia search has been made in, the language the information was written in was mapped by the researchers. This was able to give them an potential location of the person making the search.

The data was then compared with national health surveillance information on disease outbreaks. And using their Wiki-system, the team were able to forecast outbreaks of influenza, tuberculosis, and dengue fever in different countries. In eight out of 14 cases, this forecast was up to four weeks in advance of the news being announced by health officials.

Dr Sara Del Valle, lead researcher and co-author of the paper, said that a disease-forecasting system that can work on a global scale would greatly benefit the way we respond to epidemics.

“In the same way we check the weather each morning, individuals and public health officials can monitor disease incidence and plan for the future based on today’s forecast,” Dr Del Valle explained. She added that the ultimate aim of the research was to create a disease monitoring and forecasting system with both open data and open source code and that the research paper demonstrated that this was possible.

At the moment, it is unclear how the system would be able to work in a region where there is little-to-no internet access - and therefore limited access to Wikipedia. However, the team has indicated that it might be possible to create a trainable prediction model for their system. This would take the available data from one region and apply it to another region where the information is either less readily available or less reliable.

Although some experts have expressed some doubts about how such a system could be applied on a global scale, the US scientists have demonstrated that we may have to look for out-of-the-box solutions. And as the team pointed out: “Our Wikipedia-based approach is sufficiently promising to explore in more detail.”