Programma

News

July 24, 2017

Harvard University researchers use Google data to track dengue fever

The researchers with Harvard University developed disease tracking tool "AutoRegression with Google search queries" (ARGO) several years ago. Then used to track influenza, the team has now used ARGO to improve dengue surveillance. Their research paper in PLOS shows that their tool can be used to improve the tracking of dengue activity in multiple locations around the world.

Share this article

Dengue fever is a mosquito-borne disease, most commonly contracted in tropical areas of the world, including South America, Africa and the Caribbean. Almost 400 million infections occur worldwide each year, with about 96 million resulting in illness.

The symptoms include high fever, skin rash and mild bleeding. Left untreated, serious problems could develop, resulting in dengue shock syndrome. Unfortunately, it is often difficult to monitor the disease with traditional hospital-based reporting. This means that infections are often not identified fast enough, and the disease can keep spreading.

Combining traditional and new

The new tool, developed by Harvard University’s Shihao Yang and colleagues, links together multiple sources of information, using both traditional and non-traditional sources. It combines Google search data with historical government-provided clinical data. The team used the Google Trends tool to identify the top ten queries made by users, which most highly correlated with the term ‘dengue’ in each country. Both datasets were put in ARGO.

The researchers have evaluated the use of ARGO in five countries/states around the globe: Mexico, Brazil, Thailand, Singapore, and Taiwan. The goal was to estimate dengue activity one month ahead of the publication of official local health reports, retrieved from local governments.

Outperforming other methods

The results estimated by ARGO were compared to results from five alternative methods. ARGO outperformed these methods in four out of five countries: Brazil, Mexico, Thailand and Singapore. These four countries were more likely to suffer from seasonal dengue outbreaks. Taiwan, on the other hand, experienced little to no dengue prevalence for years until two epidemic spikes occurred in 2014 and 2015. These are obviously harder to predict.

The combination of historical data with Google search data seems to be the key to accurate near real-time dengue activity estimates. The method should be more finetuned before being used to make decisions on a local level, but could be used to alert governments and hospitals when elevated dengue incidence is anticipated. And that can lead to alert systems for people with an increased risk of exposure to the dengue virus, both travellers and citizens.

Share this article

Don't miss the most exciting developments