You are here
Software Uses Twitter To Track Dengue Outbreaks In Brazil
submitted by Mary Suzanne Kivlighan
Kaiser Family Foundation - July 19, 2011
The New Scientist reports on a software program that is being used "to identify a high correlation between the time and place where people tweet they have dengue and the official statistics for where the disease appears each season."
Researchers at two Brazilian National Institutes of Science and Technology worked together to create the software, which filters tweets containing the word "dengue" and user location details. "Dengue outbreaks occur every year in Brazil, but exactly where varies every season. It can take weeks for medical notifications to be centrally analyzed, creating a headache for health authorities planning where to concentrate resources," the publication notes. Using Twitter could speed up response time, according to Wagner Meira, a computer scientist at the Federal University of Minus Gerais who led the study (Corbyn, 7/18).
Where are the outbreaks? (Image: Antonio Lacerda/Corbis)
NewScientist - by Zoe Corbyn - July 18, 2011
Twitter To Track Dengue Fever Outbreaks In Brazil
Tracing disease outbreaks down to individual cities via social networking could help track dengue's spread
"MY MOTHER is suspected of having dengue," tweets a woman in Rio de Janeiro, Brazil. "I think I have dengue. Hopefully I'm wrong!" tweets a man in São Paulo, 350 kilometres away.
These short messages posted on Twitter might not seem much, but when the dengue season begins again in Brazil this November, tweets like this could help the country better control outbreaks of a virus that kills hundreds of people each year.
That's thanks to software created by a collaboration between two Brazilian National Institutes of Science and Technology, led by Wagner Meira, a computer scientist at the Federal University of Minas Gerais. The team has used it to identify a high correlation between the time and place where people tweet they have dengue and the official statistics for where the disease appears each season.
Social media has been used in real-time surveillance of diseases before; Twitter was used to follow the 2009 swine flu pandemic, for example. But this is the first time it has been used to track dengue fever, and the first time data on the scale of individual cities has been collected in this way. It is also the first attempt to pick up on people tweeting about their personal experience of a disease.
Dengue outbreaks occur every year in Brazil, but exactly where varies every season. It can take weeks for medical notifications to be centrally analysed, creating a headache for health authorities planning where to concentrate resources. Using Twitter messages could mean a much faster response, says Meira. "It isn't predicting the future but the present," he says. "This means we aren't weeks behind like we used to be."
The approach - helped by increased access to the internet in Brazil - uses software to filter tweets for those that contain the word "dengue" and information on the user's location. Tweets that express personal experience of the disease are identified using criteria such as sentence structure and wording. Those mentioning public campaigns, or telling jokes are all filtered out.
Testing the software on 2447 tweets containing the word "dengue" and a location sent between January and May 2009 shows "personal experience tweets" tightly correlated with the outbreaks identified by the Brazilian Ministry of Health. The work was presented at the Web Science Conference in Koblenz, Germany, last month.
The team now plan to analyse 181,845 tweets sent between December 2010 and April 2011, but are waiting for the ministry's 2011 data before they do so. However, Meira says that the tweets mirror the trend seen in cities where changes in the dengue outbreak are known to have occurred.
Philip Polgreen, an infectious disease expert at the University of Iowa in Iowa City, and his colleague Alberto Segre, a computer scientist, recently published a similar study on the use of Twitter to track swine flu in different regions of the US. They say Meira's work is significant. "They have included sentiment analysis and they have a fine degree of geography," says Segre.
Miera says terms such as "bone pain" and "eye pain", which are typical symptoms of dengue fever, are now being included in his team's software. He adds that the tool will also get better as more Brazilians come online: "Every year we get more data."
Last month, Google unveiled a Google Dengue Trends tool that records spikes in web searches for dengue fever, which taps into the fact that people who have dengue fever are likely to look for information about it.
Epidemiologist John Brownstein, at the Children's Hospital Boston, who worked with Google on the project, published a paper setting out the underlying science last month (PLoS Neglected Tropical Diseases, DOI: 10.1371/journal.pntd.0001206). He says the Google and Twitter methods are complementary; while Google searches are more common, Twitter provides more context.