You are what you tweet: tracking public health trends with Twitter

Two Johns Hopkins University, US, computer scientists, Mark Dredze and Michael J Paul have used 2 billion public tweets taken between May 2009 and October 2010 to determine public health trends.

They wanted to find out whether Twitter posts could be a useful source of public health information, and indeed, found that they could. The scientists fed the 2 billion tweets into computers and used software to filter out the 1.5 million messages that referred to health matters.

They reckoned that the data gave them something that perhaps not even the tweeters’ own doctors were aware of, like which medicine available over the counter they were using to treat their symptoms at home. They further sorted the tweets into “piles” and uncovered intriguing patterns about allergies, flu, insomnia, cancer, obesity, depression, pain and other ailments.

Some tweets shared the medicine used, and some displayed some people’s serious medical misconceptions, and misuse of medicines, such as taking antibiotics for the flu, which is a virus, so antibiotics will not treat the flu. This practice could contribute to the growing resistance to antibiotics.

The computer filters were also able to split allergy tweets into sneezing/sniffling types, and skin rash/hives types, as well as into geographical areas, which meant that the scientists could track the allergy and flu seasons’ beginnings and ends. The allergy season started earlier in the warmer states, and later in the Midwest and Northeastern US states.

Dredze and Paul have begun to talk about how this system could uncover more useful data, but cautioned that because users would normally only tweet once about an ailment, it would be difficult to track how long the illness lasted and whether it recurred. Twitter users are currently mostly in the States, making it only a useful tool for America. There is also a limit to what most Twitter users will share in public. However it is proving to be an interesting research tool.

Source: Futurity

Click here for more research reports


First Published July 2011

Top of page