Twitter as a Public Health Tool: language-based prediction of heart disease deaths

Twitter is a really great way to let all your followers know what's on your mind and to share cat videos, and has been used to start social movements and launch careers, but did you know it can also predict coronary heart disease? According to a new study published in Psychological Science, Twitter can be used as an indicator of a community's well-being, and may become a powerful public health tool.

Coronary heart disease is the most common kind of heart disease, caused by a build-up of plaque on the inner walls of the arteries of the heart. This accumulation restricts blood flow, which can lead to heart attacks. The risk of coronary heart disease increases with the presence of certain factors, including smoking, high blood cholesterol, high blood pressure, and stress. 

So what does this have to do with Twitter?  Well, the researchers, led by Johannes Eichstaedt from the University of Pennsylvania, found that by analyzing the language patterns on Twitter, they could more accurately predict heart disease-related deaths than models that included 10 different demographic, socioeconomic, and health-related risk factors. The authors had previously shown that language use on Twitter is as accurate a predictor of mood as are traditional questionnaires assessing emotional state. They looked further to see if Twitter could be used as an epidemiological tool for mood- and stress-related illnesses. Using tweets published in 2009-2010 with an author location, the team used emotional dictionaries and generated word clusters that reflected specific attitudes and behaviors. The tweets covered 1300 U.S. counties, representing 88% of the country's population. 

Language patterns reflecting negative emotions, such as anger, frustration, fatigue, and disengagement, were correlated with heart disease mortality. This was true even when socioeconomic factors, like income and education levels, were taken into account. Positive emotional language had the opposite effect. While this approach can't predict when heart attacks will happen, it does provide information about where to target public health efforts to address coronary heart disease.

The authors also speak to the influence our neighbors and communities have on us, since heart disease deaths were clustered to specific areas. That means that having angry neighbors may be just as dangerous to your heart health as being angry yourself.