Happy tweets, healthy hearts.

Heart disease is the leading cause of death in the United States. The most prominent risk factors for heart disease are smoking, obesity, hypertension, diabetes, low income, and low education. However, psychological science has taught us that living in a social environment that is hostile and un-supportive also contributes to poor health, specifically heart disease.

In the past ten years, the social environments we interact with have grown exponentially with the introduction of social media such as Facebook and Twitter. This introduces an entirely new dimension of social interaction but also a window of opportunity for psychological science researchers. Specifically, a recent study pursued the question:

Does language on Twitter relate to heart disease mortality?

To answer this question, Johannes Eichstaedt and colleagues collected 50,000 words tweeted between 2009 and 2010 from users all across the United States. They systematically review the words for frequency, content, and the location of the user based on their user profile. Then they gathered county specific data on rates of obesity, smoking, marital status, hypertension, income, education, race, and mortality due to Athlerosclerotic Heart Disease from the CDC for the years 2009 and 2010. The data represented in the study represents 148 million county-mapped tweets across 1,347 counties, and CDC data from 88% of the United States.

They found that combining known physical and social risk factors, including income, education, smoking, diabetes, hypertension, obesity, race and marital status, accounted for about 35% of heart disease mortality within a county. However, language used on Twitter alone accounted for about 42% of heart disease mortality within a county. Combining Twitter language and known risk factors accounted for about 43% of heart disease mortality risk. This suggests that language used on Twitter is an important indicator of health outcomes. But what were these people saying on Twitter that predicted heart disease in their county?

The research team identified 3 categories of language use that specifically predicted increased risk for heart disease mortality in their county: aggression & hostility, interpersonal tension, and disengagement. Anger and hostility was a category comprised of frequent use of expletives. Interpersonal tension was a category comprised of frequent use of words such as “hate,” “jealous,” “fake,” and “drama,” not to mention some more expletives. Finally, disengagement was a category comprised of frequent use of words related to boredom and fatigue. Each of these categories was a significant predictor of increased heart disease mortality in a county.

There were also three other categories. These categories were Skilled Occupations, with words referring to attending conferences, learning, and meeting new people; Positive Experiences, using words that refer to friends, weekends, food, company, and things described as wonderful and fantastic; and finally Optimism, which reflects the use of words reflecting possibilities, achievements, father, goals, success, strength, and courage. Frequency of Twitter content in each of these 3 categories was protective against heart disease risk in counties.

But what does this mean? Saying bad words on Twitter causes you to die of heart disease? Posting angry, hostile tweets causes your neighbors to die of heart disease?

Because this research is cross-sectional, these are just correlations, not causes of heart disease. It’s possible that pre-existing heart disease causes people to be more hostile, angry and pessimistic. In that case, language patterns on social media may be an early sign of undiagnosed heart disease that is an area for future preventive science to explore. It’s also possible that engaging with the world with more anger, hostility, and pessimism causes physiological changes to the body that lead to heart disease. Since we know that stress causes heart disease, this pathway is extremely plausible. However, the people who die of heart disease tend to be older, while the people on Twitter tend to be younger. The people in this study that were tweeting expletives were not the ones dying that year, so there’s something much greater reflected in these findings than what predicts heart disease within an individual.

What these findings really suggest to me is that older people living in communities filled with people who are angry, pessimistic, bored, tired, hostile, and curse a lot, are more likely to die of heart disease. The important assumption being made here is that people behave on Twitter the way they behave in the world. In many ways this isn’t really true. But do you think a person that is mean on social media is also the type of person who honks at older drivers when they hesitate to turn right on red, or run a yellow light? I would venture to say yes. The authors suggest that the “combined psychological character of the community” is being represented by Twitter language in this study, and it has robust associations with health. 

We all live in communities, big and small. Other people matter, but more importantly, your behavior matters in the lives of other people.

Eichstaedt, J. C., Schwartz, H. A., Kern, M. L., Park, G., Labarthe, D. R., Merchant, R. M., ... & Seligman, M. E. (2015). Psychological language on twitter predicts county-level heart disease mortality. Psychological science,26(2), 159-169.