If you are what you tweet, a lot of Americans are very, very jittery. A new study looked at nearly 80 million tweets and found that coffee was the most tweeted-about consumable in the U.S. and Starbucks was by far the most tweeted fast-food restaurant.

The study, published in the Journal of Medical Internet Research Public Health and Surveillance, used Twitter as a way to gain insight into the health of people in neighborhoods across the U.S. “Increasingly we’re seeing more and more studies looking at health beyond just disease, incorporating indicators of wellbeing,” says Quynh Nguyen, lead author of the study and an assistant professor at the University of Utah College of Health. She and her team zoomed in on geolocated tweets about food, physical activity and happiness, and compared them to census data and health surveys.

In just 140 characters, tweets about these three things said a lot about people in certain parts of the country.

Nearly 5% of the tweets, randomly collected from 2015-2016, mentioned food. The most-tweeted edibles were coffee, beer, pizza, wine, chicken, BBQ, ice cream and tacos. When researchers sorted these tweets by healthfulness, 16% were about healthy foods, and 9% mentioned fast food.

The team designed algorithms to determine food and exercise mentions as well as emotion, but they weren’t perfect. Researchers were baffled by the immense popularity of curry before realizing that Twitter was crazy for basketball star Stephen Curry, not the spicy dish.

The most popular fast-food restaurant was Starbucks—which took up nearly half of all such mentions—followed by Chipotle, Taco Bell, and Buffalo Wild Wings.

Tweets from poorer neighborhoods were less likely to mention healthy foods, and areas with more healthy-food tweets had fewer deaths and lower rates of chronic disease.

Tweets about exercise also had stories to tell. The most-tweeted kinds of physical activity were walking, dancing, running, workouts, golf, swimming, hiking, yoga and bowling. And tweets about physical activity tended to come from places with lower rates of obesity and fewer deaths.

“What was kind of nice, from a health and emotion perspective, was that tweets that mentioned food were actually happier than tweets that did not mention food, and tweets that mention healthy foods were the happiest,” Nguyen says. “Healthy food and physical activity were the happiest kind of tweets.”

Nguyen believes that with more refinement, messages from social media like Twitter can provide untapped insights into our health. “So far, we are finding that they do predict area-level health outcomes at various levels: zip code, census tract, county and state,” she says. “Our next set of analyses examine whether these social environment variables predict individual-level health outcomes.”

You can see an in-progress map of their results here.

 

This article originally appeared on Time.com.