By analyzing what people write on Twitter, researchers can help predict rates of obesity, diabetes, teen births, health insurance coverage, and access to healthy foods in counties throughout the United States. Usually demographic information–such as age and race, as well as marital status and income–are used to model health. But information from Twitter can greatly help improve public health predictions, according to a study to be presented at a scientific meeting in late April.

“Twitter activity provides a more fine-grained representation of a community’s health than demographics alone,” said Aron Culotta, a researcher at the Illinois Institute of Technology, in a statement. “The reason for this appears to come from the insights Twitter provides into personality, attitudes, and behavior, which in turn correlate with health outcomes.”

Culotta collected 4.3 million tweets from people in the 100 largest counties in the U.S. He then performed a linguistic analysis, and found a few interesting and occasionally counterintuitive things:

Researchers can also look for more obvious signs of illness, like “staying home with a sore throat.”

This is not the only instance of researchers using Twitter to examine public health. A study at the University of California, San Diego, for example, is using the platform to monitor depression and mental illness. Twitter has also been used to track HIV outbreaks and reports of food poisoning.