Mining Twitter to predict ED visits
Don’t think of it as oversharing. Researchers at the University of Arizona at Tucson and Parkland Hospital in Dallas say health-related tweets qualify as epidemiological data.
The researchers combined tweets about asthma with data taken from air-quality sensors and electronic health records and were able to predict with 75% accuracy if the Parkland emergency department staff could expect a high, low or medium number of asthma-related visits that day.
During the three-month research project, Twitter saw millions of tweets containing the words “asthma,” “inhaler” or “wheezing.” But through text-mining techniques, researchers were able to isolate those tweets generated from the ZIP codes found in asthma patients’ EHRs. Worsening air-quality conditions led to simultaneous increases in ED visits and asthma-related tweets. Asthma-related Google searches couldn’t produce the same level of predictability.
“You can get a lot of interesting insights from social media that you can’t from electronic health records,” Sudha Ram, a University of Arizona professor of management information systems, said in a news release.
Ram and colleagues plan to expand the study to 75 hospitals in the Dallas-Fort Worth area. They also want to see whether they can get the same results for diabetesrelated tweets.