September 26, 2017

Twitter is used by many to keep up with breaking news and influential views, but School of Public Health Assistant Professor Quynh Nguyen is using it for a different aim: to predict health outcomes. 

Dr. Nguyen, new this fall to the Department of Epidemiology and Biostatistics, analyzed 80 million random Twitter messages—a random 1 percent sample of publicly available geotagged tweets sent that year—to learn more about people’s health in the newly published research “Geotagged US Tweets as Predictors of County-Level Health Outcomes, 2015–2016.” The study was published online Sept. 21 in the American Journal of Public Health. 

Her goal was to understand the prevalent health culture and behaviors of communities through Twitter. Aggregating tweets collected from over 600,000 users, researchers implemented an algorithm to identify the moods of users, and what types of food and exercise people were talking about. Researchers then compared the tweet locations to county-level health outcomes such as mortality. 

Researchers found that counties with more social modeling of behaviors on Twitter around food and physical activity had lower rates of mortality, obesity and physical inactivity (accounting for county demographic and economic characteristics). Happiness and positive sentiment around healthy behaviors were linked to better health outcomes. Higher social media mentions of alcohol use in certain counties related to higher rates of excessive drinking and alcohol-related mortality. 

The study cannot conclude causality because it is based on observations, and Twitter users are not representative of the full U.S. population. While imperfect, Twitter may be useful as an additional source of information on the health of communities by helping to detect health concerns or evaluate the success of health interventions. Moreover, Twitter can be potentially used to watch changes in real time. Twitter is one of those platforms people turn to when major events happen to find out updates, sometimes even before they are broadcasted on the news,” Dr. Nguyen said. Twitter can be used to identify infectious disease outbreaks and aid in disaster response. 

In addition to Dr. Nguyen, who is in the UMD School of Public Health’s Department of Epidemiology and Biostatistics, the study’s co-authors include Matt McCullough, MNR; Hsien-wen Meng, MS; Debjyoti Paul, MS; Dapeng Li, PhD; Suraj Kath, MS; Geoffrey Loomis, MS; Ming Wen; Ken R. Smith, PhD and Feifei Li, PhD from the University of Utah; and Elaine Nsoesie, PhD from the Institute for Health Metrics and Evaluation, University of Washington.

The work was supported by the National Institutes of Health's Big Data to Knowledge Initiative (BD2K).

Article Link: 
Geotagged US Tweets as Predictors of County-Level Health Outcomes, 2015–2016
Related People
Quynh Nguyen