New research from Brigham Young University says such posts on Twitter could actually be beneficial to health officials seeking an early warning sign of outbreaks.
The study analyzed 24 million tweets from 10 million unique users. The study found that only 15% of tweets have accurate location data. This information was obtained from user profiles as well as tweets that contain GPS information. That's likely a critical number for an early-warning system to be able to detect terms such as "fever," "flu" and "coughing" in the state or city.
"One of the things this study demonstrates is that the distribution of tweets is about the same as the population distribution, so we have a decent representation of the nation," said BYU professor Christophe Giraud-Carrier.prioritygame "That's another excellent point of validity particularly if you want to examine things like the spread of diseases."
Professor Giraud-Carrier (@ChristopheGC), and his computer science students from BYU share their findings in a recent issue of The Journal of Medical Internet Research.
The researchers found significantly less information than they anticipated from Twitter's feature that enables tweets to be tagged with a location. Only 2 percent of tweets included the GPS information, according to the researchers. That's a much lower rate than the percentage Twitter users report in surveys.
Giraud-Carrier has stated that there is a gap between what you think you are thinking and what you actually do.
User profiles are often used to discover and interpret the location information. Of course, some people utilize that location field for an amusing joke, i.e. "Somewhere in my imagination" or "a cube world in Minecraft." Researchers have confirmed that the information provided by users was accurate 88 percent of the times. In addition to the jokes, a large portion of the inaccurate data is due to people tweeting while traveling.
The end result is that public health officials can capture state-level info or better for 15 percent of tweets. This is a good sign for the viability and feasibility of a system based on Twitter for monitoring diseases to complement the data that has been confirmed from hospitals with sentinel surveillance.
"The first step is to look for posts about symptoms tied to specific location indicators and begin to plot them on a map," said Scott Burton, a graduate student and lead author of the study. "You can also find out if people are discussing actual diagnoses or self-reported symptoms , such as "The doctor told me I'm sick with the flu.'
The computer scientists collaborated with two BYU health science professors on the project. Professor Josh West claims that Twitter's main advantage for health officials is speed.
"If people in a certain area have similar symptoms, and they are reported on Twitter, health officials might issue a notice to health professionals to prepare for something," West said. "Under circumstances like that, it can be very beneficial."
BYU undergraduate Kesler Tanner is a co-author on the study. He developed the code that was used to collect the data from Twitter. When he graduates in April, he'll head into graduate school to obtain a Ph.D.
