Can Social Media Predict the Next Measles Outbreak?
Posted: June 9, 2015 by Sarah Leavitt in Nursing Newsroom
Epidemiologists, researchers and other health officials are using data mined from social media to identify disease outbreaks earlier than ever before. And in some cases, they’re predicting the spread of illness before it happens.
Traditional methods of identifying an outbreak are slow. Information must be collected from doctors and hospitals and then analyzed. Because the information collected is aimed at specific diseases, these methods aren’t great at spotting new or emerging diseases or a potential outbreak before it actually begins. And traditional methods generally don’t account for people who are sick but don’t seek professional help.
Harvesting information from social media fills this gap, because even though they may not visit a doctor, these people are likely to complain about feeling sick on social media outlets. Searching location data and keywords (for example, users who mention the rash or spots associated with measles) can let health officials track cases before a diagnosis has even been confirmed.
Public health and social media
It’s not just health officials scanning and gathering data from sites like Facebook and Twitter. Using social media for public health purposes represents a back-and-forth relationship. Users can get better information about public health risks from social media sites and make decisions about the risks they’re exposed to.
For example, foodborne illness, like salmonella or other food poisoning, has long been a concern for anyone who dines out. The Centers for Disease Control and Prevention (CDC) estimates that around 48 million Americans get sick from food each year. And roughly 75 percent of those outbreaks are from food that was prepared by a restaurant or caterer.
City and county governments are responsible for inspecting restaurants and rating them for health and safety. But unless that rating is posted visibly in the eating establishment itself, which isn’t always required, it can be hard to find the score. Enter Yelp, which is now posting the scores restaurants receive alongside reviews. So now diners can make decisions about where to eat already informed about their potential choice’s violations.
Social media and food poisoning outbreaks
Review sites can help identify outbreaks as they’re happening, too. Yelp reviewers often focus on the same aspects of an eating establishment that health inspectors do, like cleanliness. So identifying key phrases like “dirty” or “got sick” can help health inspectors narrow down which restaurants they should pay the most attention to, and in the case of illness, where the problem started.
Two recent apps are making the potential a reality, by pairing food-related illness mentions on Twitter with local health department data.
- Foodborne Chicago, a partnership between volunteer app developers and the Chicago Department of Public Health, searches for tweets that may indicate a user recently contracted foodborne illness, and then encourages that user to file a simple online report. The reports are tracked, helping the city understand the size and severity of an outbreak.
- nEmesis, a similar project underway with researchers at the University of Rochester, uses location data from Twitter to find people who complain of symptoms for up to 72 hours after a restaurant visit.
Public health concerns aren’t just limited to food. Sites like TripAdvisor and Oyster, which focus on hotels, could be useful sources of information about outbreaks of bedbugs or other problems.
Is crowdsourced information accurate?
Skeptics question whether data taken from social media is as accurate as traditional reporting methods. And researchers concede that figuring out what’s relevant and what’s not is key to producing meaningful insights that can improve public health.
There are examples that cast doubt on crowdsourced illness information, like what happened with Google Flu Trends. The web service, which uses Google search data to estimate flu infections, greatly overestimated flu prevalence multiple years in a row. And though Google has since revised the algorithm, it’s still considered to be overshooting the number of predicted cases.
Another concern is self-diagnosis. Since a lot of people don’t go to the doctor, they may tweet or post on Facebook that they have the flu, when really, they just have a bad cold. This information, even though inaccurate, is picked up and potentially included in outbreak prediction modeling.
But, in a few well-documented tests, researchers were able to show a high correlation between shared disease data and reported illness rates taken from traditional collecting methods, suggesting that accuracy is possible.
Using social media to improve health care
Many researchers think it’s not a one-or-the-other answer. Instead, social media data can be used to improve and add to the information already being collected.
Foodborne Chicago’s developers point out their app is only successful with the enthusiastic participation of local health departments, who document and confirm cases.
And in the case of Google Flu Trends, it wasn’t intended to replace traditional information gathering or diagnoses methods. But it can provide a signal of increased flu infections a week or two before the CDC’s surveillance reports do. So used in partnership with more traditional methods, it can help researchers see more of the picture, earlier, of an outbreak.Learn More: Click to view related resources.
- David Lazer, Ryan Kennedy, Gary King, Alessandro Vespignani, "The Parable of Google Flu: Traps in Big Data Analysis," Science
- "Using Online Reviews by Restaurant Patrons to Identify Unreported Cases of Foodborne Illness – New York City, 2012-2013," Centers for Disease Control and Prevention
- Michael Luca and Luther Lowe, "City Governments Are Using Yelp to Tell You Where Not to Eat," Harvard Business Review
Back to: Nursing Newsroom