Can Twitter Predict Your Flu Shot?

Studying how Twitter spreads opinions about flu shots

(RxWiki News) Tweets about vaccines tend to carry strong opinions in 140 characters. If you follow a Twitter user who is against vaccination, what's the chance that you or your family won't get shots?

That's what scientists are trying to figure out, by mapping opinionated tweets about the 2009 H1N1 (swine flu) vaccine.

As fears about vaccination spread and pockets of unvaccinated Americans grow, researchers hope that identifying misconceptions and targeting areas for health information can reverse the trend.

"Ask your doctor about flu shots."

Dr. Marcel Salathé, Assistant Professor of Biology at Penn State, wanted to know if the opinions about vaccination expressed in tweets reflect attitudes, or actually help spread them. His results are publicly available in the online journal PloS Computational Biology.

Researchers have already used data from search engines to track the seasonal flu. As people search the web for information about the flu, Google Flu Trends collects the information and maps it to where people are experiencing flu symptoms with surprising accuracy.

Vaccination has become a controversial subject in recent years. Studies have found that the number of people who refuse to vaccinate themselves and their children is growing, and resisters tend to cluster in certain parts of the country.

While many of their reasons are unsupported by scientific evidence, the past years have proven that anti-vaccine attitudes translate into a change in behavior. Dr. Salathé wanted to see if the trend was also reflected in tweets.

He tracked tweets about the 2009 swine flu pandemic from August 2009 until January 2010. He wrote, “During this time, pandemic influenza A(H1N1) was spreading nationwide but a vaccine became widely available only very late in the year.”

The tweets he studied had keywords related to vaccination, as well as location information about the tweeter. His research team also collected information about who “followed” the tweeter, enabling them to see how information flowed and was responded to.

Once they had gotten 478,000 tweets, they manually categorized 10 percent of the tweets into categories: for, against or neutral about vaccination. They used those tweets to create an algorithm that classified the rest of the tweets.

When they mapped the location data to the categorized tweets, they found that their information from Twitter matched up with the vaccination rates that are estimated by the Centers for Disease Control and Prevention (CDC) with traditional phone surveys. That means that Twitter is a fairly accurate way to determine where people are – and aren't – getting vaccinated.

“Looking at the online social network, we find that both negative and positive opinions are clustered, and that an equivalent level of clustering of vaccinations in a population would strongly increase disease outbreak risks,” the study authors wrote.

In other words, people who tweet against getting vaccinated are at a higher risk for disease and their areas are at risk of disease outbreaks, because it's likely that they are unvaccinated.

Dr. Salathé told New Scientist, "If we know where people are particularly misinformed, then we know where we should do a better job at informing."

The paper was published in October 2011, and the authors stated that they have no competing interests.

Reviewed by: 
Review Date: 
July 27, 2012