Twitter Map Tracks Swine Flu Tweets in Real Time

Friday, May 01 2009

FRIDAY, May 1, 2009 ( — What do you get when you cross swine flu with Twitter, the social networking service much loved (and hated) for its rapid-fire 140-character updates known as tweets? No, it’s not a chatty new virus called Switter.

What you get is a map of the United States with the Twitterverse—at least in regards to swine flu—updated in real time. Click on the colored dots, and you can see who’s tweeting about swine flu (some guy in Texas! A lady in Glasgow!) at any given moment. (Added bonus: The site sheds light on Twitter’s popularity around the world; for example, not a lot of Twitter fans in Montana.)

The University of Iowa’s computational epidemiology research group, known as CompEpi, developed the map, and that’s just the beginning, they say. Their aim is to track disease-related content on blogs and other social media sites too, including Facebook, as part of a program known as SWIM (Social Web Information Monitoring) for Health.

The researchers have the Twitter map up and running because they are using updates that are already open to the public. (Anyone can search for swine flu tweets on Twitter’s site). Facebook, on the other hand, is more private. The researchers have asked Facebook for permission to search for health-related terms in status updates, says University of Iowa PhD candidate Alessio Signorini, who built the Twitter map.

There have been more than 10,000 swine-flu related tweets per hour on some days this week, according to the website Mashable, the social media guide.

Signorini says that social media maps could offer a snapshot of the country’s concerns or confusion (or at least among people who use Twitter and Facebook) during outbreaks. For example, many people tweeted that they were no longer going to eat pork, despite the fact that researchers have been working hard to dispel the myth that you can get swine flu, or H1N1, from eating pork.

“If you know there’s something going on like swine flu, you can capture the mood and the feelings and the perceptions of people in various parts of the country,” says Signorini. “Perhaps officials can better address those concerns in that particular zone.”

The ultimate goal is to track health-related comments of all kinds, and see if they are harbingers of disease outbreaks or health-related behavior, in much the same way that Google Flu tracks seasonal flu by looking at search terms such as “coughing” or “fever,” plugged into Google’s search engine. The University of Iowa’s researchers had already developed a map for general health terms, such as “headache,” before the swine flu outbreak.

“We are trying to understand the perception of an outbreak or [something like] the perception of a new vaccine,” says Signorini, who is also the director of search technology at the social search engine OneRiot, which provides some of the data for the project.

However, it’s still not clear if tweet tracking will be as useful as analyzing search terms. For example, Google Flu engineers worked closely with the Centers for Disease Control and Prevention to make sure that their search-term tracking does indeed correlate with actual illness.

In general, Google Flu sees a rise in flu-related searches about two weeks before the CDC sees a rise in seasonal flu cases, because the federal agency generally relies on reports that must filter from patients to doctors and then to the CDC.

In fact, Google Flu could have picked up the swine flu cases early, as search terms jumped in Mexico before the rest of the world was aware that a new type of virus was emerging. Although Google Flu usually only posts data from the United States, they recently started to include data from Mexico.

As for the Twitter map, the researchers are working on validating that health-related comments—which for swine flu range from inane musings to truly useful information and local updates—actually mean something in the real world.

“If we know if a vaccine is coming out, and we can capture how people perceive this vaccine—they’re unhappy about it, they’re waiting for it, they’re suspicious about it—do those things match with what they do?” says Signorini. “If the public on Twitter is suspicious of the vaccine does it correspond to people actually not vaccinating their child with that vaccine?”

One problem is re-tweets, which are repostings of Twitter comments that people pick up from other Twitter feeds and include in their own, says Brent Csutoras, a social media consultant in San Mateo, Calif.

“The same news report might be re-tweeted and sent around 20,000 to 30,000 times and that would end up having an effect on this,” Csutoras says.

Despite the need for filtering re-tweets, Csutoras likes the map. “It’s very interesting,” he says. “I like the fact that you can see the conversation.”

He notes that other sites are more useful for people seeking information, such as the swine flu Google Map, because it shows verified swine flu cases. Other good sites to check out for swine flu online health information include HealthMap and SickCity, says Signorini. In addition, Mashable offers a good guide to following swine flu on Twitter.

Because of the re-tweeting and casual conversations on Twitter, the map is not currently that useful to individuals but could be helpful to public health officials, Csutoras says.

“It might be interesting to identify different cities where they could use Twitter as a means to communicate to people,” says Csutoras. “You might even say, hey we’re trying to get the word out in California, we can just Twitter it and the majority of people will find it. It might help agencies in some way.”