Twitter: The Best Flu Tracker There Is?
Tweeting just got a lot more serious. In addition to giving you the latest on who just got Catfished or what Gaga’s wearing today, Twitter is proving to be a fairly accurate and useful gauge of flu trends.
Researchers at Johns Hopkins University realized a couple of years ago that they could track flu trends by analyzing tweets. Public Health experts loved the idea since the traditional method of tracking the spread of flu—via reports to the Centers for Disease Control and Prevention (CDC)—is time-consuming. The information from the CDC lags several weeks behind what's really happening.
Knowing where the flu is and how many cases are occurring helps health officials to spot potentially severe flu seasons and make plans for the distribution of vaccine and other medical resources.
Now, the same team at Johns Hopkins has tweaked their invention to make it more accurate. They've found a way to distinguish between tweets that reflect someone who has the flu from someone who's just talking about the flu.
"We've gone back and asked, 'How accurate is the data?" Mark Dredze, an assistant research professor in the department of computer science at Johns Hopkins, told TakePart. "We started to notice that a lot of things in it didn't make sense. It was giving us flu-related messages that clearly were infection. It was chatter."
Ignoring those later tweets is important. For example, one recent surge of tweets concerned L.A. Lakers basketball star Kobe Bryant's flu-like symptoms. "There was no change in the infection rate," in the population, Dredze says. "That was a tip-off that something was going on here." The computer model uses statistical methods based on language processing technologies. But Dredze and his team were able to alter the program to recognize that statements such as "I have the flu" are different from "I'm worried about getting the flu."
They recently tested the results of their new method by comparing it to CDC data on flu from November and December and found it tracked closely with the official statistics. Adds Dredze, "We're trying to understand the meaning of what people are trying to say. What we have now is a more accurate curve."
The researchers believe they can continue to refine the algorithm to make it even more accurate and detect flu trends even earlier in the season. "I think when we started working on this we didn't know what to expect. We were asking then, what can this be used for? What are the possibilities?" he says. "As we've gotten to know more about it, I think it has a potential to be a fantastic tool, and I don't think we've hit the limit for that. We can continue to make this better."
Research is now underway to look at how the program can be used to identify or track other public health trends. "I fully expect it's going to get better," he says. "I'm looking forward more to the implications this has for other problems. There are many areas of public health where we have no information or little information. I think that is where Twitter can be real exciting."
In the meantime, Twitter users have nothing to fear by the scientific use of their tweets. Dredze points out that the tweets are analyzed with computers to look at the populations—not private individuals. "These are all publicly available tweets," he says. "Nothing here is using any information that is not freely available and shared by people...We're looking for how many thousands of people are tweeting about flu."
Would you trust a flu prediction based on tweets? When you're ill, do you tweet about it?