A Friends and Family Plan for the Flu

The Friendship Paradox may provide a handy predictor for whether a flu bug will result in a mass outbreak or a few cases of the sniffles.
Author:
Publish date:

A fortune-teller looks into a crystal ball. She sees a network of people, and at the center are the trendsetters. They are contracting the flu. The seer predicts that in two weeks their friends will have the same bug.

Unlike the fortune-teller, predicting the onset of an epidemic is something the Centers for Disease Control cannot do at this time. In fact, the CDC is usually about two weeks behind the curve. But social network researchers may have found an effective predictive tool — a well-known, but to this point unused, barometer of social interaction — known as the friendship paradox.

Central to the paradox is the likelihood that at least a couple of your friends are more popular than you are. That may sound a bit harsh, but when asked to name friends, people generally name a person more connected and more social than they are. And just like their opinions of music or wines may infect others’ thinking, so may their viruses infect other bodies.

The reason for this, says James Fowler of the University of California San Diego, is that “if you’re at the center of a network, you’re a shorter number of steps from everyone else in the network. If someone just randomly infects somebody in the network, it will start to spread from that person and it will spread to the center first because those people are best connected. They really are a crystal ball in this case.”

Making the Connection

James Fowler and Nicholas Christakis aren’t the only ones to connect social media and flu bugs. In spring 2009, when concerns about swine flu ebbed and flowed, Stanford University academics Marcel Salathé and James Holland Jones tracked public emotions about the threat using an Internet survey.

Fowler and Nicholas Christakis of Harvard University tested the friendship paradox to see if a social network would predict a flu outbreak.

In a study published last month in PLoS ONE, they studied 319 randomly chosen undergrads at Harvard who responded to e-mails and agreed to participate in the study. The researchers asked them to name up to three close friends, which yielded another 425 names and became what Christakis and Fowler referred to as the “friends” group.

They tracked the groups as flu season arrived and found that people in the friends group were diagnosed with the flu 13.9 days before those in the randomly chosen group (i.e., the general population).

The beauty of this tool, Fowler said, is that the friendship paradox is a simple, inexpensive method that allows researchers to zero in on those most likely to experience an epidemic early. It also has the potential to predict the spread of other phenomena in networks — for example the toy most likely to be popular at Christmas or the rise of a political issue.

“You need a planning component to go along with the implementation,” he said, because random people have to agree to participate, name friends, and, in the case of his study, report flu symptoms and visits to the doctor.

Since the Harvard study was relatively small and contained, how would researchers go about predicting a flu outbreak across the country?

Fowler said it would be fairly easy to identify 1,000 who were then asked to name up to three friends. “A couple of dozen flu cases in this group would be enough for you to be sure that something was going on, and you’d know it by monitoring the friends group,” he said.

He noted that the same methods used in the study he and Christakis undertook could be used to predict epidemics in cities, among different strata of society and early contractors of a disease and other trends.

Fowler mused about the potential for prediction of any number of things using online tools such as Google Trends — the number of searches for, say, symptoms of the flu would definitely tell you something — and online social networks such as Facebook as ways of predicting other public health problems and phenomena that spread in networks.

“Social networks are conduits through which many things flow,” he said.

subscribe-to-mm

Related