skip navigation nih record
Vol. LX, No. 11
May 30, 2008

next story

Health Phenomena Can Spread via Social Networks

On the front page...

With a little help from our family and friends, we get by—and get chubby. Not only can loved ones influence our weight, they affect other health behaviors, as well as levels of happiness. That’s according to Dr. Nicholas Christakis, professor in the departments of health care policy, sociology and medicine at Harvard University. He recently visited NIH to share his findings in “Eat, Drink and Be Merry: The Spread of Health Phenomena in Social Networks.” The talk was part of the Behavioral and Social Sciences Research lecture series.

“Health, health care and health behaviors are not just individual but collective phenomena,” Christakis told the crowd in the Neuroscience Center at Executive Plaza.


  Dr. Nicholas Christakis  
  Dr. Nicholas Christakis  

The obesity-is-contagious story broke in a 2007 New England Journal of Medicine article coauthored by Christakis and Dr. James H. Fowler of the University of California, San Diego. Their study, supported by the National Institute on Aging, showed how a person is more likely to become obese if a close friend or family member becomes obese, even if that person lives far away.

In a close look at “a densely interconnected social network,” they analyzed data previously collected over a 32-year period in the Framingham Heart Study, an ongoing research pro-ject funded by NHLBI.

“We went to the [Framingham study] record room,” Christakis said, “took paper records and computerized them from 1971 to the present and then linked to ongoing data collection.” They tracked information on 5,124 key participants (“egos”) and their parents, spouses, siblings, children, coworkers, neighbors and close friends (“alters”) who also participated in the study.

“An amazing property of this [Framingham] study,” Christakis said, “was that only 10 people were lost to follow-up since 1971.”

Moreover, as far as he knows, Christakis said, the Framingham Heart Study Social Network dataset is one of very few, worldwide, “that allows the co-mapping of geography and social network ties across time.”

He called seeing the results of the spread of obesity in the network for the first time “the most exciting moment in my scientific career.” In animated slides, the results, showing the cascade of weight gain and weight loss through the network, resembled “a handful of rocks being thrown into a pond,” he said, “not just a single pebble.”

This “multi-centric epidemic” was sufficiently complex that various techniques borrowed from physicist Laszlo Barabasi were used to observe whether clustering was due to more than chance alone.

Christakis offers evidence for the spread of obesity in social networks.

Christakis offers evidence for the spread of obesity in social networks.

Not only was there clustering in the spread of weight gain, but also the effect was linked to social ties extending to three degrees of separation. What counted was the person’s placement in social space, not geographical space.

“It doesn’t matter if your alter lives next door to you or across the country,” said Christakis.

With U.S. adult obesity rates at 30 percent, the term “epidemic” is more than a metaphor; non-biological, person-to-person transmission is a real factor. If your friend becomes obese—that is, has a body-mass index at or above 30—then your likelihood of becoming obese increases by 57 percent, and by 71 percent if it’s a close, same-sex friendship. It increases by 40 percent among pairs of siblings; and by 37 percent in married couples. There is no effect among neighbors unless they are also friends.

As to causes, “we were misunderstood in press coverage,” said Christakis, who went on to set the record straight. He’d considered three causes: homophily, confounding and induction.

Homophily is the tendency for people to choose relationships with people who have similar attributes: “Birds of a feather flock together.” Confounding is a situation in which the effects of two processes are not separated. “My friend and I may share similar exposure to a fast-food restaurant,” he said, “but in this situation, my weight gain is not the cause of my friend’s weight gain.”

Smoking cessation, a possible factor in weight gain, did not account for the effect.

Instead, his findings supported induction. Induction involves the spread of social norms, ideas or changes in expectation. We are influenced by the behavior and appearance of people who are socially—not necessarily geographically— close to us.

In similar studies of smoking behavior, Christakis said that “people are quitting together, not as isolated people,” showing “some kind of collective decision-making, flocking behavior.”

As for the “spread of happiness,” it’s also associated up to three degrees, but works differently from the spread of obesity: “The alter has to live near you,” he said.

In related studies, Christakis has found that we influence each other in many other ways, such as “the widower effect,” in which the death of one spouse increases the death rate of the other; depression treatment in parents might improve children’s immunization rates; hip replacement in one person may reduce the disability of others to whom they are connected. In short, Christakis argued, “because people are connected, their health is connected.”

He concluded that health, health care and health behaviors are “collective social-network-based phenomena [with] specific interpersonal collateral effects.”

During Q&A, an audience member remarked: “This is some of the most interesting and innovative work in sociology that I’ve seen in two decades.”

“Half the people said [our results were] painfully obvious,” Christakis responded. “Half said: You should be tarred and feathered.

“Some use an individualistic idea of human behavior,” he concluded. “We need to understand human behaviors as properties of higher-order structures such as social networks.” NIHRecord Icon

back to top of page