Sep 16, 2010 (CIDRAP News) – Human "sensors" who are in the hubs of friend networks can detect flu outbreaks at least 2 weeks earlier than surveillance systems that track, for example, doctors' visits and may someday be a useful tool for identifying other diseases and behaviors, researchers reported yesterday.
The novel disease detecting system depends on the "friendship paradox," a theory that friends of given individuals are more popular than they are and are at the center of social webs, where they not only learn of gossip, trends, and ideas sooner, they may also be exposed to diseases earlier than those in more remote parts of friend networks.
The researchers, Dr Nicholas Christakis, professor of medicine, medical sociology, and sociology at Harvard University and Dr James Fowler, professor of medical genetics and political science at the University of California , San Diego (UCSD), tested their friend network disease detection system at Harvard University from Sep 1 through Dec 31 during the second wave of the 2009 H1N1 pandemic. Their findings appeared yesterday in Public Library of Science (PLoS) One.
They wrote that analyzing social networks and monitoring the health of central members is an ideal way to predict outbreaks, but detailed information doesn't exist for most groups, and to produce it would be time-consuming and costly.
Instead, they propose asking a random group of people to name friends, and then monitor and compare illness patterns in both groups. They emphasized that a person's position in the friend network doesn't indicate the actual flu transmission path, but it may serve as a proxy for an unobservable network of flu spread.
Using the "friendship paradox" theory, Christakis and Fowler contacted 319 randomly selected Harvard undergraduates who named 425 friends. Many who named friends were often named by others as friends, and the same person was often nominated several times by other people. In total, the study collected information on 1,789 unique interconnected students.
They monitored the two groups through self-reporting and with data from Harvard University Health Services.
They found that the friends group got sick about 2 weeks before the random-student group. Using a variation of the detection method, they found, gauged by visits to the student health service, that the friends group showed flu symptoms 46 days before the epidemic peak. A figure and movie Web link that accompany the study show flu "blooming" first in the friendship network's more central nodes.
As an alternate method of identifying a high-risk group, the researchers administered a survey to measure subjects' perceptions of their own popularity, but it did not produce an earlier flu diagnosis.
Christakis said in a UCSD press release that current surveillance systems provide only a snapshot of what's currently happening. "By simply asking members of the random group to name friends, and then tracking and comparing both groups, we can predict epidemics before they strike the population at large," he said. "This would allow an earlier, more vigorous, and more effective response."
Fowler added that current detection methods lag the real world, or, at best, report the information in real time. "We show a way you can get ahead of an epidemic of flu, or potentially anything else that spreads in networks," he said. The authors noted that the system could be used to identify drug-use behaviors or even the spread of new ideas and fashion fads.
To put such a monitoring system into practice, the authors suggest that, for example, a university health service could gather a sample of people who are nominated as friends and who agree to be passively monitored for healthcare use, such as doctors' visits. They said local, regional, or national health officials could pull together a random group of people, then enlist their nominated friends to report symptoms with brief, period text messages or an online survey system.
"Since public health officials often monitor populations in any case, the change in practice required to monitor a sample of these more central individuals might not be too burdensome," Christakis and Fowler wrote.
They added that analysts could note when disease incidence in the friends group rose above a predetermined or background rate, or they could track both the friend and random groups and watch for when the two epidemic curves first part ways, which could be an early epidemic signal.
A friend disease detection system could also be used alongside a system that tracks the online behavior in Google and other search engines to provide even better real-time information about a developing disease epidemic.
Dr John Glasser, a mathematical epidemiologist at the US Centers for Disease Control and Prevention (CDC), said in the press release that the "provocative" new study will likely prompt epidemiologists and public health officials to think more about the social contexts of disease transmission.
"This study may be unique in demonstrating that social position affects one's risk of acquiring disease," he said. "Consequently, epidemiologists and social scientists are modeling networks to evaluate novel disease surveillance and infection control strategies."
See also:
Sep 15 EurekAlert press release
Sep 15 PLoS One study
Figure of flu blooming in social network hubs
Video of flu booming in social network hubs
