Connecting the Dots: Network Analysis Helps Public Health Researchers Get the Big Picture
At first glance, the web of interacting molecules that make up a living cell and the peer relationships that influence smoking behavior in a group of middle school students would seem to have little in common. But both are examples of networks: a set of connected actors and the relationships among them.
In recent decades, a small cadre of researchers has turned to network analysis to help find answers to complex, multifaceted problems that underlie cancer and other diseases, from the molecular to the societal level.
Network analysis is the measurement, mapping, analysis, and interpretation of network structures, of the ties between actors, and of the flows that occur within and across networks. This approach, which has its roots in mathematics, is being applied to problems in a broad range of disciplines.
Molecular biologists are using network analysis to get a handle on the complex web of interactions between molecules in a cell and on how defects or changes in various components of these molecular networks can result in diseases such as cancer.
Meanwhile, public health researchers are using network analysis—specifically, social network analysis—to understand complex problems such as tobacco use, obesity, and the HIV/AIDS epidemic. Social network analysis can be used to study how infections, behaviors, attitudes, and information are transmitted among people or groups of people, as well as the adoption of new medical or public health practices.
Networking for Public Health
Public health researchers "now realize that the complexity we're dealing with can't be addressed exclusively with the traditional approaches that we've been using," noted Dr. Elizabeth Ginexi, a program director in NCI's Tobacco Control Research Branch (TCRB).
Social network analysis and other systems science approaches, such as system dynamics modeling and agent-based modeling, "specialize in getting at the relationships between components of a system and the system itself, and particularly how they change over time," added Dr. Patricia Mabry, a senior advisor in NIH's Office of Behavioral and Social Sciences Research (OBSSR).
New tools and techniques for mapping out networks have led network theorists to conclude that "most real networks in technological, social, and biological systems have common designs that are governed by simple and quantifiable organizing principles," according to a 2007 editorial by Dr. Albert-László Barabási, director of Northeastern University's Center for Complex Network Research in Boston. And, Dr. Barabási noted, "networks pervade all aspects of human health."
Signaling Pathways and Tobacco Control
Dr. Scott Leischow, associate director of the Biobehavioral and Social Sciences Research Program at the University of Arizona Cancer Center, developed his interest in social network analysis when he was chief of NCI's TCRB. While at NCI, Dr. Leischow launched the Initiative on the Study and Implementation of Systems (ISIS), which explored the application of social network analysis and other systems science approaches to tobacco control and public health.
At the University of Arizona, Dr. Leischow brings interdisciplinary teams together to do just that. "My goal in studying social networks," he said, "is to figure out where in the network one can intervene to achieve the best health outcome."
In a social network, the actors are usually people or groups of people, such as households, organizations, governments, or even societies, each of which can interact with and influence each other in various ways, directly and indirectly.
One project that Dr. Leischow leads has been using social network analysis to understand how the network of toll-free telephone-based smoking cessation services that make up the North American Quitline Consortium interact and share new knowledge. By sharing his findings on these "signaling pathways" with consortium members, Dr. Leischow hopes to help them better implement evidence-based practices so that they can improve their ability to help tobacco users quit.
Dr. Leischow also recently received a grant as part of the NCI State and Community Tobacco Control Research Initiative to work with Team Navajo, a network of Navajo health advocates and organizations whose aim is to eliminate secondhand smoke from use of commercial tobacco (as opposed to ceremonial use) in all workplaces and indoor public spaces. (The Navajo Nation is sovereign and thus exempt from state laws.) "The problem is most exemplified in places like the casinos, where they allow cigarette smoking and there is a lot of exposure to secondhand smoke, both for people who go there to gamble and among the employees," Dr. Leischow noted.
He and his colleagues plan to analyze the Team Navajo network and its interactions with other organizations and people in the Navajo Nation. By doing so, the researchers hope to help Team Navajo "identify the lynchpin people or organizations"—the thought leaders and key decision makers—with whom they should work to achieve their goal.
Whom You Know Matters
Since the 1990s, Dr. Thomas Valente, a professor in the Department of Preventive Medicine at the University of Southern California (USC), has used social network analysis to understand a range of health-related behaviors, including how peer relationships affect smoking among middle school students. He found that "one of the strongest influences on adolescents beginning and continuing to smoke is having friends who smoke."
Furthermore, Dr. Valente said, "popular students tend to be early initiators of smoking behavior. So if we want to create effective interventions to stem the spread of smoking, we need to recruit those popular students and have them actively involved."
"My goal in studying social networks is to figure out where in the network one can intervene to achieve the best health outcome."
—Dr. Scott Leischow
To see how this would work in practice, Dr. Valente and his colleagues at USC designed smoking prevention programs that use social network data to identify opinion leaders—in this case, popular students in sixth-grade classrooms—and to assign students to groups, "so that the intervention messages are received from peers rather than from randomly selected others," Dr. Valente said. Through a randomized controlled study, the researchers found that their network-based approach was more effective than two other approaches at changing attitudes toward smoking and reducing students' intention to smoke.
In an extension of this work, Dr. Valente and his colleagues "are currently estimating peer influences from networks other than friends, and designing better intervention options," he wrote in an e-mail message.
Dr. Valente also recently led a study that used social network analysis to investigate the effectiveness of a community-based NCI initiative that was designed to reduce cancer disparities among Pacific Islanders in Southern California.
It's Not All about Networks
Dr. Mabry of OBSSR and others at NIH are spearheading efforts to encourage basic research that leads to new applications of social network methods and theory for improving human health. Through its Division of Cancer Control and Population Sciences, NCI and 10 other NIH institutes, as well as OBSSR, are also participating in an NIH funding opportunity announcement that promotes the use of systems science approaches, including network analysis, in behavioral and social science research relevant to health.
"It's a slow train, there's still time for people to get on," Dr. Ginexi said. However, she noted, the goal "is not to replace the existing mathematical and statistical models, but to complement them where it makes sense."
As Dr. Valente, a self-described "avid social network analyst," said in an e-mail message: "Social network analysis will not solve all your problems or provide all the answers, but [it] should be used as a tool with other sources of information. Networks, like many things, can be unpredictable; we still have a lot to learn about network theory."