Campaigns to discourage adolescents from smoking commonly focus on the health dangers of the habit. A new study from a research team that includes two Robert Wood Johnson Foundation (RWJF) Health & Society Scholars program alumni suggests that a more effective approach might be to target the impact of peer pressure on adolescents while simultaneously finding ways to discourage them from befriending classmates who smoke.
Those findings emerge from a paper published in the October 2013 edition of Health Education & Behavior by David R. Schaefer, PhD, and Health & Society Scholars jimi adams, PhD, and Steven Haas, PhD. As part of a larger project examining the social network dynamics of health behaviors and funded by the National Institute of Child Health and Human Development, the three developed a statistical model working with data from the National Longitudinal Study of Adolescent Health (Add Health), which tracks adolescents who were in grades 7 to 12 in the 1994-95 school year. Add Health includes a wealth of information about teen’s health, habits, living circumstances, social networks, and more over a 13-year period.
Using a relatively new statistical modeling technique called the stochastic actor-based model (SABM), which allows for an examination of the impact and interaction of multiple factors in a group or network setting, Schaefer, adams and Haas identified the factors with the greatest bearing on youth's smoking behavior—both starting to smoke, and the decision to continue or quit. From there, they manipulated the model parameters to see how those factors interacted, much as an audio engineer might use the sliding controls of a sound mixer to emphasize or diminish particular instrumental or vocal tracks in a song.
The effort required sophisticated modeling techniques intended to capture the social dynamics of the teen world, and then to see how those dynamics affected smoking behavior. The original Add Health researchers "collected network data from every kid in the school," Haas explains, "so we were able to reproduce the entire network, with kids telling us about their friends, so that we could see who the friends of friends of friends were, and so on. Importantly, we are able to examine how those networks and behaviors like smoking co-evolve over time."
In the end, Schaefer, adams and Haas focused on two factors most ripe for real-world interventions. First, they observed that where smokers are socially popular, peers are more likely to emulate their smoking behavior. Second, that dynamic is magnified in social settings in which peer influence is a particularly powerful force. Significantly, while the two factors operate in concert, they might be individually manipulated.
"Our modeling method allows us to see how those two things move together over time," Haas says. "We asked, 'What if we could manipulate how desirable smokers are within a network?' You can imagine an intervention to accomplish that, and it would let us see how that would influence the level of smoking within the network. And then there might also be some sort of intervention that manipulates the amount of influence kids have on each other. So when we did that through modeling, what we found is that if you just manipulate peer influence, you don’t see a lot of difference in levels of smoking. But when you simultaneously manipulate how popular smokers are, you see a lot more impact. If kids who smoke are really popular then you see a lot more range. The two things interact with each other, producing a multiplier effect."
Overall, Haas says, the research suggests that "looking at who kids choose as friends is an important untapped source of interventions. Rather than simply trying to get kids to alter their smoking behavior, there is great potential in shaping who kids choose as friends and why. For example, in addition to trying to build up kids' resilience to the influence of their friends, we should also design interventions that make smokers less desirable as friends."
An Innovative Approach
The study is also noteworthy for its methods, because it marries the SABM approach with simulations derived from actual data. The combination, adams says, allows the researchers to “estimate the potential effects of interventions, by ... going under the hood. ... The result is that our simulated intervention effects are not solely theoretical models of what interventions could look like. Instead, estimates of the potential effects of interventions on real-world peer influence and adolescents’ friendship choices” are based on actual data about the role of those factors.