Resubmission: Social network analysis of the dynamic relationship between adolescent smoking behaviour & peer influence

There is increasing concern regarding the health behaviours of adolescents. Peer pressure is often cited as a key factor explaining adolescents? uptake of risk behaviours, yet there has been surprisingly little scrutiny of the process by which this occurs. It is most obviously exerted through overt pressure to experiment with tobacco, alcohol or illicit drugs, and interventions aim to develop adolescents? skills to resist peer pressure. Peer influence also operates more subtly, through the development of normative beliefs, in which, for example, adolescents believe that the majority of the population smoke, and that it is a normal part of teenage life to take up smoking. Adolescents are keen to fit in with their peer group, and model their own behaviour on those around them. However, there are other mechanisms at play, including peer selection, in which those with similar behaviours develop friendships. Adolescents wishing to become friends with other smokers, may take up smoking as a gateway to joining that friendship group. The combination of peer influence and peer selection leads to the development of friendship groups in which the prevalence of smoking and other risk behaviours is high. These processes contribute to marked differences between schools in smoking prevalence.

This project will use an existing dataset, collected from a recent MRC funded study of 10,000 students in 59 schools in South Wales and the Bristol area. Students were asked on three occasions to name their six best friends, and about their friendship. Using new statistical methods, these data will allow the research team to disentangle the relative importance of peer influence and peer selection. Those non-smokers most at risk of smoking uptake will be identified, in terms of their own characteristics, parental influence, and their current friendship groups. The stability of these patterns across schools will be identified. In the study, reported smoking was validated through the collection of saliva samples, and combined with the friendship data, this will allow the accuracy of students? reports of their friends? smoking to be assessed. The results of this work will have important implications for the design of future efforts to reduce adolescent uptake of smoking and other risk behaviours. There is immense public and media concern on this issue, and the study team will be sure to produce press releases of the work, as well as featuring the study and its results on the Cardiff University website.

Peer pressure is often cited as a key factor explaining adolescents? uptake of risk behaviours, yet there has been surprisingly little scrutiny of the process by which this occurs. It is most obviously exerted through overt pressure to experiment with tobacco, alcohol or illicit drugs. Social influence intervention programmes aim to develop adolescents? skills to resist peer pressure, yet their effectiveness is limited. The strong association between adolescents? smoking behaviour and the smoking behaviour of their friends may also be due to peer selection processes, or be partly due to projection effects in reporting of friends? smoking behaviour. Peer selection and influence processes will also be mediated by social network position and by the cohesiveness of the social network. These complex processes, and their relative importance, are poorly understood. Existing studies have used sub-optimal analysis methods, have used small samples, and are of limited generalisability. This project will use an existing dataset, collected from the recent MRC ASSIST trial. This was undertaken in 59 schools in South Wales and the Bristol area, in which 10,000 students completed questionnaires on four occasions over a 30 month period, following them from the age of 11-12. They were asked about their smoking behaviour, that of their friends and family, and of their attitudes and beliefs about smoking. Saliva samples were collected, for cotinine assay to validate self-report smoking data. In addition, students were asked on three occasions to name their six best friends, and about their friendship. Response rates at each data collection were at least 95% for the questionnaire data, and 90% for the saliva samples. This dataset provides an internationally unique resource to apply recent developments in the dynamic analysis of social networks and behaviour, to undertake a thorough analysis of the relationship between (i) smoking behaviour and (ii) peer influence, peer selection, social network position and social network integration. These dynamic modelling methods, implemented in SIENA software, consider selection and influence effects simultaneously, and their interactions with student characteristics. The validity of students? reports of their friends? smoking behaviour will also be assessed. These analyses will provide an important insight into the complex interplay between peer relationships and smoking behaviour that will contribute to understanding, theoretical development, and improved design and targeting of interventions to combat smoking in adolescence.