The relative importance of different measures of peer smoking on adolescent smoking behavior: cross-sectional and longitudinal analyses of a large British cohort.
Holliday JC., Rothwell HA., Moore LAR.
PURPOSE: To examine the relative importance of smoking behavior of best friends, boyfriends/girlfriends, the wider friendship group, and school year group smoking prevalence as correlates and predictors of smoking behavior (peer influence). The article also aims to assess the relative extent to which smoking behavior is associated with changes in smoking among peers who are selected to be friends (selective association). METHODS: Using two waves of data collected from 4,145 year 8 (12-13 years) and year 9 (13-14 years) students in 29 schools, logistic regression models estimated the cross-sectional association between four year 8 peer influence variables and weekly and occasional smoking, and the extent to which these peer influence variables and three selective association variables were predictors of weekly smoking in year 9. RESULTS: The smoking behavior of best friend, boyfriend/girlfriend, the wider friendship group, and school year group prevalence were cross-sectionally associated with higher odds of weekly and occasional smoking. In longitudinal multivariate influence models, only the smoking behavior of boyfriend/girlfriend and the wider friendship group were associated with weekly smoking. Weekly smoking was associated with maintaining or changing to having smokers among best friends, boyfriends/girlfriends, and the wider friendship group. In models including all variables, only selective association effects were observed. CONCLUSIONS: Although univariate analyses indicate an association of friends', peer group, and year group smoking with current and subsequent smoking behavior, multivariate longitudinal analyses indicate that simple peer influence models do not completely explain adolescent smoking and that a more complex interrelationship exists between smoking, peer's smoking and peer socialization.