Multilevel predictors of smoking initiation among adolescents: Findings from the Minnesota Adolescent Community Cohort (MACC) study

Abstract

Objective

To understand how factors at multiple levels of influence impact adolescent smoking initiation.

Method

Data from the Minnesota Adolescent Community Cohort, a population-based cohort, were analyzed. Adolescents were recruited from randomly selected geopolitical units (GPUs) in Minnesota at ages 12 to 13 (n = 1953), and were surveyed every six months (2000–2006) until 18. The association between baseline social factors and smoking initiation was analyzed using logistic regression. Linear regression was used to analyze predictors and age of initiation among smokers (n = 603).

Results

Higher proportion of 15–16 year-olds who smoke at the area-level (GPU) was associated with younger initiation (15.47 vs 15.87, p < .05). Higher proportion of the population employed and higher median household income were associated with older initiation (15.90 vs. 15.56 p < .05). Parent education, living with parents or siblings who smoke, living in homes that allow smoking, and having friends who smoke at baseline were associated with smoking initiation or younger initiation (p < .05). Participants whose parents had less than a high school education were 1.6 times more likely than those with college educated parents to have smoked at least a whole cigarette (CI = 1.06, 2.26).

Conclusion

Factors at multiple levels of influence effect adolescent smoking initiation. Smoking by older age peers and lower SES predicts earlier smoking.

Introduction

Over 80% of adult smokers tried smoking before the age 18 (CDC, 2010); two thirds experiment with smoking by age 15 (Johnson and Hoffmann, 2000). The probability of cessation among adults is inversely related to the age of smoking initiation (Breslau and Peterson, 1996). Given that most initiation of tobacco use occurs in adolescence and cessation is so difficult in adulthood, preventing smoking initiation among adolescents is critical to controlling the public health implications. Most research on predictors of adolescent smoking behaviors has emphasized individual or socio-demographic characteristics (Griesler et al., 2002, Kandel et al., 2004). Yet smoking prevention policies and programs are implemented at a broader level of communities and geographic areas. To best understand prevention requires analyzing influences on multiple intersecting levels and to establish temporal causality of smoking initiation requires using longitudinal population-based cohort studies (Wakefield and Forster, 2005).

Previous studies have examined individual, proximal, and environmental levels of predictors of adolescent smoking initiation. The proximal social environment includes friends' smoking which has been identified as a predictor of both adolescent smoking initiation (Conrad et al., 1992, Flay et al., 1994) and susceptibility to smoking (i.e. likelihood of accepting a cigarette if offered in the near future). (Bauman et al., 2001, Gritz et al., 2003, Miller et al., 2006) Other proximal risk factors to smoking initiation include living in a single-parent home, exposure to pro-tobacco messages, and living with at least one household smoker (Gritz et al., 2003). Adolescents often smoke their first cigarette in the presence of other peers (Johnson and Hoffmann, 2000) and often access cigarettes from social source (Forster et al., 2003). They smoke as a form of stress reduction, to relax, and as an expression of independence (Nichter et al., 1997).

Neighborhood-level social variables have statistically significant effects on health behaviors (Pickett and Pearl, 2001). Specifically, an inverse association was found between neighborhood-level SES and smoking prevalence, after controlling for individual-level SES (Kleinschmidt et al., 1995); with a relative risk of smoking of 1.2–1.7 in lower SES neighborhoods (Pickett and Pearl, 2001). Area-level analyses of SES (measured as employment, education, housing values, and income) have also yielded inverse associations with mean smoking prevalence in youth (Bernat et al., 2009).

Longitudinal datasets have been used to examine variation in influences on adolescent smoking initiation with particular attention to ethnicity (Griesler and Kandel, 1998, Griesler et al., 2002); parental influence (Flay et al., 1994); and peer influence (Alexander et al., 2001, Flay et al., 1994, Urberg et al., 1997). While prior research establishes the importance of area-level analyses few have used a multi-level approach to analyze adolescent smoking (Kandel et al., 2004, Klein et al., 2009, Siegel et al., 2008). No studies reported the influence of socio-economic variables and prevalence of older adolescent smoking at the area level on adolescent smoking initiation, which is the express purpose of our paper.