Introduction

Smoking is one of the leading preventable causes of early death, disease and disability [1–3]. In addition, smoking is a major contributor to socio-economic inequalities in health since smoking-attributed mortality accounts for more than a half of the difference in mortality rates between social strata in middle age [4]. Smoking is typically taken up in adolescence [5], and early nicotine exposure directly increases risks of later nicotine dependence [6]. Given that the major risk period for smoking initiation is mostly over by the age of twenty [5, 7], understanding how environmental and individual risk factors contribute to smoking initiation in adolescence is a crucial step in designing appropriate intervention and prevention strategies.

Among environmental factors, socioeconomic status (SES), a combination of factors including income, education and occupation, has been shown to impact health due to lifestyle choices [8]. Furthermore, SES has been associated with adult smoking. For example, smoking is more common among people of lower education and lower SES levels in European countries [9, 10]. Some studies found an association between parental SES and smoking initiation of adolescents [11–13]. A recent cross-sectional study of 35 countries showed socioeconomic inequalities in adolescent smoking behaviour: boys and girls from poorer families were more likely to be smokers, and this association was mediated by an unequal distribution of family factors such as family structure and relationships with parents [12]. Crucially, while smoking rates among adults and adolescents are in decline in many western countries, including Spain [14, 15], these changes do not occur equally across different socioeconomic levels [10, 16, 17]. Adults with at least some college education had a significantly greater decline in smoking prevalence than those whose highest level of education is high school or less [18].

Among individual factors, personality traits described by the Five Factor Model (FFM)[19, 20], have been associated with a variety of health behaviours [21–23], including smoking [24, 25]. A study conducted on an elderly population showed that current smokers had higher levels of neuroticism, and lower levels of agreeableness and conscientiousness than former smokers, and those who never smoked had lowest levels of neuroticism and highest levels of agreeableness and conscientiousness of all groups [26]. Higher openness and higher neuroticism have been associated with lifetime-smoking, and higher conscientiousness appears to protect against smoking progression and persistence in adults [27]. A study conducted in an undergraduate student population showed that higher extraversion, higher neuroticism, lower agreeableness and lower conscientiousness were all associated with higher current smoking behaviour, and that higher neuroticism was more strongly associated with smoking in a Mexican, than in the Mexican-American sample [28]. Furthermore, Conner et al. [24] found that adolescent non-smokers had higher levels of conscientiousness than adolescent smokers.

Finally, Chapman et al. [25] evaluated, in an adult sample, whether the relationships between the level of education, a common marker of SES, and smoking behaviour would be confounded or modified by personality traits. While they found that never smokers had lower levels of openness and higher levels of conscientiousness, and that those who quit smoking had higher levels of neuroticism, they found that the effects of education and personality were additive, and therefore these two sets of risk factors should be considered independent.

To the best of our knowledge, the role of SES on the associations between personality traits and smoking in adolescents has not been explored to date. Building on the findings that parental SES affects smoking initiation among adolescents [12] and the studies that show that personality traits are associated with adolescent smoking status [24], here we explore whether personality traits and parental education are associated with smoking initiation in a cohort of Spanish secondary school students. Furthermore, we test whether the associations between personality traits and smoking vary across levels of parental education.

Methods

Results

Descriptive statistics of the sample are presented in Table 1. Of the 842 adolescents included in the study, 62 (12.9%) self-reported being a smoker. Further 182 (21.6%) were classified as triers, and the remaining 598 (71%) as never smokers. Both triers and regular smokers were older than never smokers. Mean levels of extraversion and neuroticism were significantly higher among regular smokers and triers, and the mean level of conscientiousness was significantly lower among triers and regular smokers than in never smokers. There were significantly fewer smokers among adolescents whose parents had a university degree.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Sample characteristics by smoking status.

https://doi.org/10.1371/journal.pone.0174211.t001

To examine the independent associations of adolescent personality and parental education to smoking behaviour among adolescents, we fitted a set of three stage multinomial logistic regression models (Table 2). Adjusting for the effects of age and gender, lower levels of conscientiousness and higher levels of extraversion and agreeableness were associated with higher risk of being classified as both ever smoker and regular smoker (Model 1). The associations were stronger for the regular smoker category. Furthermore, higher parental education was associated with regular smoking (Model 2). When personality and parental education were taken together, the associations remained significant and similar in magnitude to those observed before (Model 3)

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Odd Ratios (OR) and 95% Confidence Intervals (95% CI) for the models containing personality traits and parental education, predicting adolescent smoking status.

https://doi.org/10.1371/journal.pone.0174211.t002

To test whether parental education moderated the associations between students’ personality traits and smoking status, we fit an additional multinomial logistic model. In this model (Table 3) none of the interactions between personality and parental education was statistically significant.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Odd Ratios (95% Confidence Intervals) for the model including interactions between adolescent personality and parental education predicting adolescent smoking status.

https://doi.org/10.1371/journal.pone.0174211.t003

Discussion

Our results indicate that parental education and adolescents’ personality traits are independently associated with adolescent smoking behaviour. Adolescents with higher levels of extraversion and neuroticism and lower level of conscientiousness were more likely to smoke. Furthermore, adolescents whose parents had lower educational level were more likely to smoke regularly. Finally, we found no evidence that parental education moderates the associations between adolescent’s personality and smoking behaviour.

Our results regarding extraversion and neuroticism corroborate those reported in previous studies conducted on adolescent samples [34, 35]. Of the five personality traits, neuroticism has been most consistently associated with smoking. One possible explanation is that adolescents with higher levels of neuroticism use smoking to reduce unpleasant emotions [35]. On the other hand, it is possible that individuals higher in extraversion show higher probability to smoke because both their high sociability and their increased dopaminergic activity that would be associated with increased likelihood of smoking [36]. Lastly, with respect to personality traits, we found that higher levels of conscientiousness were associated with lower likelihood of smoking in adolescence. This is not surprising, given that conscientiousness is most commonly associated with better health-related behavioural patterns in adult life, including lower prevalence of smoking [36–38].

Furthermore, our results are similar to those reported in previous studies that found an association between family SES and smoking among adolescents [12, 17]. We found an association between an indicator of family SES, highest parental educational level, and regular smoking, but we did not find an association between parental education and triers. It is possible that individual factors, such as personality, explain whether or not adolescents initiate smoking, while environmental factors, such as SES, contribute to explain why smoking eventually becomes a persistent habit. One proposed mechanism of the associations between parental affluence and adolescent health behaviours is through modelling processes [17]. To test this possibility, we utilised information on parental smoking collected two years prior to the data presented here. There were significantly more parent smokers among those with lower education (primary or less) than those of medium and higher education (36% vs 25% of smokers for mothers, and 37% vs 30% for fathers, data available upon request). Interestingly, parental education was also related to adolescent smoking when controlling for parental smoking (OR = 0.7; 95%CI 0.5–0.9, data available upon request) that could indicate also some effect of parental education independent of modelling behaviour.

Finally, parental education was associated with adolescent smoking independently of adolescent personality. This finding replicates and expands those reported by Chapman et al. [25] who utilised measures of own education and smoking behaviour in adults. Therefore, the two sets of predictors may act independently in adolescent as well as in adult populations. Future studies should also examine whether similar patterns apply to other substances [39].

This study had several limitations. First, the current analysis was limited to cross-sectional data. Although personality traits are relatively stable across the life span [40], longitudinal design is needed to better address causal direction of the effect. Future studies of smoking behaviour in adolescence should include a measure of personality traits prior to smoking initiation. Second, we include a measure of parental education as the only SES indicator. It is possible that a more comprehensive measure of parental SES would yield different results, especially with respect to interactions with personality traits [41]. Finally, we relied on self-reported measures of smoking behaviour. However, while direct methods such as coximetry exist, they suffer from reliability issues due to uneven inhalation and false positives due to exposure to environmental smoke. Coximetry was also not shown to accurately assess smoking in adolescents [42].

Our study contributes to the evidence that personality factors and parental education are important factors associated with smoking behaviour during adolescence. Furthermore, these factors should be addressed independently. Effective interventions tailored for individual personality traits and SES backgrounds could be useful to reduce future social disparities in morbidity and mortality.

Acknowledgments

The authors would like to thank the participants, schools, and all those involved in the “ITACA” study. This Project was funded by a Health Research Funds of the “Instituto de Salud Carlos III” from Spanish Government (P10/00517) co-financed with European Union ERDF funds. The work is also being supported by the Health Promotion and Preventive Activities-Primary Health Care Network, which is sustained by the Ministry of Health of Spain (ISCIII-RETCI G03/170 and RD06/0018).