Family and peer effect on young and adolescent smoking in Bangladesh

Background: Cigarette smoking among youth has become a major public health problem in many developing countries like Bangladesh. Many smokers start smoking in their teens; then they become addicted and deepen the dependency during early adulthood, maintaining their smoking behaviour. Current literature from Bangladesh has not examined how peer effect, family influence and smoking-related knowledge impact smoking behaviour of the adolescent and young adults from the urban areas of the country. This study aims to fill this gap in the literature by investigating the smoking patterns of urban male adolescents and young adults in Bangladesh. The study aims to examine the family and peer influence on the behaviour of smoking of this population group. The study also explores how knowledge on harmful effects of smoking affects the participation decision of smoking by young adults and adolescents. Methods: This study is an educational institution-based cross-sectional study within the framework of the Global Youth Tobacco Survey (GYTS). Primary data has been collected from a sample of 995 urban male students aged between 10-24 years living in the seven divisional headquarters of Bangladesh. We use logistic regression and count data regression models to examine how smoking is associated with different factors. Results: Our results suggest that a person is 45% more prone to smoke a cigarette if any of the family members is a smoker. The likelihood of smoking is about six times higher if he is offered a cigarette by his friend, whereas the likelihood of being a smoker is 15-fold if he has a smoker friend, as the findings reveal. However, institutional and family awareness are not significantly related to smoking. Conclusion: Strong evidence of peer and family effect on smoking initiation indicates the need for effective smoking-prevention intervention at the national level, specifically targeted at family and educational institutions. The findings are also relevant for other

countries which share the similar characteristics of Bangladesh in terms of adolescent smoking determinants.

Background
Individuals become more familiar with their social networks during adolescence and young adulthood period. This is a period of adopting and keeping persistent patterns of behaviours. This transition time makes them vulnerable to smoking initiation and other addictions. Adolescence is a period of smoking onset for most adult smokers, and multiple social, psychological and biological factors influence this process [1,2]. Family members and peers in the adolescents' social networks shape their socio-economic behaviour, including their smoking behaviour. The hypothesis that smoking onset is influenced by peers' and family members' smoking is well-documented in existing literature in developed countries [3][4][5][6][7][8][9][10].
On the other hand, in developing countries, the susceptibility of smoking onset and efficacy of prevention programmes differ in accord with the variation of ethnic and racial traits, socio-cultural context and socioeconomic status; accordingly, research is required in these countries [11,12]. In addition, smoking prevention and cessation programmes are not well-developed in developing countries compared to developed countries, where smoking prevention programmes are established in a comprehensive manner as they take into account both youngsters and parents within an integrated framework [13].
Bangladesh has a high smoking prevalence rate among developing countries, with 23 per cent of adult smokers [14]. The public health burden of Bangladesh is acute because 25 per cent of deaths in men is attributable to smoking [15,16]. To address the issue of tobacco control, Bangladesh was the first country to the sign the World Health Organization (WHO) Framework Convention on Tobacco Control (FCTC) in 2003 [17].
Bangladesh enacted tobacco control law in 2005 and amended it in 2013 to make it more compliant with WHO FCTC [18]. Smoke-free places, health warning messages, tobaccoadvertising ban in media, and increased tax rates on tobacco products are the main tobacco control strategies implemented in Bangladesh [19]. However, cessation service and related programmes are negligible in Bangladesh [20].
Understanding the role of family and peer effect on smoking is vital for developing smoking prevention and cessation interventions in Bangladesh. There are few studies from Bangladesh which examined the correlates of smoking behaviour among young adults and adolescents [21][22][23][24]. However, these studies did not consider peer and family members' smoking behaviour to understand the smoking pattern of this population group. This study aims to fill this gap in the literature by examining family and peer influence on the behaviour of smoking among urban young adults and adolescent males using data from an educational-institution based cross-sectional survey carried out in seven major cities of Bangladesh. We investigated not only the association of family and peer influence with the probability of smoking but also with the intensity of smoking.

Study design and study population
The research team conducted an educational-institution based, cross-sectional study within the framework of the Global Youth Tobacco Survey (GYTS). The educational institutions, considered in this study, were in the headquarters of each of the then-seven administrative divisions of Bangladesh. These headquarters were the cities of Dhaka, Chittagong (now-Chattagram), Rajshahi, Rangpur, Sylhet, Khulna and Barishal. Smoking is primarily a male-oriented activity in Bangladesh. The proportion of females aged 15 years and above who described themselves as a current smoker ranged between 1.1 per cent [14] and 1.5 per cent [25]. Therefore, the study focused exclusively on males. The target population, therefore, comprised urban male students aged 10 to 24 years of age who were studying at the relevant institutions during the period of data collection in 2014.

Sampling design and sample size
A two-staged stratified clustered sampling with probability proportional to size (PPS) sampling method was used in this study. From the institution list of the Bangladesh Bureau of Educational Information and Statistics (BANBEIS), two high schools, two colleges and one university were drawn randomly from each of the seven selected cities. This

Data collection
A structured questionnaire was developed and used to collect primary data from respondents of 35 educational institutions. The study not only adopted core components of the GYTS questionnaire but also included additional questions to allow the analysis of the influence of family and peer behaviour on the smoking outcomes of the target group.
Data were collected by trained enumerators, and the participants were assured that the data would be confidential. Consent from any of the parents or guardian was obtained where the participant was 16 years old or less. In this case the interview was conducted in the absence of any of the parents or guardian to reduce the social desirability bias.

Dependent and independent variables
During the survey, the respondents were asked whether they smoked cigarettes in the last 30 days preceding the survey. Using this information, a binary dependent variable was constructed for analysis. Independent variables included socio-demographic characteristics such as age, education, income, financial support from family, parents' education, family history of smoking, conveying of information on the harmful effects of smoking by the educational institution, smoking status of friends and explicit activities of peers to encourage smoking by the target individuals. The age variable was disaggregated into three categories: 10-15 years, 16-18 years and 19-24 years. Education was also disaggregated into three categories: school (up to and including ten standard years of education), college (11-12 standard years) and university (more than 12 standard years of education). Economic variables included income (full-time, part-time or casual work) and financial support from family. These variables were dichotomous variables having the value of 0 if the answer was "no" and 1 if it was "yes". Parents' education was divided into five categories: less than primary, secondary, higher-secondary, and tertiary. The location variable comprised the seven divisional cities, as the educational institutions attended by the students in the study were in these cities. The variable for family history of smoking, which was assigned two different value labels for "no" and "yes", considered whether the participants' father, uncle, brother, cousin or grandfather was a smoker. The influence of peers was measured by two variables: whether any of the friends was a smoker, and whether the participant was encouraged to commence smoking by a friend (by being offered a cigarette). Information provided by family or by the education institution about the harmful effects of smoking were constructed as binary variables.
Living location variable was classified into four categories-living with parents, living in the dormitory, living with relatives and living in other places.

Statistical analysis
This study used both univariate and multivariate methods for statistical analysis. The distribution of participants between smokers and non-smokers based on different characteristics was determined using the univariate technique. Multivariate regression model was used to estimate the adjusted odds-ratio(OR) to study the association between the probability smoking and independent variables. Furthermore, count data regression model was employed to estimate the incidence rate ratio (IRR) of the intensity of smoking.  Table 1 presents the descriptive statistics of smoking prevalence among the study population. The number of smokers and non-smokers was 221 and 774, respectively, representing a smoking prevalence rate of 22.2 per cent. The prevalence of smoking differed by age group and ranged from 7.8 per cent for the 10-15 years old group to 43.3 per cent for the 19-24 years old group. In addition, factors such as level of education, residence, financial support from family, income, mother's education, smoking among family members or peers/friends, education institutional awareness, and smoking encouragement by friends were found to be different between the smoking and nonsmoking groups. between being a smoker and having been offered a cigarette by a friend was also positive.

Results
A person who, when still a non-smoker, was offered a cigarette by a friend, had 6.4 times more chance of becoming a regular smoker than those who did not receive such an offer (OR = 6.41, 95% CI: 3.23, 12.70). Observations 995 995 Notes: Level of significance *** p < 0.01, ** p < 0.05, * p < 0.1 Table 3 presents the results from the analysis of negative binomial regression modelling.
The data contained 77.9 percent zero observations with all other values being positive, making count data regression a suitable choice for the modelling [26]. The results suggest that the important factors associated with smoking intensity were age, presence of a smoker in the family, association with friends who smoke, being offered a cigarette by a friend and living with others in a dormitory. However, education level and family support were not associated with smoking. The incidence rate ratio (IRR) of the variable related to family member's smoking status was 1.45 (95% CI: 1.02, 2.07) which indicates that a person was 45 per cent more likely to be a smoker if one or more of their family members was a smoker. Again, a person was six times (IRR = 6.01, 95% CI: 3.48, 10.38) more likely to smoke if the respondent was offered a cigarette by a friend, whereas the likelihood of   [24] where students in a higher secondary college in the Moulvibazar district of Bangladesh were the study participants. In another study of college students studying in private and public colleges in three towns of Karachi, Pakistan, [27,28] it was 24 per cent. Kamal [23] surveyed 474 male students at the Islamic University, Kushtia, Bangladesh, and reported that about 36 per cent were current smokers. It should be noted that, as in our case, many studies only included male students in the target population since smoking prevalence is not high among female students in many low-middle income countries.
Several studies have linked the smoking behaviour of youth with the smoking behaviour of their parents. For example, Gilman and co-authors [5] studied the influence of parental smoking on adolescent smoking initiation. The results indicated that adolescents whose parents were regular smokers had a higher likelihood of being initiated into smoking (OR = 2.81) than adolescents whose parents never smoked. It is well established in the literature that paternal and family members' smoking status influences youth smoking behaviour [29,30]. Our study also shares a similar view.
Students living with parents or siblings smoking were more likely to be initiated into smoking at an early age than are those who live in a non-smoking home environment.
Parents who hold strong anti-smoking views may directly influence their adolescent children's intention to smoke and to socialise with smokers [29]. Karimy et al. [31] used demographic variables, psychological factors, and components of the theory of planned behaviour to identify factors contributing to cigarette smoking among male adolescents in Iran. They found that parents' smoking had a significant impact on the smoking behaviour of adolescents. Our research showed that the average number of smokers in a family was higher for respondents who were current smokers. This study further revealed that the probability of smoking among young people was positively related to their number of smoker friends. It also found that an offer to smoke by friends increased the chance of becoming a smoker.
This study found that most of the current smokers reported being offered a cigarette as the trigger for their initiation into smoking. Peer smoking has emerged as one of the most significant predictors of adolescent smoking in many studies. For example, Christophi et al. [32] showed that smoking by peers was the main predictor of smoking among adolescent students attending middle and high schools in Cyprus.
We found that parental education did not significantly reduce smoking among young adolescents and adults. This is similar to the findings in Kamal et al. [23], where the authors investigated socio-psychological correlates of smoking among male students in selected Bangladeshi universities. The findings of a study conducted in Kerala, India also suggested that the father's educational attainment had no bearing on adolescent boys' smoking status [33].
This study found that obtaining knowledge about the harmful effects of smoking from the education provider does not significantly reduce the probability of smoking. The finding differs from those made in some other studies [34]. One reason for this could be that the level of knowledge sharing by the institutions is not intensive enough to be associated with a change in smoking behaviour. Another reason could be that the benefits of smoking are certain as well as immediate, while the costs are perceived to be uncertain and, if they are realised, occur a long way into the future. The place of residence was not found to be a strong factor in our study, which is similar to the findings of a Nepalese study that investigated correlates of use among junior college students in twin cities of western Nepal [34]. In this study, financial support from family was not associated with either the prevalence or intensity of smoking. This is consistent with the finding from Mall Leinsalu's study in Hungary in terms of self-generated income [35].
Our study has a few limitations. One of the limitations of the study is that biometric screening was not done to confirm the smoking status of the individuals. Therefore, there might exist some under-reporting or over-reporting, which is a common limitation of selfreported data. The present study tried to minimise this effect by employing a well-trained interviewer who was likely to convince the respondents that confidentiality would be strictly maintained. The modelling approach adopted in this study could lend itself to different kinds of interpretation. However, the coefficients of interest are not sensitive to the choice of variables, as depicted in supplementary Table 1.
This study has a number of policy implications for tobacco control in Bangladesh. First, most of the tobacco control law focuses on tobacco advertising and promotion, health warning messages and smoke-free public places. Bangladesh government is also exploring tax-raising policies to control tobacco use. However, little to no attention has been paid to prevention campaigns that involve parents or campaigns aimed at teaching young and adolescents peer refusal skill. Second, Bangladesh also lacks cessation interventions targeting adult smokers specially smoker parents. Quitting helpline and tobacco treatment interventions are also unavailable in Bangladesh. Third, studies have found that family engagement and family cohesion on anti-smoking behaviour reduce smoking among adolescents and young adults [36,37] and in line with our findings, we suggest that Bangladesh government could design interventions aimed at parents and educational institutions. While not covered in this study, smoke-free air law initiatives have been successful in other countries [38] and should be initiated as a potentially successful intervention in Bangladesh, for example, introducing smoke-free campuses in educational institutions. Future research on designing family and peer intervention would benefit Bangladesh on tobacco control.

Conclusion
The current study expanded the understanding of the Bangladeshi young and adolescents' smoking onset in the context of smoking exposure from family members and peers. The research findings validated that both family members and peers are significant factors in young and adolescents' cigarette smoking status. We found that peer smoking status was substantially associated with a higher risk than family members smoking behaviour for the Bangladeshi young and adolescents' smoking onset. In addition, the risk of smoking initiation is also higher if the respondents' friends offer to smoke. However, institutional and family awareness did not significantly reduce the risk of smoking initiation, which suggests the need for improving the quality and breadth of smoking prevention