Study area and population
In October–November 2011, we conducted a population-based cross-sectional study in Jhaukhel and Duwakot Villages in Nepal. Located in the mid-hills of Bhaktapur District, 13 kilometers east of Kathmandu, Jhaukhel and Duwakot represent the prototypical urbanizing villages that surround major urban centers in Nepal, where young people have easy access to tobacco products and are influenced by advertising. Jhaukhel and Duwakot had a total population of 13,669, of which 15% were smokers [27]. Of these, 909 adolescents between 14 and 16 years of age and the male to female ratio was 1.06:1. Among the 909 adolescents, 491 lived in Duwakot [27].
Sample size
The estimated sample size was based on an unknown prevalence of smoking (50% assumed for conservative sample size estimates), with a required maximum error totaling ± 5% units and a 95% confidence level. Our approach yielded a sample size of 384 adolescents and allowed 20% inflation to account for non-response and incomplete questionnaires. Finally, we decided on a sample size totaling 500 adolescents [28].
Among 500 potential participants, 498 responded, one refused, and one had a hearing impairment. Among the 498 respondents, 485 were nonsmokers and 13 were smokers. Among the 485 nonsmokers, 29.3% were excluded from analysis because they did not answer the questions related to susceptibility to smoking. We performed our final analysis on 352 respondents.
Study design and sampling method
This was a population-based cross-sectional study. We adopted a proportionate stratified sampling technique to select adolescents from each village. The sampling frame, which included 909 adolescents, was obtained from the baseline survey 2010 [27]. Among the 909 adolescents, 45.9% lived in Jhaukhel and 54.01% lived in Duwakot. Using these sampling fractions, we selected 230 adolescents from Jhaukhel and 270 from Duwakot, yielding 500 randomly selected adolescents. In Step 1, we obtained a sampling fraction that represented 49.8% and 53.2% of the male adolescents from Jhaukhel and Duwakot, respectively (i.e., 114 males from Jhaukhel and 150 males from Duwakot). In Step 2, we further classified the sex of the adolescents into three age groups (14-, 15-, and 16-year-olds) for each village. Among the 114 male respondents in Jhaukhel, 32.2%, 36.1%, and 31.7% belonged to the 14-, 15-, and 16-year-old age groups, respectively. Among 116 female respondents, 42.9%, 28.6% and 28.6% belonged to the same age groups, respectively. In Duwakot, 34.5%, 36.4%, and 29.1% of 150 male respondents belonged to the 14-, 15-, and 16-year-old age groups, respectively, and 30.4%, 38.7% and 30.9% of the 120 female respondents belonged to the same age groups, respectively. Finally, we used systematic sampling from each age group to select the respondents.
Tools
Our semi-structured questionnaire contained seven major sections: (i) socio demographic and individual information; (ii) smoking activities of family members, relatives, teachers, and friends; (iii) exposure to media and advertising related to tobacco and noncommunicable disease education; (iv) perception of risks and benefits of smoking; (v) smoking behavior of adolescents; (vi) smoking cessation; and (vii) health status. We adapted our questionnaire from the Global Youth Tobacco Survey 2007, the Teen Smoking Question (TSQ), and perceived risks and benefits items from Song et al. and Halpern-Felsher et al. [22, 23, 29, 30]. We modified the original questionnaire to reflect the cultural context of Nepal, and a local public health graduate translated the questionnaire into Nepalese. We made necessary modifications after pretesting the questionnaire in Chagunarayan village, Bhaktapur District. Our study specifically investigated demographic characteristics; perceived physical, social, addiction risks and perceived benefits of smoking; smoking behavior; and susceptibility to smoking.
Definition of variables
Nonsmoker
A nonsmoker adolescent is one who has never smoked, even a puff.
Nonsmoker susceptibility to smoking
We determined nonsmoker susceptibility to smoking by asking three questions, using the algorithm of Pierce et al. [14]:
-
Will you try a cigarette (taking even just one puff) sometime in the next 6 months? Response choices included (i) definitely will not, (ii) may not, (iii) may be will; and (iv) definitely will.
-
If one of your best friends offers you a cigarette, do you smoke? Response choices included (i) never, (ii) sometimes, and (iii) always.
-
Do you think you will smoke cigarettes 5 years from now? Response choices included (i) not at all, (ii) slightly likely, (iii) moderately likely, (iv) very likely, and (v) most likely.
Respondents who answered “definitely will not, never, or not at all” to all three questions were considered not susceptible to smoking (code = 0). All other respondents were considered susceptible to smoking (code = 1).
Perceptions of risks and benefits of smoking
We asked respondents to estimate their likelihood of having various perceptions of the risks and benefits of smoking in a hypothetical scenario: “Imagine that you just began smoking. Sometimes you smoke alone and sometimes you smoke with friends. If you smoke 2–3 cigarettes each day, what is the chance of getting physical and social risks and benefits of smoking?” Respondents estimated the chance (0%–100%) of experiencing seven physical risks (lung cancer, heart disease, facial wrinkles, bad colds, bad cough, bad breath, trouble breathing); two social risks (getting into trouble, smelling like an ashtray); and four perceived benefits (looking cool, feeling relaxed, becoming popular, feeling grown-up) [22, 23]. Next, we treated two components (i.e., you can quit smoking cigarettes if you want to, and you will become addicted to cigarettes) as an addiction risk and asked respondents to estimate likelihood (0%–100%), as mentioned above [22].
Training
Eight local enumerators and two field supervisors attended a 3-daytraining session prior to data collection. The training comprised objectives, ethical issues, ways to build rapport and collect data, questionnaire content, and health hazards of cigarette smoking. At the end of the training session, we pretested the questionnaire in the field and incorporated feedback into the final study questionnaire.
Data collection
First, enumerators identified households containing adolescents and met with parents to explain the study objectives. After obtaining verbal consent from the parent(s), the enumerators contacted the adolescents and explained the purpose of the study; they also assured the confidentiality of collected information. All adolescents who agreed to participate in the study were interviewed during a 60-minute, face-to-face interview conducted in a separate setting at a time convenient for each respondent.
Monitoring, supervision and quality control
Field supervisors, a field coordinator, and a PhD student regularly and closely supervised the enumerators. To ensure maximum response rates and reliable data collection, the field supervisors were responsible for spot-checking and discussing field site issues and problems with the field coordinator and the PhD student. In addition, the field coordinator and PhD student randomly cross-checked the completed forms, both in the field and in the office. Erroneous forms were returned to the field for renewed data collection.
Data management and analysis
Collected data were coded and entered into an EpiData 3.1 program and analyzed using SPSS 17.0 and STATA SE 10 software. Descriptive statistics (i.e., percentage, mean, quartiles, and standard deviation) were computed to describe both categorical and numerical variables (e.g. respondent characteristics and chance estimates for risks and benefits of smoking). Next, we used a chi-square test to compare proportion differences between different categories. Univariate and multiple logistic regressions established the relationship between susceptibility to smoking as a dependent variable and perception of risks and benefits of smoking as independent variables. Before fitting the model, we used principal component extraction with varimax rotation to confirm how well the 13 risks and benefits items loaded on their respective categories. Analysis reduced these 13 items into four meaningful categories, based on the factor scores (factor loading less than 0.04 is not reported). Categories I and II contain items related to perceived likelihood of physical risks, Category III relates to perceived likelihood of the social risks and Category IV relates to perceived likelihood of social benefits [22, 23]. Further, physical risks are categorized as physical risk I and physical risk II as bad cold, the short-term risk item, is listed in the first component where all other 3 items(Lung cancer, heart diseases and facial wrinkles) are related with long term risks [23]. Perceived physical risk I includes items describing physical problems caused by habitual smoking (long-term risks and bad cold). Perceived physical risk II comprises items describing short-term risks (bad cough, bad breath, and trouble breathing) of smoking [23]. Perceived social risks included getting into trouble and smelling like an ashtray [22, 23]. Perceived social benefits included looking cool, feeling relaxed, becoming popular, and feeling grown-up [23]. Next, we computed the composite scores for the four categories and also for addiction risk. To aid data interpretation and discussion, we coded mean scores into quartiles, where 0 = first quartile and 3 = fourth quartile [23]. Using univariate analysis, we computed the unadjusted odds ratio (OR) according to quartile score for each perception item with susceptibility to smoking. For multiple logistic analysis, we entered all five perception components, including addiction risks simultaneously into the model and computed the adjusted odds ratio (AOR) [23]. We set the significance level at 5% (alpha = 0.05) and excluded missing cases and “do not know” answers.
Ethical issues
We first sought verbal informed consent from the parents of the participating adolescents as they were less than 18 years of age. Then, we sought verbal informed consent from the participants as well. Further, we informed all respondents that their participation was voluntary and told them they were free to terminate the interview if they did not want to continue or to opt for the next question if they were unwilling to answer a particular question. We also assured all respondents about the confidentiality of collected information. At the end of each interview, we gave the participant a Nepalese-language leaflet that described the harmful effects of tobacco use. Before initiating the study, we discussed the study objectives with local authorities and leaders and obtained their permission. The Nepal Health Research Council and the Ethical Committee of Kathmandu Medical College approved this study.