Do socioeconomic differences in tobacco use exist also in developing countries? A study of Ghanaian adolescents

A cross-sectional survey was conducted from June to August 2008 on health behaviours and lifestyles of school-aged adolescents in three administrative regions in Ghana. Thirty schools were randomly sampled, ten per region, from Eastern (total number of schools in the region = 2924), Greater Accra (total number of schools in the region = 1825) and Volta Regions (total number of schools in the region = 2184). The Ghana Education Service's School Health Programme register of schools in the country was the source of the sampling frame. The sampling was done as follows: First, ten schools were randomly selected so that they comprised of four public Junior High Schools (total number in the three regions = 5325), two private Junior High Schools (total number in the three regions = 1395), three public Senior High Schools (total number in the three regions = 171) and one private Senior High School (total number in the three regions = 47) in each region in order to reflect the school types in Ghana. Second, in each school, all students whose names were found in the class attendance register of the randomly selected classes were eligible to participate in the survey. The eight page questionnaire was anonymous and self-administered. It was designed to exclude any information that will reveal the identities of the participants. One trained supervisor was assigned to each classroom during the questionnaire administration to address pupils' concerns when necessary. The survey commenced simultaneously in all the participating classes in a given school. Participants were asked to drop their questionnaires in an envelope placed in front of the class on completion. The study protocol was approved by the ethical committee of the Ghana Health Service Research Unit in Accra, Ghana.

The characteristics of the respondents are presented in Table 1. Out of the 1566 respondents who completed the questionnaire, only 13-18-year-old students were included in this study (N = 1165). They comprised of 41.5% (483) boys and 55.3% (644) girls. The mean ages for boys and girls were 15.8 years and 15.9 years, respectively. The response rate was 89.7% (the sample was based on academic year's register of pupils). Only one pupil denied answering. A convenient sample of 127 non-students in the same age group showed similar pattern of responses for most of the key indicators we measured in our survey. Among this group 3.6% and 7.3% were tawa users and smokers, respectively.

Pearson's Chi-square tests were used to test the associations between gender and each of the studied variables. Logistic regression analysis was used to model the associations between the socioeconomic indicators and tobacco use. The strength of the associations was expressed by odds ratios (OR) and 95% confidence intervals (CI). First, bivariate models (Model1) were fitted including each of the socioeconomic measures one at a time, controlling for age and gender. Second, multivariate logistic regression models were used to test whether individual SES measures were independently predictive of tobacco use. Model 2 included age, gender and all the statistically significant socioeconomic indicators and then Model 3 comprising of the indicators in Model 2 plus parental smoking. For the social mobility analyses models 2 was adjusted for family structure and model 3 was adjusted for family structure and parental smoking. In all analyses, those with the highest socioeconomic advantages were used as the reference categories.

A lower level of material affluence was associated with the likelihood of tawa use but the association was not statistically significant with smoking. Adolescents who lived in family types other than the nuclear family were more likely to smoke or use tawa compared to those who lived in the nuclear family. Lower paternal education predicted both smoking and tawa use (Table 2, Model 1). Adolescents whose fathers had primary education were more likely to use tawa compared to those whose fathers had tertiary education, albeit at borderline statistical significance. Adolescents who had illiterate fathers were more likely to smoke than those whose fathers had tertiary education. There were no statistically significant associations between tobacco use and mother's education, and father's or mother's occupation. In multivariate analysis, material affluence independently predicted tawa use when the effects of the other statistically significant familial socioeconomic measures (Table 2, Model 2) and parental smoking were controlled for (Table 2, Model 3). Similarly, family structure independently predicted smoking and tawa use.

Tawa use and smoking were related to both downward social mobility and stable low SES whether mobility was measured by material affluence scale or father's education compared to being stable in the high SES (Table 4). In a multivariate analysis, tawa use and smoking were independently related to downward social mobility and particularly being stable in low SES by both indicators of social mobility, after adjusting for family structure (Table 4, Models 2) and parental smoking (Table 4, Models 3). The only exception was that relationship between tawa use and downward mobility disappeared after controlling for parental smoking. Upwardly mobile adolescents did not differ from those stable in the high SES by smoking or tawa use.

We used a representative sample of schools both in urban and remote rural areas in three regions which are representative of the entire country, the first study of its kind in Ghana. Some of the questions we used have also been used in other studies, for example, the Global Youth Tobacco Survey (GYTS) and the Global School-based student Health Survey (GSHS) which have been conducted in many African countries. Moreover, to the best of our knowledge our study is the first of its kind which has investigated the traditional smokeless tobacco (tawa) in Ghana.

Self-report is the only way to conduct large surveys but it could lead to recall bias or intentional miss-reporting which could affect the accuracy of the reports. However, this should not affect the relationships between SES and tobacco use among adolescents. Similar methods have been used in most previous studies [4, 5]. The study was cross-sectional therefore the cause and effect relationship cannot be emphasised as an etiological conclusion, nonetheless it can be argued that at the adolescent age socioeconomic status is likely to precede tobacco use and not the reverse. During data collection an investigator was present in the classroom to address the concern of the pupils when necessary. Although we do not perceive that this might have affected the adolescents' responses, if it did, it would be more likely to have resulted in the under estimation of both the tobacco use prevalence and the socioeconomic status rather than over estimation. Our sample of students for the study was drawn from a sample of schools. The clustering of students may slightly change the standard error of our estimates, although unlikely to change neither the overall results nor the conclusion reached in this study. Due to scarce resources, only adolescents in schools have been included in this study. On the other hand, a similar pattern of responses for most of the key indicators in this study was found among a convenient sample of non-students in the same age group. Moreover, the school enrolment rate in Ghana for the age group of our respondents is high, 78.8% for Junior High Schools (Ministry of Education, Science and Sport, Ghana, 2008).