The data in this study come from the Global School-based Health Survey (GSHS), which was conducted for the first time in the Seychelles in 2007 [31
]. The GSHS, a school-based survey developed by the World Health Organization, the Centers for Disease Control and Prevention (Atlanta) and other international agencies, aims to provide a common methodology for collecting data on a broad set of risk behaviors and psychosocial characteristics among students worldwide (http://www.cdc.gov/gshs/questionnaire/index.htm
). Consistent with the GSHS methodology, a two-stage cluster sample design was used to produce a representative sample of all students in grades S1-S4 in all public and private schools in Seychelles (school is compulsory up to the S4 level in the Seychelles, which correspond to the 10th
year of school after crèche). The first-stage sampling frame consisted of all schools containing the grades S1, S2, S3, and S4. Schools were selected with probability proportional to school enrolment size. In Seychelles, all the 13 schools containing S1-S4 classes in the Seychelles were selected in the study. The second-stage sampling frame consisted of an equal-probability sampling (with a random start) of all S1-S4 classes in the selected (13) schools: 64 classrooms were selected from a total of 274. All students in the sampled classrooms were eligible to participate in the survey. In total, 1432 students from the selected 64 classrooms from the 13 schools participated in the survey, corresponding to a participation rate of 82%.
The survey took place on the same day in all selected classes. The distribution of questionnaires and collection of the answer sheets involved 33 survey officers, with one officer assigned to each participating class. Survey officers had been trained during a one-day workshop. The students and parents were not informed prior to the survey in view of the absence of invasive investigations or physical measurements, the possibility for declining participation allowed to all students, and the anonymous nature of the questionnaire. The research committee of the Ministry of Health had approved the study including the fact that parental informed consent was not necessary. There were 35 participants aged 11 years (2%), 226 aged 12 (16%), 325 aged 13 (23%), 296 aged 14 (21%), 307 aged 15 (22%), 191 aged 16 (13%), and 47 aged 17 (3%).
Data were collected using a self-administered and anonymous questionnaire that was designed along the standard GSHS methodology. The GSHS core questionnaire includes 10 core questionnaire modules containing 3-7 questions as well as several additional country-specific questions. The questions chosen from the GSHS module must be used without modification (except for translation into the local language). In Seychelles, we used questions from the modules on tobacco, alcohol, drug use, mental health, and protective characteristics.
Different cutoffs have been applied in studies that assessed risk behaviors and associated characteristics among adolescents [4
]. In this study, risk behaviors were dichotomized on the basis of definitions used by prior studies [4
]. Smoking was defined as a response of "one or more days" to the question "During the past 30 days, on how many days did you smoke cigarettes?" Alcohol use was defined as a response of "one or more days" to the question "During the past 30 days, on how many days did you have at least one drink containing alcohol?" Cannabis use was defined as a response of "one or more times" to the question "During the past 30 days, how many times have you used drugs such as cannabis, marijuana or hashish (do not include heroin, cocaine, or ecstasy)". While it should be acknowledged that positive answers may also identify students who are only engaging in experimental "one-time" experiences, it is likely that the behaviors so defined most often reflect a more habitual behavior (hence a true "risk behavior" for the majority of students). Moreover, it should be noted that substantial proportions of adolescents progress to regular substance use after occasional use, a proportion that may range from 28% to 80% [27
Four questions in the "GSHS mental health module" assess psychological characteristics: "During the past 12 months, how often have you felt lonely?", "During the past 12 months, did you ever feel so sad or hopeless almost every day for two weeks or more in a row that you stopped doing your usual activities?", "During the past 12 months, how often have you been so worried about something that you could not sleep at night" (referred to hereafter as "insomnia"), and "During the past 12 months, did you ever seriously consider attempting suicide?" Responses to these questions were on a 5-point Likert-type scale ranging from "never" to "always". Also part of the "GSHS mental health module", social interaction was assessed by asking "How many close friends do you have?" As there were only 5.8% of students who reported having no friends, the category for low social interaction was defined as having 0-1 friends (19.9% of all students).
We used the following three questions from the "GSHS protective factors module" to assess perceived parental monitoring and bonding: "During the past 30 days, how often did your parents or guardians check to see if your homework was done?"; "During the past 30 days, how often did your parents or guardians understand your problems and worries?"; and "During the past 30 days, how often did your parents or guardians really know what you were doing with your free time?" Responses to these questions were on a 5-point Likert-type scale ranging from "never" to "always". Questions on parenting practices, e.g. perceived parental monitoring or bonding, are commonly assessed among adolescents [11
] as the relationships of adolescents with their parents, as perceived by the adolescents, may affect their behaviors [33
]. Also part of this module, truancy was assessed by the question "During the past 30 days, on how many days did you miss school without permission?"
Finally, we added a question on pocket money available to the students ("How much pocket money do you get every day on average?") This variable has been employed in similar studies [16
] to reflect, to a certain extent, the socioeconomic status of the participants' parents.
Differences in the prevalence of baseline characteristics between girls and boys were tested using the chi-squared test. We considered cigarette smoking, alcohol, and cannabis use (i.e. "risk behaviors") as the response variables. In all analyses, both the response variables and the explanatory variables were dichotomized. We used age-adjusted logistic regression to examine the associations between each of the three risk behaviors and the explanatory variables, separately for boys and girls. To account for a possible clustering effect, a random classroom effect has been included in all logistic regression models, using the xtlogit command of Stata 9.2. An estimate of the intraclass correlation is also provided, which is a measure of how closely students in a classroom resemble to each other regarding the outcome variables (technically, intraclass correlation is the percentage of the outcome variability that is due to the classroom effect from the variability that is not explained by the other factors in the model; this is being measured using an underlying latent continuous variable from which the outcome is a dichotomization) [35
]. Multivariate analysis including several psychosocial variables has also been performed. Since several characteristics in the GSHS represent same constructs (e.g. several questions on parental control, etc), we used backward stepwise regression analysis to obtain parsimonious models. Age was included in all models. All participants were included in the analyses with the exception of 15 students in whom data regarding gender were missing (n = 1417). Analyses were performed using Stata 9.2 and p values < 0.05 were considered significant.