Setting
The Province of Jujuy, Argentina is characterized by a geographic configuration that includes lowlands where tobacco farms are located. Tobacco farming is an important contributor to the economy of the province, with 120 to 130 workdays by farmed hectare. The majority of the tobacco workforce in Jujuy are individuals hired by mid to large scale farmers. Only 1% are small farms with less than 2 hectares of land that depend solely on family labor [26] [27].
Sampling
Secondary schools were randomly sampled from within the three geographic areas of Jujuy. Secondary schools include 8th through 12th grades and reflect the standard educational organization in Argentina. Based upon government data, we selected a representative sample of schools containing approximately 1000 eighth grade students from within each geographic area (i.e., disproportionate stratification). The final sample included 27 schools, three of which were private. The baseline data was collected in 2004 (N=4276) among all enrolled 8th grade students, and three follow up surveys were conducted between 2005 and 2007. The response rate for each follow up was 94.2%, 91.7% and 80.0% respectively. Surveys were self-administered in class with research staff and school coordinators present as proctors. In each school, one attempt was made to survey absent students at a subsequent date. The detailed study procedures have been described in a previous publication [28]. For this report we used data from the 3234 students between ages 13 and 17 years who completed surveys in 2005 (T1) and 2006 (T2). Of these, 46 (1.4%) did not answer the questions about tobacco farming, yielding a total sample of 3188. The UCSF Committee on Human Research and an NIH-certified human subjects research board in Buenos Aires based at Centro de Educación Médica e Investigaciones Clínicas (CEMIC) approved the research protocol. Passive consent was requested from caretakers and students signed an active consent.
Questionnaire Development
The questionnaire consisted of translated items from surveys of adolescents in the U.S. [29], and questions developed through qualitative research in the target population [28]. Items in English were translated and reviewed by three Argentinean investigators and two other Spanish-speaking research staff. Pilot testing of the instrument was conducted with students in rural and urban areas evaluating situational factors, content, characteristics of the respondents, and time of administration that averaged one hour.
Demographics
Sociodemographic variables were extracted from baseline data including sex, age, ethnicity (Indigenous, mixed Indigenous and European, European), and religion. Religion was categorized as Catholic, Christian or Evangelical, and others corresponding to low frequency religions. A binary (yes/no) low socioeconomic status (SES) variable was developed by classifying the primary caretaker as having up to primary education, being unemployed, or being on welfare, versus having a higher education level or being formally employed. The location of the school was reported in the questionnaire by interviewers.
Health related factors
Health related variables correspond to T1 responses. Respondents provided a self-assessment of their health status, categorized as excellent, good, fair or poor. Another set of questions probed on the occurrence of injuries. We asked if in the previous year respondents had a serious injury, if they were injured accidentally by someone else, if they had been assaulted, and if they had been poisoned by exposure to chemical products. Local agricultural workers commonly refer to pesticides as “chemicals” and the survey question was phrased accordingly.
Smoking Behavior
For this study, we used smoking information from T1 (2005) and T2 (2006). Smoking behavior was the main outcome and questions were developed to be comparable to those used in the Centers for Disease Control and Prevention GYTS survey [29]. Respondents were considered ever smokers if they tried at least a cigarette puff in their lifetime and never smokers had not tried even one puff. Current smokers were defined as having smoked at least one whole cigarette in their lifetime and at least one puff in the previous 30 days. Established smokers were defined as current smokers who had smoked at least 100 cigarettes in their lifetime. Respondents also reported on the number of friends who smoked (none, 1 to 4, 5 or more), and whether any adult smoked in their home.
Working in Tobacco Farming
Hereby reported exposure variables correspond to measurements at T1. The youth were asked if they had ever worked in any of the tasks involved in tobacco production, growing, harvesting or selecting tobacco leaf, without discriminating the different types of tasks. Youth reported their age of initiation in tobacco farming work. Information about working in non-tobacco farming occupations was also requested.
Data Analysis
The sampling design was incorporated into all models by specifying geographic areas as strata and schools as clusters as well as including weights to adjust for disproportionate stratification. In addition, a finite population correction was applied to adjust for the relatively large proportion of available schools sampled within each geographic area. The statistical program Stata (version 14.2) was used for data analysis. Standard errors and confidence intervals were estimated via the Taylor expansion approximation using the svy procedures in Stata [30]. First, we conducted descriptive analyses by sex, to profile the sample. We calculated the prevalence of ever, current and established smoking, with chi square tests and p values at T1 and T2, and the percentage of youth who reported at T1 that they had ever worked in tobacco and non- tobacco farming. The mean and standard deviation of the age for girls and boys, and of the age of initiation in tobacco farming was calculated. Bivariate contingency tables examined the pairwise relationship of sociodemographic characteristics, health related factors and smoking behavior by sex, and by working in tobacco farming. Bivariate analysis also examined the pairwise relationship of non-tobacco farming and the smoking behavior variables.
Multivariate logistic models regressed working in tobacco farming at T1 with each of the health-related variables at T1. Separate multivariate logistic models regressed working in tobacco farming at T1 onto cigarette smoking behaviors at T2 (ever, current or established smoking). Covariates included sociodemographic characteristics (sex, age, low SES, ethnicity, religion, number of friends who smoked, adult smokers at home, and for each model, the corresponding smoking behavior at T1 (ever, current or established smoking). We estimated adjusted odds ratios and 95% confidence intervals.