We carried out a cross-sectional study in 2009 in a town that belongs to the metropolitan urban area of Belo Horizonte, capital of Minas Gerais state, Brazil.
Our sample of adolescents was drawn from a philanthropic non-governmental organization (NGO) which was established in 1966 with the purpose of connecting working adolescents to local employers. The adolescents self-refer to the services of the NGO, which maintains the following admission criteria: a) the adolescents must be aged between 16 years to 17 years and 11 months, b) they must be students enrolled in a public school, c) they must come from economically vulnerable family backgrounds, and d) they must take and complete the admission course provided by the NGO. Neither race nor religious backgrounds are taken into consideration at admission. During the admission course, the youth have an opportunity to learn about the NGO statute and its bylaws, to attend Portuguese and reading classes, and to be prepared to fit in the companies that will hire them.
By the end of the course, all the adolescents receive a qualification certificate on human relations, work team and professional behaviors. Once accepted by the NGO, the youth have an opportunity to get a part-time job and also to attend free sports classes and extra recreational activities on weekends.
The recruitment of the participants required one day and the adolescents recruited in this present study were all those present during a special weekend meeting and who had signed assent to participate in the survey (and from whom we also obtained consent from either their parents or guardians). For this present study, we used the World Bank Integrated Questionnaire to Measure Social Capital (SC-QI), which is a psychometrical validated instrument . The World Bank group study distributed some indicators into dimensions based on their previous experience with social capital survey and reading scientific literature. This instrument has 27 items and is divided into six different dimensions of social capital as follow: a) groups and network, b) trust and solidarity, c) collective action and cooperation, d) information and communication, e) social cohesion and inclusion, f) empowerment and political action. The IQ-SC questionnaire was conceptualized for the micro level (individuals) and does not collect data on the level of communities, thus it is an adequate instrument for the objective of our study.
The IQ-SC has no overall scoring algorithm. This explains the possibility of our selection of some social capital indicators among the total 27. One of co-authors (I.K.) is an expert on social capital and health studies and selected 14 of 27 items based on social epidemiological evidence and the peculiarities of adolescence age group.
All the adolescents in the NGO (N = 363) participated in the survey. Ethics approval was granted by Research Ethics Committee at Pontificia Universidade Catolica de Minas Gerais.
Our outcome measure was a one-item self-rated overall health status question. Individuals responded to the following question: "How would you describe your overall state of health these days, would you say it is very good, good, fair, poor or very poor?". We combined the categories to have a binary outcome of self-rated health where 0 = very good, good or fair and 1 = poor or very poor.
The cognitive dimensions of social capital were assessed by four questions inquiring about: a) trust in others, b) perceived helpfulness of neighbors and c) perceptions of whether the youth could borrow money from others in case of need. The variable "trust in people" was dichotomized as: 0 = people can be trusted or 1 = you can't be too careful. The variable "helpfulness of neighbors" was assessed by the extent to which youth agreed with the 2 statements that: "most people in this neighborhood are willing to help you in case of need" and "in this neighborhood, one has to be alert or someone is likely to take advantage of you". The answers (agree strongly, agree somewhat, neither agree nor disagree, disagree somewhat, disagree strongly) were dichotomized into 0 = agree versus 1 = disagree or unsure. The variable "borrow money" was determined by a five scale question asking if there was someone beyond family or close relatives willing to help the youth in case of need to borrow a small amount of money (definitely, probably, unsure, probably not or definitely not), which was dichotomized to 0 = yes and 1 = no or unsure.
The behavioral dimensions of social capital were assessed by: a) participation in community activities during the past twelve months; b) time or money contribution to a community project (0 = yes versus 1 = no); c) whether they belonged to a group (0 = at least one or more, 1 = zero); d) whether they had a close friend, and e) whether they got together with people to have food or drink in the past month 0 = at least one or more, versus 1 = no.
Finally, we assessed bonding versus bridging social capital by asking respondents to think about the people with whom they associated. Separate questions inquired whether the youth associated with others from different ethnic backgrounds, different economic or social status or different religious groups. For each of these questions, the responses were coded as 0 = yes (bridging) and 1 = no (bonding).
The variables sex (male or female), age (0 for 17 years old vs. 1 for 15-17 years old), skin color (white, black, brown, yellow, Indian - classification of Brazilian Institute of Geographic and Statistic) and educational background (first or second grade of middle school - correspond to the 11th grade American high school) were entered as possible confounders in regression models.
The SPSS statistical package version 18.0 was used for data analyses. Descriptive statistics were performed to characterize the participants including relative and absolute frequencies of the explanatory variables and confounders as well their associated odds ratio for poor self-rated health with 95% confidence intervals (95% CI).
Next, the Enter method was used for the logistic regression models. Variables with a p-value lower than 0.2 were included in the logistic regression models. Although social demographic variables did not meet the formal inclusion criterion [(sex p = 0.510), (age p = 0.714), (educational background p = 0.725), (skin color p = 0.515)], they were nonetheless retained in the regression models due to their theoretical plausibility.