Community social capital on the timing of sexual debut and teen birth in Nicaragua: a multilevel approach

This research used data from the DHS conducted in Nicaragua (years 2011/2012). The DHS is a nationally representative survey, with cross-sectional design selecting households through a two-stage sampling strategy to interview women aged 15 to 49 years and men aged 15 to 59 years. This series of surveys are carried out every 5 years and aim to generate information about maternal and child health. The information was collected by face-to-face interview, and the response rate of eligible women was 91 % [34]. In order to facilitate responses to sensitive issues and decrease social desirability, the DHS’ data collection procedure strongly recommended the respondents to be interviewed by a same-sex member who was also responsible for filling the information given during the survey [35]. In the present study, the analytic sample includes 2 766 females Nicaraguan who were aged 15 to 19 years at survey time. This sample size exceeds the minimum sample size of 1 365 adolescents required to achieve 65 events (i.e. sexual debut) and provides enough statistical power to detect differences in the distribution of the outcome across main predictors [36]. As a proxy for community we used the census tract (n = 356), which usually corresponds to a village in the rural area or to a neighborhood in the urban area [37].

This study based its definition of community social capital on Bordieu and Putnam’s theories. The former incorporates the notion of access and utilization of common resources for mutual benefit [21, 27], and the latter concerns “features of social organization that can improve the efficiency of society by facilitating coordinated actions, such as trust, norms and networks” [38]. Given that the formulation of community social capital depends on the existence of trust to others community members and the strength of connections among them [39], we hypothesized that those social processes are more fragile for people who have lived in the same community for less than 10 years (the same cutoff point of 10 years has been used in prior research using DHS) [37, 40, 41]; they may be less socially integrated which may hinder the access to emotional and material resources that operate in favor of health [28]. Therefore, our first main predictor is the proportion of people who have lived in the community for less than 10 years. This variable was created to take into account the responses from men aged 15 to 59 years and women aged 15 to 49 years: each respondent was asked “How long they have been living in the community”, and the responses for those who have lived for less than 10 years were aggregated and the mean proportion by census tract was estimated. The second main predictor is related to religious community composition. Because 50 % of Nicaraguans are affiliated with Catholicism, we focused on this religious denomination [42]. The percentages of Catholics in the community were derived using the responses from selected women and men regardless of their age. A similar procedure was followed to estimate the percentage of community inhabitants with no religious affiliation in each census tract. We anticipated this approach allows us to capture the heterogeneous composition of community social networks (e.g., religious institutions) and therefore obtain more accurate estimations of the extent to which the community-level of trust and participation influence adolescent’s reproductive outcomes. As regards social norms, a teenage girls may be influenced not only by their peers but also by adult women’s reproductive behaviors in their surrounding environment (e.g., women from her families or neighborhood), hence we estimated community social norms indicators (i.e., proportion of childbearing age women with sexual debut and having a child) based on the responses of women between 15 to 49 years in each census tract [43]. Finally, the proportion of community members with secondary school or higher education was also estimated by averaging the responses of women aged 15 to 49 years and men aged 15 to 59 years in each census tract. To facilitate the interpretation, all community-level variables were standardized.

At individual-level, we also took woman’s years of education, women’s household wealth index quintiles, marital status (at survey time), woman’s religion, residence area, female age at sexual debut, and prior history of abortion into account by statistical adjustment [12, 13].

The analyses were broken down into two parts. First, to describe the population characteristics and estimate the median survival time to first birth, we used the complex survey analysis to account for a multi-stage sampling design and correct for unequal probabilities of selection [44]; for community variables, central tendency measures were estimated. Second, we implemented survival regression models to estimate the hazard rate of the timing of sexual debut and first birth. Since adolescents are nested within communities, which leads to a violation of the assumption of independent distribution of error terms [26], the utilization of standard Cox regression would lead to imprecise estimations. To overcome this issue, we employed a two-level survival model, with teenage girls as the first level and communities as the second level. Model 1 displays the unadjusted hazard ratios from community- and individual-level variables. In model 2 each contextual variable was adjusted for individual characteristics, and finally we entered all individual- and community-level variables in model 3. For all models we incorporated probability weights at community level. Descriptive analyses were conducted in Stata 12.0 and regression analysis in R 3.2.0 using frailtypack.

All DHS protocols were approved by the ICF’s Institutional Review Board (IRB) [37]. A parent or guardian must provide consent prior to participation of adolescents [35]. The DHS guarantee respondent’s anonymity.