Background: Given the high burden of adverse sexual and reproductive health outcomes (SRH) and low levels of school attendance among adolescent girls in Kenya, this study sought to elucidate the association between school attendance and SRH outcomes among adolescent girls in Homa Bay and Narok counties. Methods: This study uses baseline quantitative data from the mixed-methods evaluation of the In Their Hands (ITH) program which occurred between September to October 2018 in Homa Bay and Narok counties. In total, 1840 adolescent girls aged 15–19 years participated in the baseline survey, of which 1810 were included in the present analysis. Multivariable logistic regression models were used to assess the association between school attendance (in- versus out-of-school) and ever having sex, condom use during last sex, and ever pregnant, controlling for age, orphan status, income generation, religion, county, relationship status, and correct SRH knowledge. Results: Across the 1810 participants included in our study, 61.3% were in-school and 38.7% were out-of-school. Compared to adolescent girls who were in-school, those out-of-school were more likely (AOR 5.74 95% CI 3.94, 8.46) to report ever having sex, less likely (AOR: 0.21, 95% CI 0.16, 0.31) to have used a condom during their last sexual intercourse, and more likely (AOR: 6.98, 95% CI 5.04, 9.74) to have ever been pregnant. Conclusions: School attendance plays an integral role in adolescent girls’ SRH outcomes, and it is imperative that policy actors coordinate with the government and community to develop and implement initiatives that support adolescent girls’ school attendance and education.
The present cross-sectional analysis uses baseline data from the In their Hands (ITH) evaluation to examine the association between school attendance and SRH outcomes. The ITH program was implemented in Kenya between April 2017 to March 2020 across eighteen counties to increase adolescent’s access to and use of high-quality SRH services through a targeted intervention. The evaluation utilized a mixed-method cross sectional design that collected both quantitative and qualitative baseline data concurrently in the periods between September to October 2018. Homa Bay and Narok counties were selected for the evaluation as the ITH intervention had not yet been implemented at baseline, and the counties had the highest prevalence of adolescent pregnancy of the 18 counties where the program was to be implemented. Homa Bay has a population of 1,131,950, and adolescents aged 10–19 comprise 28% of the population. Similarly, adolescents make up 26% of the total Narok population (1,157,873) [25]. Adolescent girls aged 15 to 19 years were recruited to participate in the intervention. The inclusion criteria were being an adolescent girl aged between 15 and 19 years, a usual resident in the study area (has lived at least 6 months preceding the study), and being a member of a sampled household. On the other hand, students who are in boarding schools and mostly stay away from their parents, and adolescents who were not competent for informed consenting were excluded from the study. The sampling approach included purposive selection of Homa Bay and Narok counties. Three sub-counties were selected within each of these two counties. Within each of the sub-counties, three wards were selected based on the distribution of ITH affiliated heath facilities. For each of the health facilities that were sampled, catchment villages served by the facility was identified. Resultantly, 22 and 24 villages were sampled in Narok and Homa Bay counties, respectively. Household listings of each of the villages were used to identify households with adolescent girls. Through random selection, only one adolescent girl was interviewed from each household where at least an adolescent girl was listed. The baseline dataset included 1840 adolescent girls aged 15–19 years who were residents in the study areas for at least 6 months prior to the study. The response rate for participation was 97%. The survey targeted 1897 adolescents. 1840 adolescent girls were successfully recruited and participated in the cross-sectional survey; 57 (3%) adolescent girls selected did not participate due to lack of parental consent, unavailability, or refusal to participate. The present analysis was restricted to adolescent girls who had ever attended school. Participants who had never attended school (n = 30) were excluded. Resultantly, a total of 1810 participants were included in this analysis. The study protocol and data collection instruments were reviewed and approved by AMREF Health Africa Ethics and Scientific Review Committee. Furthermore, research clearance was granted by Kenya’s National Commission for Science, Technology and Innovation. Additional approvals were obtained from local commissioners and the Ministries of Health and Education in the respective counties where the study was conducted in. Individual consent was sought from adolescents who were aged 18–19 years or were emancipated minors. For adolescents younger than 18 years old, both parental/guardian consent and adolescent assent was obtained prior to interviews and data collection. Quantitative data was collected from a representative sample of adolescent girls living in urban and rural ITH program areas. Research assistants were trained on all aspects of the study protocol. An interviewer-administered structured questionnaire was used to collect quantitative data to understand adolescent girls’ use of SRH outcomes and services, as well as their access to information, prior to the implementation of the ITH program. Interviewers used a tablet to collect the information during the face-to-face interviews. These interviews were conducted in a private setting to ensure confidentiality. While the study tools were programmed in the SurveyCTO in both English and Kiswahili, respondents who consented to participate were asked about their preferred language of interview. Furthermore, since the interviewers were recruited on their knowledge of the study area and ability to communicate in any of the dominant local languages spoken in any of the study area, where the respondents were not fluent in both languages of interview, such interviews were conducted in the local languages by competent interviewers. The questionnaire was piloted to assess consistency, appropriateness, readability, and ease of understanding of the questions. The main exposure of interest was current school attendance, defined as either currently attending school or not, at the time of the survey. This was a self-reported measure assessed at baseline by asking participants ‘Are you currently attending school?’ and was reported as a binary variable: currently in-school and out-of-school. Our SRH outcomes of interest were: (1) ever had sexual intercourse, (2) condom use during last sexual intercourse, and (3) ever been pregnant. Responses for these three main outcomes were reported and categorized as binary categorical variables (yes/no). Of those who reported ever having sexual intercourse, participants were asked if a condom was used during last sexual intercourse and if they have ever been pregnant. To guide the analysis, a conceptual model was developed a priori through the identification of potential confounders and effect measure modifiers (EMM). Based on the current literature, the present study aimed to adjust for predictors of SRH outcomes and sociodemographic variables: including age, orphan status, engagement in income generating activities, religion, county, and relationship status. In addition, we hypothesized that having correct SRH knowledge may be an EMM, as knowledge is necessary but not sufficient to improve adolescent’s health promoting behavior [26]. Age was self-reported and later recoded as a binary categorical variable [15–19]. Participants’ county was documented by interviewers. To determine orphan status, participants were asked if their biological parents were alive, and orphan status was categorized into single, double, or non-orphaned. A single orphan was defined as a child whose mother (maternal), or father (paternal) had died or was absent from their life. A double orphan was defined as a child having both parents who had died or were absent from their life. If participants reported having both parents alive and/or present in their life, they were categorized as not orphaned. Participants were asked if they had engaged in any activities for which they got money or any kind of payment in the last 6 months (yes/no) to measure engagement in income generation. Participants were asked what religion they practiced and categorized as Catholic, Protestant/Other Christian, Islam/tradition/no religion. Relationship status was classified into the following categories: currently married/in union, has boyfriend/engaged, never been in a relationship, currently not/had past boyfriend, divorced/separated/widowed. SRH knowledge was assessed by asking participants which time a woman is more likely to become pregnant if she has sexual relations. Participants who answered ‘2 weeks after her period’ were classified as having correct menstrual cycle knowledge. All other options that were reported were classified as incorrect. As a result, correct menstrual cycle knowledge was recoded into a binary variable (correct/incorrect). Descriptive statistics were used to determine frequencies and proportions for categorical variables and means and standard deviations (SDs) for continuous variables. We conducted bivariate logistic regression to examine the relationship between school attendance (in-school vs. out-of-school), SRH outcomes (ever had sexual intercourse, condom use during last sexual intercourse, and ever been pregnant), and potential confounders including, age, county, orphan status, religion, relationship status, income generation, and correct menstrual cycle knowledge. We then conducted multiple logistic regression producing three models, one for each SRH outcome of interest, controlling for confounders, to produce adjusted odds ratios (AOR) and 95% confidence intervals (CIs). Multicollinearity of the models was assessed at a threshold of 0.8, and not observed. EMM was assessed for correct SRH knowledge. Likelihood ratio tests were performed with full and nested models to determine whether interaction terms should be included; if they were not statistically significant at an alpha level of 0.05, then the reduced model was chosen as in this case. However, given that correct SRH was statistically significant, it was included in all three models as a covariate. To ensure data quality, all variables were assessed for missingness. If variables had a low proportion of missingness (characterized as less than 5%), observations were assumed to be missing completely at random [27]. As no variables had greater than 5% of missing data, a complete case analysis was employed as it was likely to not bias the estimate and sample size of each model run. All analyses were conducted using RStudio Version 1.4.1106.
N/A