The dataset for this study was obtained from the measure DHS program after permission was granted at http://www.dhsprogram.com. A total of 36 sub-Saharan African countries’ most recent DHS datasets from 2006 to 2019 were used in this study.
Data from the southern region of Africa (Lesotho, Namibia, Swaziland, and South Africa), the central region of Africa (Angola, Democratic Republic Congo, Congo, Cameroon, Chad, Gabon, Sao Tome & Principe), the Eastern region of Africa (Burundi, Comoros, Ethiopia, Kenya, Madagascar, Malawi, Mozambique, Rwanda, Tanzania, Uganda, Zambia, and Zimbabwe), western Africa (Burkina-Faso, Benin, Cote d’Ivoire, Ghana, Gambia, Guinea, Liberia, Mali, Nigeria, Niger, Sierra Leone, Senegal, and Togo) was included.
Each country’s survey consists of different datasets including men, women, children, birth, and household datasets. For this study, we used the individual records data set (IR file) where data on women’s health is recorded. Demographic and health survey is conducted at the five-year interval and follows a common execution procedure in each country.
A two-stage stratified sampling procedure is adopted to select study participants in each survey. In the first phase, Enumeration Areas (EAs) were selected based on the sampling frame of each respective country. In the second stage, a sample of households is drawn from each EA. Then eligible study participants were interviewed in the selected household. The detail of the sampling procedure has been documented elsewhere . For the current, study a total of 17,797 (weighted sample) adolescent girls and young women [15,16,17,18,19,20,21,22,23,24] having a pregnancy at the time of the interview in 36 sub-Saharan African countries were included (Table 1). During analysis sampling weight was applied using individual sample weights recorded in the data set to produce reliable estimates by adjusting the over and under-sampled regions.
The dependent variable for this study was unintended pregnancy. It was measured in such a way, by asking women about their pregnancy to state just when they wanted their pregnancy (then, later, or not at all). Those women responding to the above question as ‘wanted later’ or ‘not wanted at all were considered to have an unintended pregnancy and those who responded by saying ‘wanted then’ were considered to have intended pregnancy. Therefore, unintended pregnancy was coded ‘1’, and intended pregnancy was coded ‘0’ for further statistical analysis.
Individual and community-level variables were retrieved from DHS datasets. Age [15,16,17,18,19,20,21,22,23,24], educational level (no education, primary, secondary, and higher), marital status(single, married,), wealth index (Poorest, Poorer, poor, Richer, and Richest), media exposure (yes, no), heard about family planning from media (yes, no), knowledge of contraceptive methods (no, traditional, modern), distance to health facility (big problem and not a big problem), smoking (yes, no), covered by health insurance (no, yes), sex of household head (male, female) and occupation (not working, working) were individual-level variables. Whereas residence (urban and rural) and SSA region (South Africa, Central Africa, East Africa, and West Africa) were community (country) level variables.
The descriptive statistics was presented in Table 2. The overall prevalence of unintended pregnancy among adolescent girls and young women in sub-Saharan Africa with 95%CI was reported. A multilevel logistic regression model was fitted to assess the factors associated with unintended pregnancy. Consequently, four models were fitted. First, the null model without explanatory variables was fitted by using the country as a group variable to assess the community (country) level variance and the applicability of multilevel analysis. Model II and model III were adjusted for individual-level variables and community-level variables respectively. In model IV, both individual-level and community-level variables were fitted simultaneously. Deviance was used for model comparison. Accordingly, the final (Model IV) was the best-fitted model. In bi-variable analysis variables which are eligible for multivariable analysis were selected at a p-value of 0.2. The multi-collinearity was checked using the variance inflation factor (VIF) to avoid the inflation of the effect size of independent variables. In the multivariable analysis, an Adjusted Odds Ratio (AOR) with 95% CI was reported and variables with a p-value ≤0.05 were considered significant determinants of unintended pregnancy.