Data
Data were extracted from four consecutive Nigeria Demographic and Health Surveys (NDHSs) which were implemented between 2003 and 2018. The NDHS is a subset of Demographic and Health Survey (DHS) programme executed nationally in more than 90 countries with more than 300 surveys successfully conducted so far [38]. The quality of data emanating from the DHS Programme has been widely adjudged to be credible and of high quality [39–41]. The National Population Commission (NPC) is responsible for implementing the DHS programme in Nigeria while the financial and technical aspects are supported by ICF International through major international development partners [42]. The DHS programme seeks to make available reliable and accurate information on countries’ demographic and health characteristics, which are mostly used by policymakers to assist countries to monitor improvement in health and family planning programme [42]. The NDHS as part of the DHS programmes provide reliable national information on fertility, nutrition, marriage, family planning, mortality, HIV/AIDS, female genital mutilation and anthropometrics information in all the six geo-political zones and the 36 states of Nigeria including the Federal Capital Territory [42].
The sampling procedures in the four rounds of the NDHS were based on the same methodology that employed a multi-stage sampling process. The 36 administrative units of the country and the Federal Capital Territory were stratified into urban and rural areas from which some urban and rural areas were randomly selected. In the selected urban and rural areas, localities used as Enumeration Areas (EAs) in the population census were randomly selected and used as the primary sampling unit (cluster). In the selected clusters, households were listed and selected randomly for the surveys. Eligible men and women were then randomly selected in the households but different numbers of men and women were covered in each round of the surveys. Further details of the survey methodology of the NDHSs have been published elsewhere [43–45] In all the four rounds of the NDHS covered in the study, the numbers of single mothers were pooled for analysis. The inclusion criteria were being sexually active, never married, separated or divorced with at least one living child. The resulting sample size was thus a weighted sample of 7,215 women.
Data Pooling
Based on the inclusion criteria, the sample size of single mothers was relatively small across the four datasets. As a result, the four consecutive NDHS (2003–2018) were pooled in order to increase the observational cases, statistical power and representativeness of the results [46]. This pooled data could increase the ability of weak but scientifically important variables to predict the response variable [47]. This method has been extensively used by researchers to investigate rare issues affecting sub-grouped of population which might be practically difficult in individual data or studies [48–51]. The demerit attached to this approach is the difference in the sample size of each dataset which was technically addressed by applying the weighting factors with unique primary sampling unit [52–53]. Also, the fear of heterogeneity of the datasets and possibility of yielding spurious results were eliminated by the similarities of the datasets in terms of variable measurement, context, characteristics of respondents, sampling design, procedures and implementation and objectives [54–55].
Research Variables
The outcome variable was current modern contraceptive utilisation with two possible responses of yes or no. Single mothers who currently use any modern contraceptive were grouped as ‘yes’ and coded ‘1’ while those who are not currently using a modern method were grouped as ‘no’ and coded ‘0’. The explanatory variables are sets of selected socio-economic and demographic characteristics. The socio-economic characteristics are maternal education (none, primary, secondary and higher), household wealth quintile (poorest, poorer, middle, richer and richest), religious affiliation (Christianity, Islam and tradition/others), employment status (employed or unemployed), place of residence (urban or rural), geographic region (northern or southern) and exposure to mass media (low, moderate and high). The three geo-political zones in the southern parts of the country, namely, southeast, south-south and southwest zones are combined as southern region while the zones in the northern parts of the country, namely, northcentral, northeast and northwest zones are combined as the northern region. Exposure to mass media was generated from three variables, namely, frequency of reading newspaper, listening to radio and watching television. Single mothers who do not access these outlets or accessed the outlets less than once a week were grouped as low exposure. Those who accessed the outlets at least once a week were grouped as moderate exposure. Other single mothers who accessed the outlets almost every day were grouped as high exposure.
The demographic characteristics are age group (15–24, 25–34 and 35–49 years), age at sexual debut (less than 18 years or 18 or older), parity (one child, two-four children, and five or more children) fertility desire (wanted more children or wanted no more), and child living arrangement (lives with mother or lives elsewhere). These variables are selected based on their significance in previous studies [33–34, 56–60]. The control variables are nature of singlehood (premarital or post marital singlehood), sexual activity (active or inactive) and most recent sexual partner (boy/man friend, commercial worker and casual friends).
Data Analysis
Data analysis was carried out using Stata version 14 [61]. Univariate analysis was carried out to assess the prevalence and use of modern contraceptives and to describe the sample characteristics. At the bivariate level, the relationship between each of the explanatory variables and the outcome variable was examined using the Unadjusted Odds Ratio (UOR) of binary logistic regression. Any variable that reveal no statistical significance at p < 0.025 was excluded from further analysis. The multivariable logistic regression analysis was used to examine the factors influencing the outcome variable using the Adjusted Odds Ratio (AOR) with 95% confidence interval. Statistical significance was set at p < 0.05.