Modern contraceptive utilization and its associated factors among married women in Senegal: a multilevel analysis

Senegal, located in West Africa, is well-known as the “Entry to Africa” [40, 41]. Up to half of its 15.4 million people (2016) live in and around Dakar and other urban areas [40]. In Senegal, there is a hierarchical health system model, where regional medical officers oversee the activities of health districts in every region. Health districts are made up of a health centre and a linkage to rural health posts, often under the supervision of allied community-level facilities [42]. These health facilties are operated by community health workers [43]. In senegal, family planning usage has been low- only about 18% of women with partners aged 15–49 used a modern contraceptive method in 2013 [10]. In 2015, it increased marginally to 21.2%, and 21–23% had an unmet need-non-use of contraceptives among fecund and sexually active women who do not want any more pregnancy or have the intention to delay the next pregnancy [44], for family planning [45].

Data from the 2017 Continuous-Demographic and Health Survey (C-DHS) [45] was used in this study. The C-DHS is a nationally representative data undertaken by the Senegal Agency of Statistics and Demography (ANSD) in collaboration with ICF International [45]. A two-stage stratified sampling procedure is adopted in selecting research participants. In the first phase, clusters/enumeration areas (EAs) are selected, guided by a sample frame developed during the 2013 census [45]. In all, 8800 households, made up of 4092 located in urban areas and 4708 in rural areas were considered in the survey [45]. For this study, a sample of 11,394 married women aged 15–49 with information on modern contraceptive use were considered.

The main outcome variable of this study was modern contraceptive use. To derive this variable, married women were asked of the contraceptive methods they use. Those who mentioned that they used modern contraceptives were coded as modern contraceptive method users while those who used traditional/folkloric/no method were coded as non-modern contraceptive users. Details of the types of contraceptives considered as modern, traditional or folkloric have been published in previous studies [2, 4, 46]. Married women who were sterilized and declared as infecund were excluded from the analysis. Several individual and community-level explanatory variables were chosen based on prior evidence [6, 14‐16, 18, 19, 36, 47‐50].

These included demographic and socio-economic variables such as women’s age (15-19, 20-24, 25-29, 30-34, 35-39, 40-44, 45-49) women’s educational level (No formal education, primary school, secondary school, higher), husband/partner education (No formal education, primary school, secondary school, higher), number of living children (no children, 1–2 children, 3–4 children, five or more children), ideal number of children (no child, 1 child, 2 children, 3 children, 4 children, 5 children, 6 children and above and non-numeric response (e.g upto God/Allah) and desire for more children (wants within 2 years, wants after 2 years, wants but unsure of timing, undecided, wants no more). Others were distance to health facility (a big problem, not a big problem), ethnicity (Wolof, Poular, Serer, Mandingue/soc., Diola, Sonink, other Senegalese and other), religion (Muslim and Christian) and media exposure (newspaper, radio or television [TV]). Media exposure was derived from a combination of three variables; reading of newspaper or magazine, listening to radio and watching television. DHS grouped all of the three variables into three categories not at all (0), less than once a week (1) and at least once a week (2). We then, re-categorized these into two (binary) for all the three variables. Those who didn’t read newspaper/magazine, or listened to radio or watched television at all were coded as zero (0), and those who read newspaper/magazine, listened to radio or watched television for less than once a week and at least once a week were coded as one (1). We then generated, media exposure by adding the three variables. If the respondent had no exposure at all for all three variables we coded as ‘no’. If the respondent had exposure to at least one of the three variables, we coded as ‘yes’.

Other individual level variables included wealth index (poorest, poorer, middle, richer, richest), attitude towards wife beating (refuse versus accept) and women’s empowerment [51, 52]. Women empowerment as a variables was derived from respondent’s decision making capacity on own health care, large household purchases, and visits to family and relatives. Respondents whose responses to these three decision making variables showed that their decisions are made by their husband/partner alone or other were considered as having “no decision-making”. Respondents who mentioned that they took decisions alone or with their husband/partner on two of the decision making variables were put in the category “decision making one”. Finally, women who decided alone or with their partner/husband on all the three decision making variables were coded into “decision making two”.

Place of residence (urban, rural), region (Dakar, Ziguinchor, Diourbel, Saint Louis, Tambacounda, Kaolack, Thies, Louga, Fatick, Kolda, Matam, Kaffrine, Kedougo and Sedhiou), community-level literacy-proportion of women who can read and write (low medium, high), community-level socioeconomic status-the proportion of women within the richest wealth quintile (low, medium, high) and community-level knowledge about modern contraceptive-the proportion of women with knowledge about modern contraceptives (low, medium, high) were considered as community-level variables.

This study employed both descriptive statistics and multi-level logistic regression models in the data analyses. The descriptive analysis employed percentages to describe the study variables. This was followed by the use of chi-square test to show the independent associations between the individual and community level variables and modern contraceptive use using p-values < 0.05 as cut of points. Finally, multilevel logistic regression analysis made up of four models was carried out using the “melogit” stata command. Results were presented using adjusted odds ratios (aOR) at 95% confidence intervals (CI). The multilevel logistic regression models consisted of fixed effects (measures of association) and random effects (measures of variability) [19, 53, 54]. Muli-collinearity test was done among the independent variables using the variance inflation factor (VIF) and the results indicated no evidence of high collinearity among the explanatory variables (Mean VIF = 1.43, Minimum VIF = 1.01, Maximum VIF = 2.26). Previous studies have shown that VIF values less than 10 are tolerable [55‐57]. Sample weights were employed and the ‘svyset’ command was used to correct for under and over-sampling [26, 58]. All analyses were carried out in stata 14 for windows.

Data for this study was derived from the website of MEASURE DHS through a standard national survey. Permission to use the data was granted by MEASURE DHS review board. All data were anonymized prior to the authors receiving the data.

Out of the 11,394 married women, 26.3% were using modern contraceptives. Results on the distribution of modern contraceptive use across the explanatory variables show that most of the women who used modern contraceptives were aged 35–39 (29.62%), had higher level of education (36.68%), had husbands with higher level of education (34.83%), lived in urban areas (31.36%) and belonged to the Serer ethnic group (28.47%). The results further showed that most of the women who used modern contraceptives were Christians (217.61%), had richest wealth index (33.84%), had five or more children (32.75%), were against wife beating (27.62%), lived in Dakar (42.05%) and were exposed to media (24.19%). Further results also indicate high modern contraceptive use among women with 2 children (34.23%), those who considered distance to the health facility as not a big problem (25.13%), those who lived in high literacy level communities (33.47%), those who lived in communities with high socio-economic status (32.65%) and those who lived in communities where knowledge of modern contraceptives was high (28.94%). The results of the bivariate analysis indicate that all the explantatory variables had statistically significant associations with modern contraceptive use at p < 0.05 (Table 1).

The results on the individual level factors of modern contraceptive use showed that women’s age (45–49 years-aOR = 0.44, 0.30–0.63), women’s educational level (higher-aOR = 1.88, 1.28–2.76) and husband’s educational level (higher-aOR = 1.43, 1.10–1.85) were associated with  modern contraceptive use among married women in Senegal. Number of living children (5 or more children-aOR = 36.01, 22.73–57.05), ideal number of children (2 children-aOR = 1.95, 1.13–3.35), desire to have more children (wants no more-aOR = 2.46, 2.06–2.94), ethnicity (Diola-aOR = 0.70, 0.50–0.99) and media exposure (yes-aOR = 1.44, 1.16–1.79) also had significant associations with modern contraceptive use among married women in Senegal. Other individual level factors associated with  modern contraceptive use were wealth index (richer-aOR = 1.31, 1.03–1.67) and decision making power of women (decision making two-aOR = 1.20, 1.02–1.41). In relation to the community-level factors, region (Matam-aOR = 0.35, 0.23–0.53), place of residence (rural-aOR = 0.76, 0.63–0.93), community literacy level (high-aOR = 1.31, 1.01–1.71) and community knowledge level of modern contraceptive (high-aOR = 1.37, 1.13–1.67) were found as significant factors (see Model 3 of Table 2).

In terms of model fitness, the complete model (Model 3), which included all the individual and community level factors had the highest AIC of 10,226.48 and the lowest log-likelihood ratio of − 5048.23 and was thus considered as the best fit model for predicting the occurrence of modern contraceptive use among married women.

The key strength of this study lies in the use of nationally representative DHS data and a large sample size. Nonetheless, some study limitations were also observed. The study design (cross-sectional) does not give us the opportunity to establish causality between variables. Second, since we used secondary data, other determinants such as quality of care including counseling about side effects and accessibility of supply, and cultural factors were not included in the models. To deal with the limitations associated with the cross-sectional design and the absence of some essential variables in this study, a longitudinal study that uses primary data and gathers information on all relevant factors associated with the use of modern contraceptives is recommended.