Empowerment and use of modern contraceptive methods among married women in Burkina Faso: a multilevel analysis

Data were retrieved from individual women’s files available from the 2010 Burkina Faso Demographic and Health Survey (DHS) [38]. This survey collected nationally representative data on women’s empowerment, modern contraceptive use and socioeconomic characteristics. It employed a two-stage cluster design stratified by urban and rural areas. The first stage consisted of sampling clusters based on the enumeration areas (EAs) delineated in the 2006 population and household census. A total of 574 clusters were selected by probability proportional to size sampling; 176 of the clusters were in urban areas, and 378 were in rural areas. However, one EA located in the Sahel region was not surveyed. The second stage involved the systematic sampling of approximately 25 households from each cluster, which yielded a total sample of 14,947 households; among these households, 14,424 were interviewed. All women aged 15–49 years in the households were eligible, yielding a final sample of 17,087 women from 14,242 households and 573 clusters, with response rates of 99.8, 99.2 and 98.4%, respectively.

The use of contraceptive methods was measured among MWRA aged 15–49 years with family planning needs, that is, those who wanted to avoid unintended pregnancies. According to the DHS, MWRA are women who are fecund, are currently married, and want to postpone their next birth for two or more years or to stop childbearing altogether. However, the definition also includes pregnant women whose pregnancies are mistimed or unwanted and amenorrhoeic women whose last birth was mistimed or whose last child was unwanted. After the removal of missing values, the unweighted final population included 4714 MWRA at risk of or with experience of unintended pregnancies (Fig. 1).

The outcome variable was mDFPS among MWRA. To assess the use of contraceptive methods, the DHS first asked about any current use of contraception and then the type of method used. According to the DHS, modern contraceptive methods include female and male sterilization, intrauterine devices, injectables, implants, the pill, male and female condoms, the lactational amenorrhea method, emergency contraception, diaphragms, foams and jellies [38]. Rhythm and withdrawal are considered traditional methods. The outcome variable is binary. It takes the value “0” if a woman used a traditional method or did not use contraception and “1” if a woman used a modern method.

Our main exploratory variable is empowerment, measured at the community level as factors of gender-based discrimination and opportunity and measured in marital relationships as women’s agency.

Based on previous research, we explored the following relevant and context-specific components of women’s agency: influence in family decisions, attitudes toward domestic violence and freedom of movement in public spaces [33]. In the 2010 DHS, the ability to participate in household decision-making was an essential indicator of agency. It was assessed based on women’s autonomy (jointly or alone) to decide on their own health care, make major household purchases, and visit family or relatives. However, this autonomy could be overridden when they faced domestic violence. Another section of the survey instrument assessed women’s attitudes toward domestic violence by asking whether (yes, no, or don’t know) a husband is justified in hitting or beating his wife in each of the following five situations: if she burns food, if she argues with him, if she goes out without telling him, if she neglects the children, and if she refuses to have sexual intercourse with him. Furthermore, other factors may prevent women from accessing healthcare during their reproductive lives. To assess these problems, one section asked women whether (yes or no) each of the following factors would be a significant problem for them in seeking healthcare: permission to go for treatment, money for treatment, distance to a health facility, and a desire to not go alone. All these sections on women’s ability were systematically included in all DHS regardless of the country and without an assessment of the robustness and consistency of the instrument. The construction of the indicators of agency among MWRA is shown in Additional file 1.

In terms of gender-based inequality within communities, we derived community-level variables of inequality from individual-level data, as the DHS does not collect cluster-level data. Similar studies on community effects on contraceptive use have aggregated the values of all members of a cluster or Primary Sampling Unit (PSU) to create a community mean [9, 10, 27]. The PSUs constitute a closer approximation of the actual community. The state of discrimination against women was assessed based on the cluster averages for acceptance of domestic violence, marriage before the age of 18, female genital mutilation, unpaid work and high fertility expectations. On the other hand, access to opportunities for women was evaluated based on the cluster averages for house/land ownership, secondary education, exposure to family planning messages, and contact with family planning workers. An operational description of the community-level variables is provided in Additional file 2.

We accounted for the socioeconomic factors in each cluster by adjusting for women’s age and education as well as their household wealth and place.

The weighting, stratification and clustering factors defined by the DHS were used in analyses, except in component analysis and Cronbach’s alpha testing. Descriptive statistical analyses of mDFPS, empowerment indicators, and socioeconomic factors were performed. All analyses were performed with Stata version 14 [39].

First, we applied principal component analysis and Cronbach’s alpha test to explore the components and assess the consistency of married women’s agency in the Burkina Faso context. We then explored community-level indicators of gender equality by assuming that individuals in the same community were likely to be similar than individuals in different communities [40]. Community averages were dichotomized into “high” and “low” groups based on national medians for descriptive purpose (Table 2); they were used as continuous variable in logistic regression (Table 3). Then, the chi-square test was used to assess bivariate relationships. In addition, the variance inflation factor (VIF) was checked for multicollinearity using a threshold of less than five (< 5).

The hierarchical structure of the DHS data requires the consideration of clustering within communities. The percentage of variance related to clustering is represented by intraclass correlation (ICC), which derived from the null model in which only cluster variable was included as random-effect term. We accounted for unobserved community-level variations using multilevel analysis [40].

The regression strategy involved the fitting of three models. The first (Model 1) included the components of agency in marital relationships. Model 2 included community-level indicators of gender equality as continuous. Model 3 controlled for socioeconomic factors. Robust standard errors were used.

The proposed analysis was exempted from review as described in the enforcement rules issued by the Institutional Review Board of National Yang-Ming University.

Overall, less than one-third (30.7%) of all 4714 MWRA who had family planning needs used modern methods. Meanwhile, two-thirds (66.9%) used no contraception at all, and 2.4% relied on traditional methods.

In the exploration of the dimensionality of agency among married women, three components were retained. These three components matched the three sections of the DHS on participation in household decision-making, problems accessing healthcare and attitudes toward domestic violence; the eigenvalues were 1.7, 2.2, and 3.1, respectively. The evaluation of the internal consistency of each dimension yielded Cronbach’s alpha coefficients of 0.65, 0.83 and 0.70 for participation in household decision-making, problems accessing healthcare and attitudes toward domestic violence, respectively (Table 1). Following this analysis, we grouped each dimension into higher and lower levels of agency. For instance, women who solely or jointly participated in decision-making (regarding family visits, their own healthcare, or household purchases), had no substantial problems accessing healthcare (permission to go, money needed for treatment, distance to the health facility, or a desire to not go alone) or did not agree with domestic violence at all in any of the listed situations (the wife going out without telling her husband, neglecting the children, arguing with her husband, refusing sex, or burning food) were considered to have higher levels of agency (Table 2).

Overall, 61% of MWRA participated in some family decisions, and 22% of them had no problems accessing healthcare, while 56% did not agree with domestic violence at all. All three dimensions were positively associated with mDFPS. With respect to the community-level indicators of gender equality, there were lower proportions of mDFPS in communities with high levels of acceptance of domestic violence, early marriage, female genital mutilation, unpaid work, fertility preferences, and asset ownership. In contrast, a higher proportion of mDFPS was reported in communities with a high level of secondary education and exposure to family planning messages. For socioeconomic factors, there was a higher proportion of mDFPS reported among women living in wealthier households, women living in urban areas, older women, and more educated women (Table 2).

In our regression strategy, we sequentially added components of women’s agency in marital relationships (Model 1), community-level indicators of gender equality regarding discrimination and access to opportunities for women (Model 2), and socioeconomic factors (Model 3). The results are displayed in Table 3.

Model 1 showed significant positive associations of all three components of agency with mDFPS. The odds of mDFPS increased by 30, 62 and 36% for women who participated in some family decisions, faced no problems accessing healthcare, and were opposed any type of domestic violence, respectively. In Model 2, the addition of community-level indicators decreased the estimates in Model 1; however, the significance and directionality remained. Women in communities with higher prevalence of female genital mutilation (OR: 2.59; 95% CI [1.60–4.18]), were significantly more likely to report mDFPS. In contrast, living in communities with higher fertility expectations (OR: 0.70; 95% CI [0.60–0.82]) lowered the odds of reporting mDFPS. In parallel, residing in a community with higher average access to secondary education (OR: 4.91; 95% CI [2.08–11.55]) and exposure to family planning messages (OR: 2.95; 95% CI [1.83–4.74]) was positively associated with mDFPS. Finally, when socioeconomic factors were added in Model 3, the effect size and the statistical significance of agency components were greatly reduced, but the directionality remained the same. Although opposition to domestic violence and participation in family decisions lost their significance, they were still marginally and positively associated with mDFPS. In contrast, previous estimates of community-level indicators remained essentially unchanged in their significance and directionality, except for secondary education. Furthermore, household wealth, as well as women’s age and education, were positively associated with mDFPS.

Less than one-third of MWRA who had family planning needs were able to use modern methods. In Burkina Faso, Choi et al. [15] found 40% of mDFPS, but the pace of the growth was three times slower than that required to reach 75% by 2030.

Similar to previous research, we found that participation in household decision-making, a lack of problems accessing healthcare, and negative attitudes toward domestic violence constituted separate components of women’s agency [33, 41]. Moreover, the variables for each component were fairly consistent within our study population. In addition, three-fifths of the MWRA either were able to participate in household decisions or held negative views on domestic violence, while only one-fifth had no problems accessing healthcare. According to Wayack-Pambe, few married women in Burkina Faso are empowered; only 59% of them participate in decision-making, while 61% live under psychological pressure [42]. Furthermore, we found that the odds of mDFPS increased with the presence of each component of women’s agency, especially the freedom to access healthcare. The associations between women’s agency and contraception have been studied fairly extensively and seem to be mostly positive [35, 36]. However, no study has systematically accounted for the dimensionalities and consistency of indicators of agency as related to mDFPS. Considering these parameters in our study not only strengthened the results but also revealed the struggle that MWRA have undergone to avoid unintended pregnancies. In addition, our study may have identified freedom in accessing healthcare as a contextually salient component in the relationship of women’s agency with mDFPS. However, previous conceptualizations of empowerment have considered women’s freedom of movement to be unlimited in the sub-Saharan African context [43, 44]. In fact, heath-seeking behavior is not trivial for married women, as it requires financial independence and more equal spousal power dynamics. In Burkina Faso, previous research found that married women with home ownership or bargaining power were more likely to meet their healthcare needs that were less obvious to husbands, including contraception [19, 45]. It should be noted that these components of agency are also influenced by levels of inequality in communities [42].

Adding community-level of gender equality reduced the influence of women’s agency on mDFPS especially participation in household decision-making and attitudes toward domestic violence. Previous research suggested that in communities that emphasized on traditional gender norms and relationships, perceived benefit gained from using family planning diminish [46]. In fact, it is possible that women who participate in household decisions and have equitable gender norms may more carefully consider, beyond the health risks, the social costs of losing their independence due to modern contraceptive adoption. In contrast, the lack of significant reduction on the effect of women’s access to healthcare may indicate that there is a weaker control of community gender norms and relationships on women’s health seeking behavior [47]. Conversely, Yaya et al. found in sub-Saharan Africa that women who have high decision making power and medium acceptance of wife beating still have increased odds of ever use of contraception, after controlling for community and country levels of socioeconomic factors [26]. However, the authors focused on past experience of contraception regardless of the type of methods. This approach may have overlooked the effects of contraceptive past experience on the process of empowerment. In addition, seeking modern and traditional methods of contraception may involve different levels of empowerment. Furthermore, our findings emphasized on the need to consider the effects of gender equality in communities rather than that of crude socioeconomic factors in relation with modern contraceptive use.

Consistent with socioecological theories, community-level gender inequality also shaped mDFPS. Previous studies found that discrimination against women negatively impacted modern contraceptive use, while greater opportunity for women had the opposite effect [10, 27, 28]. We found substantial variation in mDFPS across communities. Women in communities with higher fertility expectations were significantly less likely to address their family planning needs with modern methods. This finding points to African pronatalism as the most common and salient obstacle to modern contraception, as it prioritizes women’s reproductive roles and denies women the choice to limit their fertility [9, 48]. Through social conformity, women may adhere to the prevailing fertility expectations and refrain from using modern contraceptives.

Furthermore, women may conform to the prevailing fertility norms to gain social approval and avoid violence. Consistent with previous research, we did not find significant relationship between violence justification at the community level and modern contraceptive use [10]. Despite an overall negative association between acceptance of domestic violence and modern contraceptive use in African communities, within countries disparities may still exist [9]. In fact, the impact of both acceptance and experience of violence on women’s contraceptive behavior is complex to capture due to stigma which may lead to under-reporting. Further work is needed to develop tools that can capture specific and reliable elements of violence in communities.

Unexpectedly, a higher practice of female genital mutilation in the community was positively associated with the use of modern contraception. We ruled out the possibility of the ecological fallacy (Additional file 3) by controlling for individual-level genital mutilation status. As female genital mutilation is seen as an indicator of the violation of women’s rights, we expected to find a negative association with modern contraception [49]. Although the practice of female genital mutilation is discriminatory against women, women may have undergone the procedure as an act of conformity, and there is even a sense of pride in the practice in many African societies, including Burkina Faso [50, 51]. Therefore, female genital mutilation may not necessarily deter a household from using modern contraceptives. On the other hand, it is possible that in communities where the practice of female genital mutilation is widespread, the complications of pregnancies and childbirths are better understood and may have encouraged women to adopt more effective contraceptive measures [52]. Further research is required to examine the impact of other community-level influences that may mediate the association between female genital mutilation and modern contraceptive use.

Unsurprisingly, our findings on the association between community-level access to rights and resources for women and mDFPS were consistent with the current literature [9, 10, 27]. In particular, we found that greater access to assets and exposure to family planning messages in the community was also associated with higher odds of mDFPS. Our results indicate independent effects of community-level rights and opportunities for women as well as violence and discrimination against women on mDFPS. Therefore, the expansion of mDFPS should take into account both community factors, such as political changes, and cultural barriers regarding gender equality [53]. In addition, women’s access to rights and resources relative to those of men should also be considered. In fact, an increased male-to-female ratio may reflect higher gender inequality and privileging of the reproductive over the productive roles of women [27].

This study used a socioecological framework, a systematic analytical strategy, and rigorous statistical methods to assess mDFPS and its link to women’s empowerment. Nonetheless, we acknowledge several limitations. Due to a lack of data, we were unable to control for the community-level of access to family planning, which remains unexplained. Additionally, most of the empowerment indicators in the DHS questionnaire were developed in South Asia [54]. Therefore, the empowerment indicators may not adequately reflect the Burkina Faso context. Future research may examine ethnographic evidence and use more appropriate variables. However, our results are consistent with those of previous studies in a similar context. Moreover, as a cross-sectional study, this analysis does not support any inferences about causality in the relation between women’s empowerment and modern contraceptive use. A longitudinal study may enhance our understanding of the dynamic interactions between the dimensions of empowerment and socioeconomic factors. Last, community-derived gender inequality indicators are based on physical boundaries, which may not fully represent the concept of community in terms of sociocultural entity [55].

Despite the limitations, this study also has strengths. First and foremost, the nationwide survey with representative sampling allowed multilevel analyses, which provided a more comprehensive outline for understanding the influence of gender inequality on women’s reproductive behavior, especially in resource-constrained countries. This study conceptualized power dynamics at multiple levels and further identified multiple and relevant components of women’s agency in the Burkina Faso context. Such results could be useful in the future conceptualization, measurement and interpretation of empowerment. As the study population was limited to those with some interest in family planning per the selection criteria, we might have neglected the need for family planning in the non-selected sub-population. We therefore compared woman’s agency, empowerment and socioeconomic indicators between the selected and non-selected groups (Additional file 4). We found that non-selected married women of reproductive age reported significantly lower levels of participation in household decision-making, freedom in accessing healthcare and opposition of domestic violence. Moreover, we also found significantly higher proportion of non-selected women living in communities with high levels of acceptance of domestic violence, early marriage, female genital mutilation, unpaid work and fertility expectations. In contrast, non-selected women more frequently lived in communities with low levels of secondary education, and exposure to family planning messages. Women in the non-selected group were also significantly poorer, younger, less educated and more concentrated in rural areas. It is possible that these women had not perceived family planning as desirable or possible, and therefore the need for empowering women and promoting gender equality may be greater than the results of this study suggest. Last, this study provided insight into the need to adopt a multilevel and comprehensive approach to addressing socioeconomic development, gender inequality and sexual and reproductive rights.