Women’s Status and Empowerment within the Household as a Determinant of Adequate Antenatal Care Visits in Malawi. A Population-Based Study


 Background Research has linked women status and empowerment to an improved health status of the general population. However, there is a deficiency of pragmatic data to support how females’ position within the household may impact adequate antenatal care (ANC) visits in Malawi. Therefore, this study explored whether the women’s status and empowerment influence adequate ANC visits in Malawi.Methods The current study sampled and analyzed ever-married women of reproductive age from the 2010 and 2015–16 Malawi Demographic and Health Surveys (MDHS). The two-level multilevel multivariable analyses were performed by means of the generalized linear mixed models (GLMMs) to investigate the relationships between the women’s status and empowerment within the household and adequate ANC visits while controlling for covariates.Results About 46% and slightly over 50% of participants had adequate ANC visits in 2010 and 2015–16 respectively. The GLMMs shows that women with no formal education (adjusted odds ratio [aOR]: 0.71; 95% confidence interval [CI]: 0.54–0.94) and primary education (aOR: 0.65; 95% CI: 0.53–0.80) had reduced odds of achieving adequate ANC visits. Additionally, younger women, women from the poorest households, women who did not have access to media and women who resided in the northern region had reduced odds of achieving adequate ANC visits. The GLMMs also revealed variations in the communities with respect to the odds of achieving adequate ANC visits.Conclusion The association of women’s education on adequate ANC visits underscores the need for policymakers’ to think through incorporating education sector in ANC settings to increase health-related knowledge among women of reproductive age in order to inform them of the importance of achieving adequate ANC visits.

95% con dence interval [CI]: 0.54-0.94) and primary education (aOR: 0.65; 95% CI: 0.53-0.80) had reduced odds of achieving adequate ANC visits. The GLMMs also revealed variations in the communities with respect to the odds of achieving adequate ANC visits.
The association of women's education on adequate ANC visits underscores the need for policymakers' to think through incorporating education sector in ANC settings.

Background
Appropriate care throughout the pregnancy is essential for the better health outcomes for the woman and her children. Usually, maternal and child health outcomes are used as measures to describe the overall health status of the general population [1] [2]. According to the World Health Organization (WHO), about 830 women are dying each day from pregnancy-related complications worldwide [3]. In 2015, approximately 303,000 women died in the course and following pregnancy and its outcomes.
Unfortunately, a majority of these deaths happened in low-income countries [4]. Globally, over the past two decades,  maternal death fell by around 44% i.e., from 385 to 216 deaths per 100,000 live births [3][5] [6]. In Malawi maternal mortality has been dropping steadily. For instance, statistics taken from the population-based surveys conducted in Malawi indicated that pregnancy-related mortality has dropped from 1,123 deaths per 100,000 live births in the early 2000s to 439 deaths per 100,000 live births in 2016. In the year 2010 which is the year before the 2015-16 Malawi Demographic and Health Survey (MDHS), the maternal mortality was estimated to be at 675 deaths per 100,000 live births [7] [8]. Despite this steady decline in maternal mortality, Malawi remains one of the countries with high maternal mortality rates in the sub-Saharan Africa region. This is true as the United Nations Children Fund (UNICEF) stated that the sub-Saharan African region still has excessive levels of maternal mortality than the rest of the world [6]. Speci cally, over the last two decades, maternal mortality in sub-Saharan Africa has slightly dropped from 987-546 deaths per 100,000 live births [6].
Maternal mortalities are by and largely preventable since strategies to avert or manage pregnancy-related complications are well recognized [3][5] [9]. The WHO reported that about three-quarter of the bulk of the maternal mortalities is predominantly caused by pre-eclampsia and eclampsia, infections, severe bleeding, complications from delivery, and unsafe abortion [9]. Additionally, the WHO stated that the remaining proportion of mortality is mostly due to infections in the course of pregnancy such as malaria, and acquired immunode ciency syndrome (AIDS) [5] [9]. Antenatal care (ANC) has been reported to be effective in the reduction of maternal perinatal morbidity and mortality. For example, women who have pregnancy-related complications (placenta previa, gestational diabetes, eclampsia, and eclampsia etc.,) are detected and treated appropriately. Furthermore, the ANC provides the medium through which a number of conditions are assessed and diagnosed such as fetal growth, human immunode ciency virus (HIV), sexually transmitted infections (STIs) and hepatitis. Likewise, mothers are also provided with tetanus toxoid vaccination and intermittent preventive treatment for malaria during pregnancy (IPTp). Moreover, women are also encouraged to practice some of the better health behaviors such as early initiation of breastfeeding and vaccination [10]. Previous studies have emphasized the effects of individual-level factors such as maternal age, maternal education level, birth rank, parity, self-employment, age at 1st delivery, obstetrical history [11] [12] [13] and household characteristics including, religion, caste, type of families, family size, urban residence, exposure to media [12] [13] [14] [15]. However other in uences such as women status, empowerment, and neighborhood level characteristics have rarely been explored.
The status of women in the household is a vital in uence on the development, well-being, and enhancement in the standard of living [7]. Women's status and empowerment have been frequently de ned as the socio-economic position of women in the society and their capability of making informed choices as well as getting to grips with matters that affect their well-being [16]. Women's status and empowerment encompass education and employment status, participation in decision-making, control over their own earnings, and attitudes towards intimate partner violence [7]. Prior researchers have reported the association of the components of women status on health care utilization. For instance, in Nigeria, women's decision making on own health and an employment annually were associated with the higher likelihood of getting skilled ANC, whilst those who con rmed physical intimate partner violence were less likely to receive both skilled ANC and delivery services [2]. Similarly, women who had a discussion of family planning with their spouses had an increased chance of getting skilled ANC and delivery care meanwhile women who had secondary and above education were strongly associated with the use of health care in Nepal [17].
In Malawi, over the last two decades, the use of ANC has been unsatisfactory. For example, the 2000 MDHS reported that 56% of women had the adequate number (4 and above) of ANC visits [18]. In 2004, 57% of women had an adequate number of ANC visits [19], whilst in 2010 was 45% and 2015-16 was 50% respectively [7] [20]. It is commonly emphasized that an equality in the household is essential for better maternal health [21]. In most sub-Saharan African countries, the poor socioeconomic status of women is recognized as a burden to the progress towards health and population policy targets of the nation [2]. Even though it is highly anticipated that the better equality within the household may lead to the better use of maternal health care services. Unfortunately, there is no empirical evidence to prove these hypotheses among Malawian women. Thus, the current study aimed to explore whether the women's household status and empowerment in uence adequate ANC visits in Malawi while controlling for a wide range of covariates both at the individual and community level.

Data source
The current study utilized two cross-sectional datasets taken from the 2010 and 2015-16 Malawi Demographic and Health Surveys (MDHS). Comprehensive procedures used in these surveys can be found in details elsewhere [7] [20]. Brie y, the sampling frame used in these surveys was taken from the Malawi Population and Housing Census (MPHC) which was conducted in Malawi in 2008. This census sampling frame contained a complete list of all census standard enumeration areas (SEAs). The MDHS produced nationally representative samples using strati ed two-stage cluster design. The rst stage was employed to select 849 SEAs (158 in urban areas and 691 in rural areas) in 2010 and 850 SEAs (677 in rural areas and 173 SEAs in urban areas) in 2015-16 . The second stage was employed to select a  systematic sample of 27,345 and 27,516 households in 2010 and 2015-16 respectively. The surveys interviewed 23,020 respondents in 2010 and 24,562 respondents in 2015-16 representing response rates of approximately 97% and 98% respectively. Using personal interviews, women of reproductive age who had children less than 5 years before the surveys were interviewed on the measures of socioeconomic, demographic, environmental, population health, anthropometry etc. The current study focused only on married women and excluded missing cases from independent variables. Final samples analyzed were 4253 for 2010 MDHS and 2612 for 2015-16 MDHS. Figure 1 shows

Independent Variables
To understand women status and empowerment as determining factors of adequate ANC visits in Malawi, the modi ed framework of health-seeking behavior established by Andersen and Newman [22] combined with the framework on women status and empowerment was used [23] Fig. 2 presents the modi ed framework. Women status and empowerment are two confusing terms and at times used interchangeably. Women's status is described as an overall position of the women in the general public. The term includes women's educational status, cultural beliefs, economic position, etc., in a given population [16]. On the other hand, women's empowerment is commonly de ned and described using proxy procedures. These proxies include access/control over household income; women's involvement in household decision-making; perceptions of gender norms; and perceived equity in their power and resources [16] [24]. Prior literature has considered women's empowerment as the process through which those who have been left without the capacity to make informed life preferences obtain such capability comprising of several concepts, such as the capacity to access and make use of resources, the ability to move, agency, and autonomy [24].

Main Independent Variables
The main independent variables included in the current study were; women's educational level (no formal education, primary education, and secondary and above education); decision on how to spend money (respondent alone, respondent with other persons, and other persons); partner's earnings (wife > husband, wife < husband, wife = husband, and don't bring cash); nal say on own health care (respondent alone, respondent with other persons, and other persons); nal say on making large household purchases (respondent alone, respondent with other persons, and other persons); age at rst marriage in years (< 18, 18-24, and ≥ 25 years) and wife or husband's attitude toward wife beating (whether justi ed or unjusti ed). Attitude toward wife beating was created from ve items on whether the husband was justi ed in beating his wife that is if the woman had disregarded the rules on the established gender roles in the household. The ve situation that permitted or justi ed wife beating were: [1] if wife neglects the children, [2] if the wife goes out without telling the husband, [3] if the wife refuses to have sex with husband, [4] if the wife argues with the husband, and [5] if the wife burns the food. Justi cation of wife's beating was calculated by summing up the ve items. Speci cally, the current study focused on the percentage of all married women of reproductive age who agreed that a husband is justi ed in beating his wife with at least one speci ed reason.

Control Variables -Individual-level Factors
Individual-level variables included; age of women in years (15-24, 25-34, and 35-49); religion (Roman Catholic, Protestant, and Other religion); husband's educational level (no formal education, primary education, and secondary and above education); sex of the household head (male or female); household wealth index (poorest, poorer, middle, richer, and richest); media exposure (media exposure or no media exposure). The wealth index was created using information on the possession of a number of household's assets. Generated through the procedure known as principal component analysis (PCA) by MEASURE DHS household assets were assigned scores. The procedures on how to derive the wealth index have been described in details elsewhere [25]. The media exposure was classi ed based on the frequency the women accessed newspaper/magazine, listened to the radio or watched television. A composite measure for mass media was created by combining access to newspaper/magazine, radio or television (less than once a week/at least once a week/almost every day).

Control Variables -Community-level Factors
To de ne the clustering effects on women living within the similar geographical environments, the term community was used. Community-level characteristics were built by grouping maternal-and householdlevel data to the higher level. Speci cally, a community was classi ed based on the sharing of the same primary sampling unit (PSU) within the DHS data [26]. The PSUs are habitually created from enumeration areas identi ed and used in a preceding national and housing census. The geographical regions namely northern, central and southern as well as the place of residence namely urban and rural were considered as the community residence. Additionally, the four aggregated variables were considered and assessed the community female education, wealth, media exposure, and distance to the nearest health facility. The community wealth was de ned as the proportion of households in the community considered as the richest (upper 40% of quintiles). The community female education was classi ed as a proportion of women in the community who had a secondary and post-secondary education. Furthermore, the community exposure to media was de ned as a proportion of households in the community who had media exposure whilst the community distance to the nearest health facility was described as a percentage of household in the community who perceived distance to the health facility as a big problem. All the aggregated variables were characterized as "low", "medium" and "high" (tertiles) in order to provide the results that are more easily understood in the development of health-related policies.

Statistical analysis
Firstly, the baseline statistics for all the categorical characteristics were carried out and presented as frequencies and percentages. Secondly, bivariate analyses were conducted by means of Pearson Chisquare (χ2) to examine the probable factors that are independently associated with the woman achieving adequate ANC visits. A Cochran-Armitage test for trend was applied where necessary to test for trend within the ordered categories of the variable. Pearson correlation test was conducted to assess multicollinearity among continuous community factors. Thirdly, the multivariate analyses were conducted using two-level multilevel logistic regression. The generalized linear mixed models (GLMMs) were tted to concurrently estimate the effects of women's status and empowerment on adequate ANC visits while controlling the in uence of individual-and community-level covariates. The GLMMs models with the logitlink function and binomial distribution were speci ed in the analysis owing to the binary nature of the dependent variable. Because of the complex sampling technique, the GLMM was also used in the current study in order to take into account the clustering, and strati cation in the survey design [27]. All of the analyses were performed using SAS version 9.4 (SAS Institute Inc., Cary, NC, USA). Five models were constructed for each cohort year. The rst model (null model) had no predictors. The model was exclusively used to calculate the total variance in the community without taking into any independent variables. Second model comprised of women's status and empowerment namely; women's educational level, decision on how to spend money, partner's earnings, nal say on own health care, nal say on making large household purchases, age at rst marriage, and wife or husband's attitude toward wife beating. Third Model contained women's status and empowerment within the household and individuallevel factors including, age group of women, religion, and husband's educational level, sex of the household head, household wealth index, and media exposure. Fourth model, contained women's status and empowerment within the household and community level factors namely, geographical region, place of residence, community wealth, women's education, media exposure, and distance to o cial health facility. Fifth model controlled for women's status, empowerment within the household, individual-and community-level factors. Due to the multilevel nature of the statistical technique employed in this study, all models contained a random-intercept xed-slope. The xed effects (measures of association) between the women's status and empowerment within the household and adequate ANC visits were reported as adjusted odds ratios (aOR) 95% con dence interval 95% (CI) with their p-values and after considering possible confounders. Alpha level of less than 0.05 was used to decide the statistical signi cance in the models. The possible random effects (measure of variation) were examined by the Median Odds Ratio (MOR), Area variance (AV), Intra-Cluster Correlation (ICC), and Proportion Change in Variance (PCV). The techniques used for calculating MOR and PCV had been described in detail elsewhere. The ICC was calculated by the linear threshold according to the formula used by [28] which represent the proportion of variance at the group level divided by the sum of the variances at the individual and group levels while the MOR quanti es the variation between clusters by comparing two persons from two randomly chosen, different clusters. The xed and random effects presented in the result section are those from the nal models (Model 5). The Deviance Information Criterion (DIC) was used to assess the appropriateness of the models. The multicollinearity was tested by the use of variance in ation factor (VIF). All VIF values were < 10 suggesting no multicollinearity problems (supplementary table).

Ethics Statement
The National Statistics O ce (NSO) implemented the data collection. The procedures for the survey were approved by the Malawi Health Sciences Research Committee (NHSRC) and the Institutional Review Board (IRB) of ICF Macro. ICF IRB ensured that the survey complied with the U.S. Department of Health and Human Services regulations for the protection of human subjects (45 CFR 46), while the NHSRC ensured that the survey was conducted in line with Malawian laws and norms [29]. At the beginning of each interview, participants were asked to give verbal and written consent (informed consent). Informed consent statement emphasizes that participation was voluntary and that the respondent may opt to refuse to answer any question. The DHS program provided the permission for the use of the data above and beyond the primary purpose it was collected.

Results
Population characteristics Table 1 presents the baseline characteristics of the study participants. A total of 4253 married women were analyzed in the 2010 MDHS and 2612 married women were analyzed in the 2015-16 MDHS. About 46% of participants had adequate ANC visits in 2010 whilst slight above 50% of the participants had adequate ANC visits 2015-16. In both cohort year, over 50% of participants had primary education, the respondents were earning less than their partner's. Furthermore, in both cohort years, the respondents had problems with decision making about their own health as well as on the purchase of large household items. A majority of respondents in 2010 were less than 18 years when began cohabitating with their partners. And in both cohort years, respondents reported being unjusti ably beaten. Prevalence Of Adequate Anc Visits Table 2 displays the distribution of adequate ANC visits across the main explanatory variables as well as covariates. In 2010, the prevalence of adequate ANC visits was observed to be highest in women with secondary and above education, in women who made a decision on how to spend money with other persons, in women who were 25 years and above at their rst marriage. Furthermore, the prevalence of adequate ANC visits was highest among Protestant women, in women whose partners had secondary and above education, in women who resided in the richest household and had access to mass media. On the community level, the prevalence of adequate ANC visits was highest in women from central and urban dwellers. In addition, the prevalence of adequate ANC visits was highest in women resided in communities with a high proportion of household regarded as richest, high female education, high proportion of household exposed to media, and a low percentage of women perceived the distance to nearest health facility as a big problem. In 2015-16, the prevalence of adequate ANC visits was observed to be highest in women with secondary and above education, in women who made decision on own health care with other persons, in women who made nal say on making large household purchases with other persons, and in women who were 25 years and above at their rst marriage. Furthermore, the prevalence of adequate ANC visits was highest among women aged 35-49 years, in women whose partners had secondary and above education, in women who resided in the richest household and had access to mass media. On the community level, the prevalence of adequate ANC visits was highest in urban dwellers, in women resided in communities with a high percentage of household categorized richest, high percentage of women with a secondary or higher education, high proportion of household exposed to media, and a low percentage of women perceived the distance to nearest health facility as a big problem. Table 3 revels the results of Pearson correlation among the community-level factors. The correlation was strong and positive between the community female education and wealth. Furthermore, the correlation was moderate and positive between community media exposure and community wealth, community media exposure and community female education. Whilst the correlation was moderate and negative between a community with perceived distance to the nearest health facility with community wealth, community female education, and community media exposure. Both cohort years exhibited similar patterns of results.

Discussion
The current study employed a multilevel technique on the nationally representative data to explore whether the women's status and empowerment within the household in uenced adequate ANC visits in Malawi. It has been demonstrated already that proper care in the course of pregnancy is essential for the better health and well-being of the mother and the growth of the fetus [1]. Hence, an operational continuum of care which is easy to access and contains high-quality care before the pregnancy and beyond is necessary for the prevention of pregnancy-related complications [30]. The [16], the current study also found that participants with secondary and above education had an increased odds of having an adequate ANC visits. In a society, an accomplishment of formal education by women is mostly used as the measure of women's socioeconomic position [16]. Several mechanisms have been hypothesized by which maternal education affect health-seeking behavior. The mechanisms include increased awareness on the signi cance of health services and improved capacity to opt for the most suitable services [14]. Maternal education is strongly associated with improved autonomy, different views on traditional norms and beliefs, as well as ability to have power over resources within the household [31] [32]. Furthermore, better-quality education enables women to understand health-related information more swiftly [32]. Prior literature elucidates that women with improved education are more likely to live in developed areas such as cities, urban areas etc. where availability, accessibility, and affordability to the health facilities maybe not a big problem [33].

Individual level factors
Apart from women status within household, some of the individual control variables were signi cantly associated with adequate ANC visits. The present study found that younger women and women from the poorest households were less likely to achieve adequate ANC visits. Previous studies have also reported the positive effects of better household wealth and the use of health services including ANC [10] [34]. For example, in Colombia, it was reported that women from wealthier households had increased likelihood of attending adequate ANC than poorer counterparts [35]. Additionally, due to better socioeconomic position, women from richer households' maybe residing in developed areas where resources may be easily to be accessed [35] compared to poor counterparts. As regards age, younger women may not be able to achieve adequate ANC visits because they may not have prior knowledge or experience as regards pregnancy. It was reported that the use of ANC may be in uenced by complications experienced during earlier pregnancies [34].
Community level factors.
In addition to women status and individual level factors, the current study revealed that community level factors such as geographical region and media exposure were also other determinants of adequate ANC visits. Women from central region of Malawi and were exposed to media had higher chances of achieving adequate ANC visits. Previous research reported that region inequalities in health outcomes may be as the result of geographical location, economic, political will, environmental and various sociocultural norms and values [32]. Factors like the distance and transportation to the health facility may be much easier for the developed region as the result the use of the health care services may be affected [36]. On the other hand, media is frequently regarded as a reliable source of health-related information whereby the key messages such as maternal health care utilization may be addressed [37]. Thus, it is expected that women who have access to media may have a high chance to achieve adequate ANC visits.

Contextual in uences
The ndings of the current study also revealed evidence of variations at community levels. This suggests that women from same neighborhood are subject to common contextual in uences. In both cohort years 56% of the total community-level variance were explained by the women status and empowerment-, individual-, and community-level factors. However, the signi cant variations were observed in 2015-16.
Regarding the ICCs, about 2% and 4% of the total variance of adequate ANC visits remain unexplained even if the women status and empowerment-, individual-, and community-level factors were taken into account. Prior researchers on multilevel analyses have reported that any ICC of at or above 2% indicates signi cant group-level variation [38].

Strengths And Weaknesses
This was the rst multilevel study to simultaneously incorporate women status and empowerment together with individual-and community-level factors as such the issue of the methodological problem was addressed. The use of the nationally representative sample makes these ndings generalized in Malawian context. However, the weakness of the current study was inability to draw causal inferences due to cross-sectional nature of the study design as well as in ability to incorporate other covariates that could have explained ANC utilization owing to the use of already collected data.

Policy Implication
The present study reveals that education as the measure of women's status was an essential factor associated with the use of adequate ANC visits in Malawi. Policy makers in maternal and child health need to consider incorporating education sector in ANC settings. Furthermore, the signi cant variation in the log odds of adequate ANC visits in the community in 2015-16, warrant policymakers not only to target individual needs but also their communities. This could be by increasing health facilities in communities so as to reduce the distances women travel to access health services. Moreover, awareness regarding the bene ts of adequate ANC visits should be emphasized since the current study revealed that communities with low percentage of house exposed to media were less likely to achieve adequate ANC  [7]. Before this study was conducted, the authors sought permission from the MEASURE DHS for use of the data beyond the primary purpose by which data were collected. At the beginning of each interview, informed consent (both written and oral consent) was obtained from all eligible participants.

Consent for publication
Not applicable Availability of data and materials The datasets generated and/or analyzed during the present study are available in The DHS Program repository, https://dhsprogram.com/data/available-datasets.cfm

Competing interests
The author declares that he has no competing interests Figure 1 shows a owchart of the sample inclusion and exclusion criteria. Figure 2 shows the modi ed framework for women empowerment and health care utilization.

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download. TableSuppliment.doc