Women's Education and Health Inequalities in Under-Five Mortality in Selected Sub-Saharan African Countries, 1990–2015

Background The aim of the study was to analyse trends in the relationship between mother’s educational level and mortality of children under the year of five in Sub-Saharan Africa, from 1990 to 2015. Data and Methods Data used in this study came from different waves of Demographic and Health Surveys (DHS) of Sub-Saharan countries. Logistic regression and Buis’s decomposition method were used to explore the effect of mother’s educational level on the mortality of children under five years. Results Although the results of our study in the selected countries show that under-five mortality rates of children born to mothers without formal education are higher than the mortality rates of children of educated mothers, it appears that differences in mortality were reduced over the past two decades. In selected countries for our study, we noticed a significant decline in mortality among children of non-educated mothers compared to the decrease in mortality rates among children of educated mothers during the period of 1990–2010. The results show that the decline in mortality of children under five years was much higher among the children born to mothers who have never received formal education—112 points drop in Malawi, over 80 in Zambia and Zimbabwe, 65 points in Burkina Faso, 56 in Congo, 43 in Namibia, 27 in Guinea, Cameroon, and 22 to 15 in Niger. However, we noted a variation in results among the countries selected for the study—in Burkina Faso (OR = 0.7), in Cameroon (OR = 0.8), in Guinea (OR = 0.8) and Niger (OR = 0.8). It is normally observed that children of mothers with 0–6 years of education are about 20% more likely to survive until their fifth year compared to children of mothers who have not been to school. Conversely, the results did not reveal significant differences between the under-five deaths of children born to non-educated mothers and children of low-level educated mothers in Congo, Malawi and Namibia. Conclusion The decline in under-five mortality rates, during last two decades, can be partly due to the government policies on women’s education. It is evident that women’s educational level has resulted in increased maternal awareness about infant health and hygiene, thereby bringing about a decline in the under-five mortality rates. This reduction is due to improved supply of health care programmes and health policies in reducing economic inequalities and increasing access to health care.

Despite these declines and progress, Sub-Saharan African countries are far from reaching the MDG 4. The Sub-Saharan African countries still record the highest under-five mortality rates [18] and over 4.4 million of these deaths are primarily caused due to infectious diseases, which can be avoided by practicing healthy habits [18]. Accounting for only 15% of the world's population, the Sub-Saharan African countries have more than 41% of children younger than five years, and countries like Nigeria, DR Congo and Ethiopia are among the top 10 countries in the world that record maximum under-five deaths. Although under-five mortality rate and malnutrition are continuing to decline in most Sun-Saharan African countries large inequalities between poor and better-off children exist both between and within these countries [19]. Evidence also shows alarming disparities in under-five mortality rates within countries. Additionally, the risks of under-five mortality increase for children born to uneducated mothers in remote areas or poor households. Survey data show that the under-five mortality rates for the poorest fifth of the population average around twice as high as the average rates for the richest fifth. These inequalities, which appear to be widening, call into question the strategies for child mortality reduction relied upon until date [19].
The variables that are used to measure health inequalities in developing countries include educational [20,21] level of the mother, socioeconomic status [22,23] and place of residence (urban vs. rural) [24]. While recent studies have focused on inequalities related to socioeconomic status, very few recent studies have examined inequalities in under-five deaths and maternal education level. Most of these studies were conducted in the 1980s and early 1990s, during which several Sub-Saharan African countries witnessed the introduction of educational reforms that aimed at increasing the educational level of women and girls. The studies of Caldwell [25,26] and Hobcraft [27] revealed the significance of maternal education in determining of child health, spurring extensive studies in developing countries on the subject. These latter studies also confirmed maternal education to be a very strong and consistent predictor of reduced child morbidity and mortality. The existing literature highlights that female education has been a key to reducing infant and child mortality and fertility rate; increasing consumption of health services; and improving socioeconomic status [28]. Additionally, maternal education increases a mother's awareness of good health care practices, children's illnesses, and availability of health services, and hence is considered as a factor that may positively affect a child's health. The authors [27,28] have argued that education could change attitudes and behaviours of women, thereby enabling them to attain greater autonomy and efficiency. The authors also emphasise on the fact that educated mothers have a greater ability to identify healthcare services for treating their child's illnesses; higher receptivity to new health-related information; familiarity with modern medical culture; access to financial resources and health insurance; better decision-making power; and increased self-worth and self-confidence.
Several empirical studies [25,26,[29][30][31] have confirmed the positive effects of mother's education on a child's survival. However, the results of some studies are mixed. A few studies show a very low impact of education on child's health, while other studies have shown that there is no link between the educational level of mother and a child's health [32]. This finding implies that although mother's educational level influences the health and survival of the child, the effect may vary from one country to another and from one context to another. In investigating the pathways of influence, research confirms that the causal linkages between maternal education and child's health are far from clear, and the relationship between these two factors is simply not a reflection of a co-occurrence of education with other socioeconomic variables [27,32,33].
Using data from selected Sub-Saharan countries of several birth cohorts, this study aimed to analyse trends in the relationship between the mother's educational level and inequalities in under-five mortality over the past two decades in Sub-Saharan African countries (1990 to 2015). This study also aimed to evaluate the effect of maternal educational level on the mortality rate of children under five years of age. The study hypothesises that inequalities in underfive mortality related to mother's educational level has declined over the last two decades in Sub-Saharan Africa, and the gap in the risks of under-five deaths between children born to best educated mothers and children born to less educated mothers is narrowing other time.

Data
Huge progress has been made over the last decades to improve the literacy status of adult women around the Sub-Saharan Africa [34]. While gender disparities in adult literacy rates remain wide in Sub-Saharan Africa due to many factors including culture and lack of infrastructures, evidences indicate that some progress has been made at regional and country levels to improve the literacy level of women [34]. Fig 1 presents the trend of female primary education completion rates in the countries selected for this study. The results show that although the level of women's education has greatly improved during the period 1980-2010, there is a slower increase in the West African countries. The results show that the rate of increase in female education has been 30.7% in Niger, 34.3% in Burkina Faso and 47.5% in Guinea, for the period 2005-2010, compared to other countries that record an increase of above 56%.
Data used in this study come from different waves of DHS of Sub-Saharan countries ( Table 1), including Burkina Faso, Niger, Guinea, Cameroon, Congo, Malawi, Zambia, and Namibia. The countries were selected according to a sub-regional basis selecting at least two countries from each sub-region of the Sub-Saharan Africa (Western Africa, Central Africa, Eastern and Southern Africa). The selected countries have already realised at least two DHS rounds in the period 1990-2015.
First, DHS waves of survey data of each selected country were pooled, and then data from all the countries were pooled for the purpose of analysis. DHS data were collected using a standardised questionnaire, which was used in all the countries and for different waves of DHS. This offers an advantage for data analysis and comparison of results across countries. This study used birth history data from DHS. Table 2 given below presents the number of children under five years and the survival status of each birth cohort by country. The birth history (birth cohort) dataset contained information on the date of birth of all the children born to a woman during her lifetime, starting from the first child to the total number of children born at the time of the survey. Additionally, information on child's survival (dead or alive) was available in these datasets [35]. Birth histories were collected from a sample of women aged 15-49, at the time of the survey. The complete birth histories (including date of the birth and survival status of each child born to these women until the time of survey) of women aged 15 to 49 years old is considered to be useful data for computing child mortality indicators. DHS data are cross-sectional data, and therefore the survey data represents the entire population. In this context, the information collected during the study included demographic, socioeconomic, and maternal and child health data. The sample design of these surveys is a nationally representative sample and a stratified two-stage cluster design.

Ethics statement
The datasets used in this study were obtained from the DHS program thanks to the authorisation received to download the dataset on the website (https://dhsprogram.com/data/availabledatasets.cfm).

Variables
The dependent variable of this study is the child survival status (alive or dead) during the duration of the survey. The wealth indexes, the mother's educational level, and the place of residence (rural/urban) are the main explanatory variables that assess the inequalities in child mortality. Households were grouped into five categories of wealth index (poorest, poor, middle, rich and richest). In terms of maternal education, we categorised the mothers into three groups (not attended school, 1-6 years of education and more than 6 years of education). The other covariate variables included sex and birth order of the child, parity, mother's age during childbirth and the size of the household.

Statistical methods
Descriptive and multivariate analyses. First, the study used concentration index as well as absolute and relative ratios of mortality rates to measure inequality in health. The concentration index ranges between -1 and 1, and a negative value in this study means the deaths are more concentrated among children born to less educated mothers. Two logistic regression models were built to estimate the effect of mother educational level on the probability of a child to die before reaching his fifth birthday: Model 1 gave the unadjusted effect and Model 2 gave the adjusted effect of the co-variables.
Decomposition analysis. This study used the method proposed by Buis [36] which decomposes the total association between a categorical, discrete or continuous exposure variable, and an outcome from a direct effect and an indirect effect. The decomposition method proposed by Buis is described in his article published in 2010. In the current study, it is assumed that the dependent variable Y represents the child's mortality and X represents the mother's educational level. X is the main dependent variable by which we seek to quantify the direct effect, and Z is the indirect effect of the effect of all other covariates in our study. OR is considered as the risk of child to die before reaching his fifth birthday.

Results
Trend of mother's educational level and under-five mortality inequalities The results reveal that, in general, children of mothers who did not attend school have a higher rate of death compared to mothers with formal education. However, mortality rate differentials are reduced from the older birth-cohort of children (1986)(1987)(1988)(1989) to the younger birth-cohort of children in each country. The results also show a variation among the countries (Fig 2). Cameroon, Burkina Faso, Niger and Guinea are countries showing large inequalities in mortality by education level of the mother. Indeed, in these countries, the concentration index is less than -0.  Mother's educational level and reduction of under-five mortality Table 3 presents the comparative results of the under-five mortality rates of birth cohort according to different levels of mother's education. The results show that the decline in mortality of children under five years was much higher among the children of mothers who had not been to school. The findings show a 112 points drop in Malawi, over 80 in Zambia and Zimbabwe, 65 points in Burkina Faso, 56 in Congo, 43 in Namibia, 27 in Guinea and Cameroon, and 22 to 15 in Niger. With regards to mothers with 1-6 years of education, the declines in mortality rates for children under five years were less compared to the groups of children born to non-educated mothers. The findings reported significant differences in under-five mortality comparison between the older birth cohort and younger birth cohort. These differences were 125 points for Malawi, 91points for Zambia, 54 points for Burkina Faso, and 40 points for Congo. However, the study reported an upward trend for Cameroon (-42 points), Niger (-11 points), and Guinea (-8) points. A downward trend was noticed for children born to educated mothers. There was a significant difference in decline in under-five mortality rates between old birth cohort and the young birth cohort 104.4 points difference in Malawi, more than 40 points difference in Zambia and 24 points in Congo. The declines were not significant in Burkina Faso, Niger, Guinea, Cameroon and Namibia. Tables 4 & 5 shows the multivariate results. The study aimed to examine the gross effect of maternal education on child mortality, and subsequently obtain the net effect after controlling In all these cases, children of more educated mothers are not likely to die before the age of five (OR = 0.4 in Guinea and in Niger, OR = 0.5 in Burkina Faso and Cameroon, OR = 0.7 in Malawi and Congo, and OR = 0.8 in Zambia and Namibia) compared to children born to mothers without any formal education.

Effects of mother's educational level on inequalities in under-five mortality
With regards to Burkina Faso (OR = 0.7), Cameroon (OR = 0.8), Guinea (OR = 0.8) and Niger (OR = 0.8), children of mothers with 0-6 years of education are about 20% more likely to survive until their fifth birthday compared to children of non-educated mothers. There are no significant differences between children of mothers with 0-6 years of education and those of mothers who have not been educated who die before their 5th birthday in Congo, in Malawi and in Namibia.
The results (Model 2 in Table 5) depict the net effects (the adjusted OR after controlling the confounding variables. As in the case of Model 1, the results vary by country. With the exception of Namibia and Congo, where the mother's educational level loses its significance, the mother's educational level is significantly associated with under-five mortality in other countries. While there is a 50% to 20% likelihood that children born to educated mothers might not die before their fifth birthday, it might not be the case for under-five children of mothers without any formal education. Table 5 shows that although the effect of maternal education declined in Burkina Faso while controlling the effect of the covariate, the impact of maternal education on under-five mortality still remains significant. However, there is no significant difference in mortality between children of mothers with 1-6 years of education and children born to non-educated mothers. Variables such as the birth cohort, maternal age during childbirth, the birth interval, multi-births, the child's sex, and place of residence are proved to be significantly associated with the mortality of children under-five years. Children born between 1990 and 1999     3)) are more likely to die before their fifth birthday than female children and those living in urban areas. With regards to Niger, the results of the final model show that maternal education has dropped slightly but still remains significant after the introduction of control variables. It appears that children born between 2000 and 2009 (OR = 0.9, 95%CI (0.8, 1.0)), and post 2009 (OR = 0.5, 95%CI (0.4; 0.5)) were less likely to die before the age five compared to children born between 1990 and 1999. Also, the risk of mortality reduced for children born to f mothers who delivered between 20-34 of age (OR = 0.7, 95%CI (0.6, 0.8)) and maintained more than 24 months of birth interval. However, this was not the case for children born to young mothers ( 19 years) and children of first-time mothers. The results also show that, compared to twins, non-twins (OR = 0.2, 95%CI (0.2; 0.3) had about 80% less risk of dying before reaching their fifth birthday. Like mother's educational level, the variables that were significantly associated with underfive mortality rate in Guinea included birth cohort, parity, and maternal age at delivery, interbirth interval, and type of birth (twin vs. single), the child's sex and socioeconomic status. The observed relationships are similar compared to those found in Burkina Faso and Niger. In Cameroon, the variables significantly associated with mortality were birth cohort, parity of the mother, the mother's age at delivery, inter-birth interval, type of birth, the sex of the child and, the socioeconomic status. While in Congo, birth cohort, the type of delivery, and socioeconomic status were significantly associated with the mortality of children under five years. In   Malawi and Zambia, in addition to the mother's educational level, the birth cohort, inter-birth interval, the type of delivery, and the child's sex were significantly associated with under-five mortality. The maternal age at birth, birth interval, multi or single birth, socioeconomic statuses were determinants of mortality in children under five years in Namibia.
Direct and indirect effects of mother's educational level on under five mortality Table 6 and Fig 3 show the decomposition results of the direct and indirect effect of mother's education on under-five mortality. The results show that the direct effect of mother's educational level varies between 35% (in Namibia) to 79% (in Congo). The direct effect of mother's educational level is not significant in Namibia and Congo. In the case of Namibia, the education of the mother is not a determinant of mortality in children under five years of age. However, more than three-fourths difference in mortality was observed among children of educated mothers, with seven years of education, and children of non-educated mothers. The differences accounted were 78.3% in Burkina Faso, 73.6% in Malawi and 71.5% in Guinea. This difference is greater than half for Zambia (66.0%), Cameroon (63.7%) and Niger (52.3%). These differences could be due to the direct effect of the educational level of the mother.

Discussion
The study results clearly show the influence of maternal education on under-five mortality while decomposing the direct and indirect effect of maternal education on mortality of children under five years. The results of our study have shown that the trend of mortality rates by level of mother's education varies from one country to another and we did not get the same trends in all the countries. The trend of under-five mortality by level of mother's education is similar in Burkina Faso, Niger and Guinea high mortality among children of non-educated mothers, low mortality risk among children of educated mothers (!7 years schooling), and an intermediate position for children of mothers with 1-6years of education. The trend for other countries is not consistent with this finding. The heterogeneity of the results reveals the importance of unobserved factors which may affect maternal education and child health simultaneously [37]. Additionally, the inconsistency in findings also highlights the importance of contextual effects for explaining the demographic events such as under-five mortality [38,39]. These include the health policies implemented across selected countries; the presence, access and affordability of healthcare services; the availability of community services; a mother's health status and awareness.
Country specific results confirm the existence of a difference in mortality rates between children of mothers who have been to school and children of mothers without formal education. However, these results varied across countries. The results confirm an inverse relationship between mother's educational level and the risks of under-five mortality [26,27,40]. With the exception of Namibia, the results showed that children of educated mothers (seven years or more of education) are less likely to die before reaching their fifth birthday compared to children of less educated mothers. These results are in line with the results of previous studies on infant mortality in developing countries [29,30,40,41].
Our study also analysed the trend of the under-five mortality of birth cohorts of children born between 1986 and 2013. The results showed a significant decline in mortality among children of mothers without formal education (between 14 points in Niger and Malawi 112 points), while the decrease was less pronounced among children born to mothers with 1-6 years of education (-42 points in Cameroon to 124 points in Malawi), and among children of mothers with seven or more years of schooling. The largest decline among children of non-educated mothers could be explained by the reduction of geographical and financial barriers to accessible healthcare for children under-five year of age in many countries.
The remoteness of health centres is identified as an unfavourable factor for accessing to healthcare services for children, especially in rural areas [42][43][44][45]. The current study also identifies the need to reduce delivery costs of healthcare services for promoting the use of maternal and child services, especially for children from poor households [46,47]. The multivariate results, after controlling for other variables, showed that the effect of educational level of the mother is an important determinant of child mortality in several countries. These analyses also showed the importance of variables such as the generation of child birth, parity of the mother, the mother's age at delivery, inter-birth interval, the type of birth, the sex of the child and the household socioeconomic status. Our study has some limitations in addition to the limitations often assigned to DHS data. These limitations include event omissions; misreporting of ages and dates of birth and death, especially, for new-borns; and selectivity bias owing to the lack of DHS data on survival or death of children born to mothers who had died at the time of the survey. Additionally, mother's educational level variable represents only the educational level of mothers at the time of the interview, which cannot be reflective of mother's education at the time of childbirth. It is therefore possible that the educational level of a mother during childbirth might be slightly different than her educational level at the time of the survey. However, we believe that the bias is relatively small and will not impact the results. In this study, we did not pretend to take into account all the determinants of child mortality. Therefore, the study does not consider education level of parents, ethnicity, religion and administrative region, assistance at delivery, and complete or uncertain vaccination.

Conclusions
The study demonstrates that mothers' educational attainment is an important determinant of child mortality, and the results confirmed that children of more educated mothers have lesser risk of dying before their fifth birthday compared to children born to mothers without formal education. The analysis of the trend in mortality educational level of the mother showed a significant reduction in mortality gaps between children of mothers with more than seven years of education and children of non-educated mothers. These observed reductions are certainly the result of programmes and policies designed for reducing unequal access to health care and improving the supply of health care, especially, to the most underprivileged communities. If universal access of women to school could induce a reduction in under-five mortality rates in Sub-Saharan countries, it is clear that it should be accompanied by access to policies and healthcare programmes, which can reduce the gaps between children of educated mothers and non-educated mothers.
Supporting Information S1 File. Demographic and Health Surveys data agreement. (DOCX)