Maternal education is an important influence on child survival, growth and development [1, 2]. Education is an underlying determinant of dietary intake and care practices, and, if improved, helps protect children at risk of poor health and nutrition in low- and middle-income countries [3]. For many studies evaluating the relationship between maternal education and child health and nutrition, the main outcome of interest has been either child mortality or child stunting (low height-for-age) [2, 4–12]. For example, in a recent analysis of Demographic and Health Survey (DHS) data from 19 countries, Ruel and Alderman found that the risk of stunting was significantly lower among children of mothers with some primary or secondary education compared to those of mothers with no formal education after controlling for wealth, residence, and child age and sex [2]. In this paper we extend that research to address the association of women’s education with micronutrient deficiencies in children and women, including an assessment of population-level outcomes.

Causality between nutrition and education runs both ways, and both are influenced by national income that facilitates public and private investments in food, health and schooling [13]. Nutrition and education in turn both contribute to higher productivity and earnings, with estimates of undernutrition’s cost ranging from 2.5% of potential national income in India to about 6% in Uganda and over 16% in Ethiopia [14–16]. A significant proportion of the economic costs of undernutrition are due to the effects of poor child nutrition on educational performance and lifelong earnings. That is, impaired child growth and cognitive development due to undernutrition hinders school enrollment at an appropriate age, causes absenteeism or early drop-out due to ill-health or poor learning, and prevents optimal learning and skills development [17, 18].

Education levels have continued to rise in recent decades. The proportion of children up to age 15 worldwide ever attending primary school rose from 81% in 2000 to 85% in 2010, and by 2015 the net enrollment ratio in primary education was estimated at almost 92% of all children [19]. Importantly, the fastest gains have been among girls and women. Between 2000 and 2015, the number of girls for every 100 boys has risen from 92 to 97 in primary education, and from 91 to 97 in secondary education [20]. This change has been accompanied by an even larger decline in the prevalence of stunting, which from 1990 to 2015 is estimated to have declined by about 40% [21]. These promising global trends are driven primarily by rapid success in some regions, as large parts of Africa and Asia continue to suffer from both undernutrition and low school attainment.

Improving nutrition would support improved education outcomes, while enhanced educational performance - particularly of girls, who will become the next generation’s mothers - would help improve intergenerational effects of malnutrition [4, 22, 23]. However, the nature of this bi-directional relationship is not fully understood in terms of sequence, mechanisms, or magnitude of impacts. This is especially true of the links between women’s education and micronutrient deficiencies.

The existing literature focuses heavily on the association between maternal education and child stunting. In this paper, we address the composition of changes differentiated among six indicators of micronutrient deficiencies. Although maternal education is commonly identified as a predictor of anemia, serum retinol, serum zinc, and other measures of micronutrient status among children [24–31], few studies have explicitly compared how maternal education affects each micronutrient deficiency individually. In one study that did look specifically at maternal schooling and maternal nutrition knowledge as determinants of child anemia in Indonesia, both were significant determinants, and maternal nutrition knowledge could substitute for schooling in this relationship [32]. Similarly, another study evaluated maternal education and socioeconomic status (SES) as indicators of anemia in urban Korea and found that maternal education, but not other SES indicators, was inversely related to anemia among 10 year olds [33]. As Block has pointed out, the ability to understand and identify micronutrient content and quality in foods may require more knowledgeable or educated consumers, thus maternal education could be an important determinant of child micronutrient status [32].

This paper seeks to unpack the relationships between maternal education and micronutrient deficiencies and anemia globally, with a specific focus on child outcomes. We focus on women’s educational attainment, with the understanding that this represents the cumulative result of all girls’ education as total number of years spent in school. Where data relate specifically to mothers’ education we use the term ‘maternal education’; otherwise, we apply the term ‘women’s education’. The primary objective of this paper was to define the global associations between women’s education and key micronutrient deficiencies and anemia among children, non-pregnant women, and the general population, describe patterns across time and examine the impact of income on these relationships.

We identified recent databases with observations of variables of interest at the national level and globally representative when possible and data access was requested when necessary. We used country-level datasets that included information on educational attainment, gross national income (GNI) per capita, and estimates for anemia and micronutrient deficiencies. Each of the datasets used in this analysis are described below and outcome definitions are summarized in Additional file 1: Table S1.

VAD was classified as being of severe public health concern in 41.9% (n = 39) of the countries in the sample and the risk of zinc deficiency was high in 18.7% (n = 26) of the countries included in the analysis related to zinc deficiency. When we applied the cut-offs for adequate, inadequate or excessive iodine levels, iodine range was inadequate in 20.5% (n = 26) of the countries and 6.3% (n = 8) of countries were at risk for excessive iodine.

All regression models accounted for GNI per capita, which was significantly associated with the outcomes in all models. We tested for collinearity between GNI per capita and schooling in each model and the results suggested that collinearity was not a problem (VIF < 10). The implication is that while national wealth matters both for investments in educational systems and for building the food and health systems needed to support good nutrition, the degree to which individual girls are able to gain years of education matters independently of wealth.

Average years of schooling among women was significantly negatively associated with anemia among children, anemia among non-pregnant women, VAD among children and zinc deficiency among the country’s population, but not with median UIE of the population or the proportion of children protected against iodine deficiency using the most current years available. No schooling for women was significantly associated with all the outcomes evaluated (Fig. 1).

In describing the most current relationships between women’s educational attainment and micronutrient deficiencies, we tested for interactions between country income level and women’s educational attainment for each outcome. Income level was a significant effect modifier for the effect of years of women’s education on child anemia, but not on other outcomes. When this model was stratified by the national income levels, women’s years of schooling was no longer associated with child anemia in upper middle income countries (n = 33; β(SE) = − 0.34(0.90), p = 0.71), but remained a significant predictor of child anemia in all other income categories: low (n = 23; β(SE) = − 3.86(1.30), p = 0.008); lower middle (n = 34; β(SE) = − 2.78(0.98), p = 0.008); and high income countries (n = 51; β(SE) = − 1.29(0.54), p = 0.02)) (Fig. 1a).

When the interaction between no schooling for women and income level for each model was tested, we found that the relationship between no schooling in women and zinc deficiency in the population differed by national income level. When this model was stratified by income level, no schooling was no longer a predictor of zinc deficiency among low income (n = 35; β(SE) = − 0.05(0.08), p = 0.57) or lower middle income countries (n = 36; β(SE) = 0.18(0.09), p = 0.06), but was a significant predictor among upper middle (n = 28; β(SE) = 0.30(0.12), p = 0.02) and high income countries (n = 40; β(SE) = 0.45(0.09), p < 0.001) (Fig. 1d).

When trends in the relationships between women’s educational attainment and anemia and micronutrient deficiencies across time were assessed, we found the relationship between the percentage of no schooling among women and the prevalence of anemia among children did not differ significantly, while the relationship between the average years of schooling in women and the percentage of anemic children did differ by year (Fig. 2a). When the model was stratified by year, average years of schooling in women remained significantly negatively associated with anemia in children for all years, though the coefficients and standard errors decline toward zero over time, indicating that the relationship remains precisely estimated by the magnitude of the relationship declines over time (β(SE);P for 1985 = − 3.50(0.59); p < 0.001; 1990 = − 3.28(0.48); p < 0.001; 1995 = − 2.97(0.45); p < 0.001; 2000 = − 2.60(0.46); p < 0.001; 2005 = − 2.17(0.42); p < 0.001). In the same way, the relationship between women’s educational attainment and the prevalence of anemia in non-pregnant women differed across the years (Fig. 2b). The coefficients for average years of schooling in women ranged from − 2.36 (0.36) in 1985 to − 1.74 (0.34) in 2005, with the magnitude of the relationship decreasing consistently over time while remaining precisely estimated and significant. For the model with percentage of no schooling in women as the main predictor, coefficients increased from 0.25 (0.03) in 1985 to 0.29 (0.04) in 2005, remaining precisely estimated and significant across time.

In the case of both VAD and zinc deficiency, the relationships with women’s educational attainment did not differ by year for either predictor of either outcome (VAD: 1990–2005; zinc: 1985–2000).

Due to the lack of credible and comparable data on covariates within these pathways, this paper was not able to explore which of these, or other, mechanisms may be at work for different nutrients. This is an important knowledge gap given the large ongoing investments by the international community tackle globally-relevant micronutrient deficiencies.

We found evidence of the importance of GNI per capita on the relationship between women’s education and anemia and micronutrient deficiencies globally. Household SES is an important parameter that frames the pathway whereby maternal education can influence children’s nutritional status and the nutritional status of the girl child can influence girl’s education performance. Several studies have shown that children from wealthier families perform better in school, and similarly, higher levels of educational attainment lead to greater income [42]. Using 1998 Bolivia DHS data, Frost et al. found that SES factors were the most important pathways through which maternal education was linked to child nutritional status, though attitudes about modern health care utilization also facilitated this relationship [49]. Yet, the interaction between income and education differs by nutrient of interest. In the case of zinc deficiency, we found that a lack of formal education in women does not predict zinc deficiency in the general population in the low- and lower-middle income countries. This outcome is supported by country-specific studies suggesting that a base level of resources may be required before education can influence nutrition [57, 58]. In other words, education and income may not, in very low income settings, override the fundamental food insufficiency, unsanitary environments and lack of public services that are needed to ensure quality diets and healthcare [59]. Thus, as indicated by the many pathways proposed by the literature, many other things matter in the relationship between women’s education and nutrition beyond wealth.

For example, we may expect that the way in which self-confidence and empowering social networks can improve child growth outcomes (i.e. mothers making decisions about diets and health seeking), would similarly affect micronutrient intakes. The strong links between schooling, delayed marriage, delayed first birth, and fewer, healthier children may impact the micronutrient status of the mother and her children, although this is not a relationship that has been studied as much as effects on child anthropometry and maternal anemia [60].

Similarly, it has been shown that maternal education was associated with adequate compliance with micronutrient supplementation in children aged 3–24 months in Mexico [48] and was associated with the quality of children’s diets (3-day food diaries) in Cyprus [61]. A study in Kenya found that maternal cognitive tests scores - but not maternal education - were positively associated with the quality of the diets of school-age children [62].

There has been some work towards parsing out differences in maternal schooling vis-a-vis health and nutrition knowledge. Block found that in Indonesia, nutrition knowledge and maternal schooling were associated with child anemia and in cases where schooling and income were low, nutrition knowledge was still a determinant of child anemia [32]. Ruel concluded that the effect of maternal schooling on weight-for-age z score was mediated by the nutrition knowledge of the mother in households with access to resources, suggesting that both maternal schooling and the path through which schooling impacts child growth are important [58].

In the case of micronutrient deficiencies, there may be differences in the mechanisms through which maternal education influences each micronutrient outcome, depending on whether knowledge, resources or other constraints impact the acquisition or use of the nutrients, and at what level these constraints work (e.g. individual, household, community, or country). Child vitamin A status, for example, may be influenced by the primary caregiver’s knowledge of and access to high-dose vitamin A supplements, knowledge of vitamin A rich foods, resources to acquire vitamin A rich foods, etc., thus the mechanism through which women’s schooling is associated with child VAD could be through increased knowledge or through increased income and resources that allow for a vitamin A sufficient diet.

Some countries have or are in the process of implementing ‘universal’ supplementation programs as part of health service delivery, and/or establishing regulations for mandatory fortification of target foods with various micronutrients. In such cases, the importance of education to the ability of women to identify and procure specific foods rich in key micronutrients may be lessened, but the value-added of education to micronutrient status would remain, albeit operating via alternative mechanisms. For example, in our analysis, iodine status was not associated with years of women’s schooling, perhaps because in most contexts iodine status is less dependent on dietary choice than on national regulations for universal salt iodization. Once a country successfully reaches the threshold used to define ‘universal iodization’, the consumer does not decide between consuming iodine or not.

Our findings also indicate that years of schooling among adult men showed the same pattern of significant negative associations with anemia and VAD in children, anemia in women, and zinc deficiency in the general population. The significant association between no schooling of men and these outcomes was stronger than that with no schooling of women. It could be that men’s education is working primarily through the mechanism of increased income, which may be less diluted at the population level, whereas women’s education could be influencing micronutrient status through other pathways discussed above. We are not able to tease this out with the current analysis; however, such gendered differences in transmission pathways should be explored further.

While the results presented here are novel and strong, there are a number of limitations to this analysis that should be considered. This was an ecological study designed to quantify relationships at the global level highly informative, it cannot capture the nuances of relationships studied. Additionally, we anticipate variation at the individual level and between countries. In a 19 countries analysis, country-specific regressions indicated that there was diversity among countries in relationships between parental education and child’s height, indicating context specificity of this relationship [4].

Furthermore, we are unable to speak to the effect of an individual mother’s education on her own child’s micronutrient status, which should be a focus of future studies. Much of the data used are derived from modeling, including anemia and VAD in children aged 6–59 months, and anemia in non-pregnant women. Finally, the outcomes of interest were measured using different approaches and each included a number of inbuilt assumptions. For example, VAD, anemia, and UIE were measured using biochemical indicators, but zinc status was based on dietary availability at the country level, and the proportion of infants protected against iodine deficiency was based on a household level diet indicator. As well, we consider anemia as a proxy of iron deficiency, while recognizing that there are a number of nutritional and non-nutritional causes of anemia including non-modifiable factors such as genetic disorders. There remains a critical need for greater data comparability globally, standardization of thresholds used across countries, and disaggregation in many instances by child and adult age and by sex [63].

Despite these limitations, this analysis has a number of important strengths. To our knowledge, it is the first analysis that examines the relationship between women’s education and micronutrient deficiencies and anemia across countries. We account for national per capita income and changes over time, thereby helping to explain how women’s education modifies the relationship between income and malnutrition [64].

Education, especially women’s education, should remain at the forefront of the global development agenda. Many studies have clearly demonstrated the important relationship between maternal education and children’s survival, growth and development. We add to this body of knowledge by identifying global trends specific to women’s education and micronutrient deficiencies and anemia; we find that as average years of women’s schooling increases, the prevalence of anemia and VAD in children, anemia among non-pregnant women, and zinc deficiency in a country’s general population all decrease. While primary education enrollment is important, this study shows that years of education was associated with micronutrient status in children, women and populations, suggesting the importance of not just enrollment but also of continuation in school. Beyond the clear value of formal education to nutrition, schooling offers an entry point for specific actions that can support nutrition gains more broadly. Educational settings can be used as platforms for nutrition-specific interventions, for example providing nutritious foods at school such as in Food for Education/School Feeding Programs [65] or providing iron and folic acid supplements, deworming prophylaxis, and nutrition education as in the Adolescent Girls Anemia Control Programme in India [66]. Enabling and promoting women’s education can enhance the progress made with nutrition-specific interventions via the mechanisms discussed in this paper.

Harnessing efforts to ensuring girls education – especially good quality secondary education - that empowers women to make choices, while scaling-up the many well-documented cost-effective nutrition-specific interventions will accelerate progress towards the global nutrition goals [67] while bringing about great gains in human capital and sustainable development globally.