Article Text

Download PDFPDF

How consistent are associations between maternal and paternal education and child growth and development outcomes across 39 low-income and middle-income countries?
  1. Joshua Jeong1,
  2. Rockli Kim2,
  3. S V Subramanian2,3
  1. 1 Department of Global Health and Population, Harvard T.H. Chan School of Public Health, Boston, Massachusetts, USA
  2. 2 Department of Social and Behavioral Sciences, Harvard T.H. Chan School of Public Health, Boston, Massachusetts, USA
  3. 3 Harvard Center for Population and Development Studies, Cambridge, Massachusetts, USA
  1. Correspondence to Mr Joshua Jeong, Department of Global Health and Population, Harvard T.H. Chan School of Public Health, Boston, MA 02115, USA; jjeong{at}mail.harvard.edu

Abstract

Background Maternal and paternal education are associated with improved early child outcomes. However, less is known about how these relative associations compare for preschool children’s growth versus development outcomes; and across country contexts.

Methods We analysed data from 89 663 children aged 36 to 59 months in 39 low-income and middle-income countries (LMICs). We used linear regression models with country fixed effects to estimate the joint associations between maternal and paternal education and children’s growth and development outcomes. Additionally, we examined the variability in these relationships by each country and within subgroups of countries.

Results In the pooled sample, maternal and paternal education were independently associated with 0.37 (95% CI 0.33 to 0.41) and 0.20 (95% CI 0.16 to 0.24) higher height-for-age z-scores, and 0.31 (95% CI 0.29 to 0.34) and 0.16 (95% CI 0.14 to 0.18) higher Early Childhood Development Index z-scores, respectively (comparing secondary or higher to no education). Associations were stronger for maternal education than paternal education but comparable between child outcomes. In country-specific regressions, we found the most heterogeneity in the associations between maternal education and children’s growth; and between paternal education and children’s development. Subgroup analyses suggested that these associations may be systematically patterned by country-level adult illiteracy, infant mortality and food insecurity.

Conclusion Our findings highlight variability in the statistical significance and magnitude of the associations between caregivers’ education and children’s outcomes. Further research is needed to understand the sources of variation that may promote or constrain the benefits of caregivers’ education for children’s early health and development in LMICs.

  • child health
  • education
  • international hlth
  • social epidemiology

Statistics from Altmetric.com

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

Introduction

Globally, it is estimated that approximately 250 million children under 5 years of age are at risk of not fulfilling their developmental potential.1 Inadequate stimulation, nutritional deficiencies, poor environmental conditions and maternal depression, among other co-occurring risk factors, have been found to compromise children’s early growth and development in low-income and middle-income countries (LMICs).2 3 Maternal education has been underscored as one key protective factor for promoting children’s early brain functioning, physical growth and long-term developmental trajectories in LMICs.2 In fact, many studies have shed light on a variety of mechanisms through which maternal education may improve child health and development outcomes. For example, studies have found that more educated mothers make greater investments in child health promoting activities4; provide more learning opportunities for the child5; demonstrate higher quality interactions with their children6; possess greater knowledge about child health and development7 8 and experience less parenting stress.9

While the importance of maternal education has been well-established, fewer studies have also considered the role of paternal education and examined the joint associations of maternal and paternal education with early child outcomes in LMICs. Three recent global studies have documented independent relationships between maternal and paternal education and children’s nutrition using the Demographic and Health Surveys data10 11 and development outcomes using Multiple Indicator Cluster Surveys (MICS) data.12 However, these studies have primarily focused on average associations in pooled samples, and less is known about the variability in the extent to which mothers’ and fathers’ education levels benefit their young children’s early health and development outcomes across countries. Considering the heterogeneity in socioeconomic and societal differences across LMICs—and particularly how school attainment, food access and security and learning opportunities are all lower for children and their parents in poorer countries13—it is reasonable to hypothesise that these relationships between caregivers’ education and child outcomes may vary across country contexts.

In this study, we aimed to estimate the main associations between maternal and paternal education and early child outcomes in our global sample, and explore the degree to which these relationships vary across 39 LMICs. We focus on how maternal and paternal education specifically relate to linear growth and developmental outcomes of children aged 36–59 months. Moreover, we explored country-level adult illiteracy, gender inequality, infant mortality, and food insecurity as factors potentially moderating these relationships across LMICs.

Methods

Data

We used data from Unicef’s MICS programme, which is an international household survey programme that collects information about health, nutrition, education and early development and well-being of children and their families in LMICs. For this study, we combined all publicly available national surveys from MICS rounds 4 and 5 (2010–2014). As the Early Childhood Development (ECD) module in the MICS was only asked regarding children aged 36–59 months, we restricted our data to this age group of children to enable comparability between children’s cognitive development and undernutrition outcomes.

Outcome variables

The two primary outcomes relating to early child health and development considered in this analysis were height-for-age z-scores (HAZ) and Early Childhood Development Index (ECDI) z-scores. HAZ is a measure of child’s linear growth, calculated as the child’s height minus the median value for that child’s age and sex, divided by the standard deviation (SD) of this age-specific and sex-specific group in the WHO Multicentre Growth Reference Study population.14 Biologically implausible values (HAZ <−6 or >6) were excluded. The ECDI was developed by Unicef for the MICS household survey programme. It is comprised of 10 caregiver-reported, dichotomously scored questions that are relevant to children aged 3 and 4 years regarding four developmental domains: cognitive, socioemotional, literacy-numeracy and physical development. These 10 items were determined through multicountry field tests, validity and reliability studies, and deliberation with experts.15 In this analysis, we exclude the two physical development questions, based on prior research that has revealed poor measurement model fit statistics for these two particular questions and for being less relevant to children’s early developmental skills.16 Instead, we focus on the eight questions relating to children’s literacy-numeracy, social-emotional and learning development. A composite score for ECD was created (ranging from 0 to 8) by summing the number of positive responses across the literacy-numeracy, social-emotional and learning domain items,12 and normalised to a z-score (mean of 0 and SD of 1) for ease of interpretation and direct comparability to the standardised scale of HAZ. Weight-for-age z-scores (WAZ) and weight-for-height z-scores (WHZ) were considered as secondary child anthropometric outcomes.

Exposure variable

The primary caregiver of the child reported on the highest level of formal education completed by the mother and father of the child. Because school grade levels vary across countries, paternal and maternal education were categorised into three groups: no or incomplete primary education, completed primary education or completed secondary education or higher.

Analysis

First, we conducted a complete-case, pooled analysis and specified two linear regression models to estimate the associations between maternal and paternal education and each child outcome variable of interest: HAZ, WAZ, WHZ and ECDI z-score. Model 1 only adjusted for child age (in months), sex and country fixed effects (to control for any country-level differences that may confound the relationship between caregivers’ education levels and child outcomes). Model 2 further adjusted for a full set of covariates: maternal age (categorised in 5-year age groups from 15 to 49 years), maternal marital status (three categories: currently married, formerly married, never married), household wealth index (quintiles within each country: calculated as a principal component of a group of assets owned by the household17) and place of residency (coded 0 for urban and 1 for rural).

In addition to the pooled analysis, we conducted separate country-specific regression models to examine the heterogeneity in estimates between caregivers’ education and the primary outcomes of child HAZ and ECDI z-score across country contexts (results for WAZ and WHZ are presented in the online supplementary appendix). In particular, we focused on standardised mean differences in child outcomes comparing caregivers with secondary education or higher with caregivers with no education in country surveys with >100 observations.

Supplementary file 1

Finally, we conducted several subgroup analyses based on country-level variables, indicating low versus high adult illiteracy (≤70% vs >70%, respectively), gender inequality index (≥0.50 vs <0.50), infant mortality (≥50 vs <50 per 1000 live births) and global food security index (≤50.0 vs >50.0). National statistic data for adult illiteracy, gender inequality index and infant mortality came from the United Nations Development Programme’s Human Development Reports18 and data for food security index came from The Economist Intelligence Unit.19

Standard errors across all models were clustered at the primary sampling unit (PSU) level and we accounted for the complex MICS survey design. All analyses were conducted using Stata, V.13.

Sensitivity analyses

We also conducted two sensitivity analyses for the overall pooled estimates of model 2 to examine the robustness of our findings. First, we restricted our sample to households in which children were living together with both their mother and father. Second, we respecified our model and additionally adjusted for PSU-level fixed effects to account for geospatial differences in socioeconomic characteristics of local enumeration areas that are common to all respondents from that area (eg, local diet, access to and quality of local education and cultural and social norms).

Results

On restricting to national surveys that included the ECD module or assessed children’s anthropometry, 388 656 children from 42 countries were identified. Of this sample of children under 5 years of age, 142 385 children were aged 36–59 months. After excluding children from Guinea Bissau and Ukraine where there was no variation in levels of caregiver’s education and those who had missing or invalid information for any of the covariates or outcomes, the final sample comprised 89 663 preschool-aged children with valid data for at least one of the four child outcomes from 39 countries (see online supplementary appendix figure S1).

Table 1 presents descriptive statistics of the final analytic sample. The average age of children was 47 months and approximately half of the sample was male. The majority of households (64.0%) resided in rural areas. In our pooled sample, the mean HAZ was −1.39 (SD 1.5), with 32.4% of the children classified as stunted, and the mean ECDI z-score was 0.0 (SD 1.0). HAZ and ECDI z-score were moderately correlated at the individual level (r=0.23). Overall, 30% and 22% of the children had mothers and fathers with no education, respectively. A greater proportion of fathers had completed secondary education or higher (47%) compared with mothers (38%) (table 1). The distribution of education level varied substantially across countries, ranging from nearly all mothers having secondary education or higher (99%) in Kyrgyzstan and Moldova to <10% in Sierra Leone and Chad (see online supplementary appendix figure S2). Similarly, the proportion of fathers having completed secondary or higher education ranged from >95% in Belarus, Barbados, Jamaica, Kyrgyzstan, Moldova and Kazakhstan to around 20% in Chad (see online supplementary appendix figure S3).

Table 1

Sample characteristics (n=89 663)

In table 2, we present adjusted associations between levels of maternal and paternal education and early child outcomes. There were two key findings from the pooled analysis. First, both maternal and paternal education (for all levels) were significantly associated with child growth and developmental outcomes, and the magnitude of the association was stronger for maternal education than paternal education. Second, the associations for both caregivers’ education were comparable with regard to children’s HAZ and ECDI outcomes. After adjusting for child’s age, sex, maternal age, marital status, household wealth and place of residence (model 2), children whose mothers completed secondary education or higher were associated with 0.37 (95% CI 0.33 to 0.41) higher HAZ, compared to children whose mothers had no education. Children whose fathers completed secondary education or higher were associated with 0.20 (95% CI 0.16 to 0.24) higher HAZ. Consistent results were found for WAZ and WHZ, although the role of caregivers’ education was weaker for these outcomes, compared with HAZ. Similarly, children whose mothers and fathers completed secondary education or higher were associated with 0.31 (95% CI 0.29 to 0.34) and 0.16 (95% CI 0.14 to 0.18) higher ECDI z-scores, respectively (table 2).

Table 2

Associations between maternal and paternal education and children’s HAZ, WAZ, WHZ and ECDI z-scores

In figures 1 and 2, we present country-specific results for the associations between maternal and paternal education and children’s HAZ and ECDI z-scores, respectively. Findings revealed notable variation underlying the global estimations, in terms of the statistical significance and magnitude of the associations across countries. Maternal education was significantly associated with children’s HAZ in 13 countries and ECDI z-scores in 23 countries. The magnitude of the positive associations between higher maternal education and children’s HAZ ranged from 0.18 (95% CI 0.02 to 0.34) in Ghana to 1.35 (95% CI 0.68 to 2.03) in Kosovo. For children’s ECDI z-scores, the positive associations ranged from 0.13 (95% CI 0.07 to 0.18) in Iraq to 0.66 (95% CI 0.09 to 1.23) in Jamaica. Paternal education was significantly associated with HAZ in 9 countries and with ECDI z-score in 15 countries. The magnitude of the positive association between higher paternal education and children’s HAZ ranged from 0.12 (95% CI 0.03 to 0.21) in Iraq to 0.40 (95% CI 0.21 to 0.59) in Togo. For children’s ECDI z-scores, the positive associations ranged from 0.08 (95% CI 0.01 to 0.15) in Nigeria to 1.10 (95% CI 0.06 to 2.15) in Macedonia. Considerable variations in country-specific results were also found for WAZ and WHZ (see online supplementary appendix figure S4,S5 respectively).

Figure 1

Associations between maternal and paternal education and children’s height-for-age z-score (HAZ) based on country-specific regression models. Estimates represent differences in children’s HAZ between caregivers with secondary or higher education and caregivers with incomplete primary education or no education. For each country, results (point estimate and 95% CI) with a circle (in red) correspond to maternal education and with a square (in blue) correspond to paternal education.

Figure 2

Associations between maternal and paternal education and children’s Early Childhood Development Index (ECDI) z-scores based on country-specific regression models. Estimates represent differences in children’s ECDI z-scores between caregivers with secondary or higher education and caregivers with incomplete primary education or no education. For each country, results (point estimate and 95% CI) with a circle (in red) correspond to maternal education and with a square (in blue) correspond to paternal education.

Exploratory subgroup analyses by country-level social and economic development indicators suggested that the associations between maternal and paternal education and early child outcomes may be systematically patterned (results for HAZ and ECDI z-scores in table 3; WAZ and WHZ in online supplementary appendix table S1). For example, maternal and/or paternal education were more strongly associated with children’s HAZ and ECDI z-scores in countries with lower health and development indicators (ie, higher adult illiteracy rate, infant mortality rate and food insecurity). On the contrary, maternal and paternal education were more strongly associated with ECDI z-scores in countries with lower gender inequality compared with those with higher gender inequality.

Table 3

Subgroup analyses for associations between maternal and paternal education and HAZ and ECDI z-scores by national development indicators

In our sensitivity analysis restricted to children who were living with both their mother and father (n=76 680 for HAZ analysis and n=84 202 for ECDI z-score analysis), our findings remained largely the same (see online supplementary appendix table S2). Moreover, the main results were robust to further adjusting for PSU/cluster-level fixed effects, with only a slight attenuation in the standardised mean differences (see online supplementary appendix table S3).

Discussion

In this study, we examined the joint associations between maternal and paternal education and children’s linear growth and development outcomes in a sample of 39 LMICs. We used the largest nationally representative and internationally standardised household survey programme that collects both early child anthropometric data and development data across LMICs. Extending beyond previous research that primarily focuses on pooled analyses and average associations across all LMICs, we additionally conducted country-specific regression models to present the extent of heterogeneity in associations across country contexts.

Our analysis revealed three main findings. First, we found that the majority of mothers and fathers (61.7% and 52.9%, respectively) in the pooled sample of LMICs had less than secondary education. Second, in the pooled sample, each caregiver’s education independently predicted both linear growth and development outcomes among preschool-aged children; associations were stronger for maternal versus paternal education. Third, these relationships were highly variable in terms of statistical significance and magnitude of association across LMICs. Results additionally suggested potential systematic patterning by several country-level indicators.

Several recent studies have similarly underscored the importance of maternal and paternal education for just undernutrition outcomes among children 0–59 months of age in 62 LMICs,11 just linear growth among children 25–59 months of age in 56 LMICs10 and just development outcomes among children 36–59 months of age in 44 LMICs.12 Our study provides an important extension to this knowledge base by concurrently investigating and comparing these relationships with both children’s growth and development outcomes in the same dataset and common population of preschool-aged children. In the pooled sample, we found that maternal education—particularly maternal secondary education or higher—predicted both child outcomes stronger than paternal education. Previous global studies have similarly found that the associations between maternal education and child outcomes are stronger than paternal education.10 12 20 Our results from the pooled sample revealed no differences in the relative associations of maternal and paternal education for preschool children’s growth versus development outcomes.

Our country-specific findings, moreover, added a more nuanced interpretation to the results from the pooled analysis, which alone suggested comparable associations between each caregiver’s education and the two child outcomes. Our study is one of the first in this global literature to compare the relative associations between maternal and paternal education and child growth and development outcomes both within individual countries and across the pooled sample of LMICs. More specifically, country-specific regressions revealed that the association between maternal education and children’s growth was more heterogeneous across countries than that for children’s development; the opposite was true for paternal education. Of note, while the magnitude was consistently greater for maternal education in the pooled analysis, country-specific results did not indicate significant mean differences between maternal and paternal education in the majority of countries.

Additionally, we found that the relationships between caregivers’ education and children’s health and development outcomes varied by several national development indicators. More specifically, maternal and paternal education were more strongly related to both child outcomes in high adult illiteracy countries; maternal and paternal education were more strongly related to specifically HAZ in high infant mortality countries and high food insecurity countries. These findings are consistent with a recent study by Alderman and Headley that also demonstrated that caregivers’ education was more strongly related to particularly children’s HAZ in countries with higher stunting burdens and higher education quality.10

The heterogeneous results of our country-specific findings and subgroup analyses by national indicators together suggest that some of the mechanisms driving the associations between caregivers’ education and child outcomes may be specific to each country. For instance, we found null associations between either or both caregiver’s education and child outcomes (with some of the greatest degree of uncertainty) in Bosnia, Macedonia, Montenegro and Panama. These countries ranked above average in our sample across each of the human development indicators that we examined. For countries with stronger social systems that more equitably reach and benefit young children’s early development, maternal and paternal education may perhaps matter less. On the other hand, we found more significant independent associations between either or both caregiver’s education and child outcomes in countries that were of poorer resource such as Bangladesh, Chad and Togo. Considering how children in lower-resource countries are more likely exposed to a constellation of co-occurring risk factors of poor development (ie, infectious disease, low dietary intake, poor housing quality, poor quality health and education services, etc),2 our results point to how both maternal and paternal education may serve as stronger protective factors for promoting children’s health and development, especially in such country contexts of adversity.

In light of our findings, additional research investigating and comparing the caregiving processes by which maternal and paternal education influence children’s health versus development outcomes is needed within and across countries. For example, some mechanisms may commonly underlie the relationship between caregivers’ education and children’s growth and development outcomes: such as the quality of the home learning and physical environment,21 22 maternal and paternal parenting knowledge regarding child care, health and development8 23 and caregivers’ emotional well-being. Whereas other mechanisms may be more specific to children’s cognitive and socioemotional development (ie, maternal and paternal stimulation12 and responsiveness); or linear growth and nutrition (ie, mothers’ and fathers’ own health and nutritional status, immunisation decision for the child23 or maternal and paternal feeding practices10). Future research should additionally explore how similarly or uniquely these mechanisms drive the varying associations between maternal and paternal education and children’s cognitive and nutritional outcomes.

There are a number of limitations to our study. First, the primary variables of caregivers’ education (coded as three categories) and the ECDI (comprising few yes/no items for each developmental domain) were crude measures, given the population-level scope and design of the MICS programme. Furthermore, these measures are based on the primary caregiver’s report, which raises the possibility of common rater, recall or social desirability biases. Second, the small sample sizes in select countries (Barbados and St Lucia) limited our power to detect differences in several country-specific regression models. Third, we were only able to pool MICS country surveys that have valid data for both maternal and paternal education and at least one early child outcome; therefore, our results are not representative of or generalisable across all LMICs. Finally, the MICS data are cross-sectional surveys, and the directionality and causality of the associations cannot be established.

Overall, our results underscore the importance of investigating beyond pooled associations between maternal and paternal education and children’s early health and development outcomes in LMICs. The considerable variation detected in terms of statistical significance and magnitude of the associations between each caregiver’s education and early child outcomes suggests complex mechanisms underlying these relationships across countries. Further understanding of the different contextual factors and caregiver-specific pathways that promote or constrain the protective role of maternal and paternal education are needed in order to effectively improve children’s growth and development outcomes within and across LMICs.

What is already known on this subject

Maternal and paternal education have been found to independently predict both early child health and development outcomes in low-income and middle-income countries (LMICs). However, much of this literature in LMICs has examined average relationships in pooled data across multiple countries. Little is known about the variability and consistency in the associations between maternal and paternal education and children’s growth and development outcomes across countries.

What this study adds

Examining beyond average associations in a pooled sample of LMICs, we found substantial variability in the associations between maternal and paternal education and children’s developmental and linear growth across LMICs using country-specific regression models. In particular, we found the most heterogeneity in the association between maternal education and children’s growth; and the association between paternal education and children’s development. Future global analyses using multicountry survey data should present country-specific estimates, in addition to pooled associations, which can shed light on the universality of risk and protective factors across LMICs, and inform country-specific interventions and policies for improving early childhood development.

Acknowledgments

The data for this research were collected by UNICEF’s Multiple Indicator Cluster Survey Program.

References

Footnotes

  • Contributors JJ, RK and SVS conceptualised the study and analysis plan. JJ conducted the analyses and drafted the manuscript. RK and SVS reviewed and edited the manuscript. All authors read and approved the final version submitted for publication.

  • Competing interests None declared.

  • Ethics approval This study was deemed exempt from ethics review by the Harvard T.H. Chan School of Public Health Institutional Review Board, as the MICS data used are publicly available and fully de-identified.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data sharing statement The data are from the Multiple Indicator Cluster Survey (MICS) Program and are available to registered MICS data users.