Menu
Menu
Menu
Menu
Background: This study examines socioeconomic inequality in children’s health and factors that moderate this inequality. Socioeconomic measures include household wealth, maternal education and urban/rural area of residence. Moderating factors include reproductive behavior, access to health care, time, economic development, health expenditures and foreign aid. Methods: Data are taken from Demographic and Health Surveys conducted between 2003 and 2012 in 26 African countries. Results: Birth spacing, skilled birth attendants, economic development and greater per capita health expenditures benefit the children of disadvantaged mothers, but the wealthy benefit more from the services of a skilled birth attendant and from higher per capita expenditure on health. Conclusion: Some health behavior and policy changeswould reduce social inequality, but the wealthy benefit more than the poor from provision of health services.
The Demographic and Health Surveys (DHS) for Africa are the primary source of data for the analysis (http://www.measuredhs.com). Data collected since 2003 from 26 countries are analyzed to examine the impact of social determinants on child health. We focus on this time period because some of the measures in DHS are comparable for this period (the wealth index and a more detailed measure of maternal education.) Using this time period also allows for the assessment of change, as more attention has been given to social disparities in health outcomes. Several countries have multiple surveys. DHS surveys are co-sponsored by USAID, the governments of the countries where the surveys are conducted, and several other foundations. Surveys are based on national probability sampling so that results can be generalized to the country level. Trained interviewers visit selected households and conduct interviews with men and women of reproductive age. Interviewers also prepare a household roster with basic information for all members of the household. These surveys have become widely accepted sources of information for a variety of health related topics. The key child health outcomes of interest are neonatal mortality (coded 0 or 1), the hazard rate of child survival until age five, and nutritional status as indicated by height-for-age Z-score multiplied by 100 to facilitate reporting of significant digits (HAZ). Measures of social status include maternal education treated as a interval level variable (no education, incomplete primary, complete primary, incomplete secondary, complete secondary, and post-secondary), wealth, a reflection of the household standard of living, as measured by household assets such as appliances and home building material sanitation facilities and housing construction, and urban/rural residency. Specific factors included in the wealth index vary from country to country (for details see http://www.dhsprogram.com/topics/wealth-index/Wealth-Index-Construction.cfm). Key moderating factors include prior birth interval (minimum of 24 months between births), presence of a skilled birth attendant (presence of doctor or nurse) at delivery, immunization (coded 1 if children received recommended immunizations including BCG, DPT 1 and Polio 1, and 0 otherwise) within 2 months of birth, year of the survey, per-capita income (GDP per capita), per capita expenditure on health and per capita expenditures on foreign aid in the 3 years prior to the survey. We only consider the first round of immunizations so we can include the youngest children in the analysis. The national level data on per capita income and per capita health expenditures were gathered from the World Bank [67]. If data for a specific DHS survey year were not available for a country, values within 3 years of the DHS survey year were used. Data on foreign aid were obtained from the AidData.org database [68]. Initially, we categorized aid by sectors including agriculture, health, reproductive health, water development, and all other aid. Per capita aid in all of these sectors except water were weakly associated with poor child health outcomes. Because we are interested in moderating factors that improve child health, analysis reported here only includes per capita foreign aid for water development. Several other household and child characteristics are associated with children’s health in developing countries [18]. This analysis includes maternal age, child’s age (in the models for nutritional status), child’s birth order, sex of the child, whether the child was a twin, presence of the father, marital status of the mother, household size, maternal employment and whether or not the father has at least some secondary education. Younger mothers may not be as likely to have resources and experience they can use to promote greater health for their children. As children age, their nutritional status (height-for-age Z-score) deteriorates relative to the WHO standard (see Fig. 1) because they do not receive adequate nutrition and are at risk of infections leading to diarrhea. Twins and children with more older siblings are at higher risk of mortality and undernutrition. Male children have higher rates of mortality but there is generally not a great gender difference in access to calories. The presence of a father in the home has been shown to be associated with better child outcomes [69–71]. For example, Dearden et al. found that children who saw their father daily or weekly at both one and 5 years of age had higher HAZ scores than children who saw their fathers less often at either or both ages (2012) [70]. Finally, father’s education provides an additional resource that may benefit children independent of maternal education and household wealth. We also include marital status of the mother, household size, and maternal employment to adjust for household structure and mother’s time availability. We considered including breastfeeding practices but measurement of this variable in DHS is not sufficient to capture the timing of exclusive breastfeeding and introduction of other foods into the diet. Educational inequality in children’s nutritional status (height/age z-score) Three regression models are used depending on the distribution of the measure of child health. Logistic regression is used to predict a dichotomous variable indicating mortality in the first month, Cox regression is used to predict child mortality measured in months, and linear regression is used for height for age z-scores. All countries and years are pooled. Regression models for neonatal mortality and nutritional status use multi-level models with country as the level two unit of analysis to account for intra-group correlations within countries. The Cox-regressions include fixed effects for each country. Stata 14.1 was used to estimate all models. Year of the survey, GDP per capita, health expenditures per capita and per capita aid for water development are measured at the national level. Forty-two percent of the households have more than one child under age 5. We estimated models adjusting standard errors for household clustering. Design effect statistics are all well below 2.0 (deff). Moreover, the standard errors in these models were only slightly larger and did not affect our conclusions. Regression coefficients for the three social determinants, maternal education, wealth and urban residence, indicate the degree of socioeconomic inequality in health outcomes: larger coefficients show a steeper gradient of difference between more and less advantaged children. For example, a coefficient of 4.88 for maternal education implies that a child whose mother has post-secondary education will score .25 standard deviations higher on height-for-age than a child whose mother has no education ((4.88*5)/100 = .244), indicating substantial educational inequality. A coefficient of 2.0 would only imply a .10 standard deviation difference between children of the most and least educated mothers. Interaction terms between each of the moderating factors and the social determinants show the degree to which these factors have potential to reduce inequality. If coefficients for interaction terms run counter to the coefficients for social determinants then mitigation is implied. In other words, if the influence of social determinants becomes smaller as the magnitude of moderating variables increase then the main effect of the social determinant and the interaction effect will work in opposite directions. For example, if the coefficient for maternal education is 5.0 and the interaction between birth spacing (coded 0 for short interval and 1 for long interval) is -3.0 then the education gradient is 5.0 for children with a short birth interval and only 2.0 (5 + -3*1 = 2) for children with a longer birth interval, implying that a longer birth interval reduces educational inequality in child nutritional status.
N/A
