Ambel et al. International Journal for Equity in Health (2017) 16:152 DOI 10.1186/s12939-017-0648-1 RESEARCH Open Access Examining changes in maternal and child health inequalities in Ethiopia Alemayehu A. Ambel1* , Colin Andrews1, Anne M. Bakilana2, Elizabeth M. Foster1, Qaiser Khan1 and Huihui Wang1 Abstract Background: Ethiopia has made considerable progress in maternal, newborn, and child health in terms of health outcomes and health services coverage. This study examined how different groups have fared in the process. It also looked at possible factors behind the inequalities. Methods: The study examined 11 maternal and child health outcomes and services: stunting, underweight, wasting, neonatal mortality, infant mortality, under-5 mortality, measles vaccination, full immunization, modern contraceptive use by currently married women, antenatal care visits, and skilled birth attendance. It explored trends in inequalities by household wealth status based on Demographic and Health Surveys conducted in 2000, 2005, 2011, and 2014. The study also investigated the dynamics of inequality, using concentration curves for different years. Decomposition analysis was used to identify the role of proximate determinants. Results: The study found substantial improvements in health outcomes and health services: Although there is still a considerable gap between the rich and the poor, inequalities in health services have been reduced. However, child nutrition outcomes have mainly improved for the rich. The changes observed in wealth-related inequality tend to reflect the changing direct effect of household wealth on child health and health service use. Conclusions: The country’s efforts to improve access to health services have shown some positive results, but attention should now turn to service quality and to identifying multisectoral interventions that can change outcomes for the poorest. Keywords: Maternal and child health, Health inequalities, Health care utilization Background According to a 2012 UN report, all the MDG targets in Improving maternal and child health was integral to the Ethiopia were either on track or likely to be on track [2]. Millennium Development Goals (MDGs) for 1990–2015: The 2014 mini-Demographic and Health Survey (DHS) Goal 4 called for a two-thirds reduction in under-5 mor- found reductions in child undernutrition and child mor- tality, and Goal 5 for a 75% reduction in maternal mortal- tality and increased coverage of maternal, newborn, and ity. The goals received global attention as countries and child health (MNCH) services like antenatal care, contra- their international development partners mobilized sup- ceptive prevalence, and skilled birth attendance (SBA) [3]. port to, e.g., expand childhood immunization and increase With the MDG period behind us and new targets set in the availability and utilization of maternal health services. the Sustainable Development Goals (SDGs), attention is Ethiopia made considerable progress towards achiev- now turning to whether the achievements recorded were ing the targets. In 2011 the Center for Global Develop- inclusive [4]. Of the few studies on health inequalities in ment reported satisfactory progress on all goals and Ethiopia, most analyzed just one or two indicators at the ranked Ethiopia 33rd of 137 countries, with an MDG national level [5–8], while others examined one or two in- progress index of 4.5 on a scale of zero to 8 points [1]. dicators for a specific region or city [6, 9]. In this study, we provide evidence of the dynamics of MNCH inequalities. The study contributes to the empir- ical evidence by adding a more detailed inequality analysis * Correspondence: aambel@worldbank.org 1 The World Bank, 1818 H Street, Washington, DC 20046, USA using data from a series of comparable recent surveys. Full list of author information is available at the end of the article © The Author(s). 2017 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated. Ambel et al. International Journal for Equity in Health (2017) 16:152 Page 2 of 16 The surveys allow us to analyze the changes in inequalities sample for the 2014 survey was about half that of the previ- over a period that overlaps with most of the MDG period. ous rounds. Much of the data used in this study comes We also examine the contribution of socioeconomic de- from the women’s questionnaire, which compiles a compre- terminants of maternal and child health outcomes. hensive birth history for each woman, from antenatal care and delivery attendance through child survival and vaccin- Methods ation. The questionnaires used in the different surveys are Indicators and definitions standard and comparable. The data provide nationally rep- Table 1 presents the six health status and five health ser- resentative information on the variables we selected for this vice indicators analyzed in this study, chosen based on study. their relevance to the health MDGs and on the availability The primary qualifying variable is wealth ranking. We of data. For health status, there are three child undernutri- looked at socioeconomic inequalities in health by wealth tion and three mortality indicators: stunting, wasting, ranking between the worse-off (bottom 40%) and the underweight, neonatal mortality rate (NMR), infant mor- better-off (top 60%) and between the poorest (1st quintile) tality rate (IMR), and under-5 mortality rate (U5MR). The and the richest (5th quintile). These are computed from service indicators cover child immunization and maternal the household wealth index available with the data [10]. health services; they are measles vaccination, full To evaluate child undernutrition, we computed an- immunization, prevalence of modern contraceptive use by thropometric indicators based on the WHO 2006 growth married women, four or more antenatal care visits from a standards: We calculated height-for-age, height-for- skilled professional (ANC4+), and delivery assistance from weight, and weight-for-age z-scores and then stunting, a skilled birth attendant (SBA). Additional file 1: Table 1a wasting, and underweight levels for children aged 0 to and b present details of the indicators and terms as used 59 months. Child mortality rates (IMR, NMR, and U5MR) in the analysis. are calculated using the standard DHS methodology, using data on all child deaths in the 5-year period preceding Data sources and variable construction each survey [8]. We calculated the prevalence of modern The data are from four Ethiopia Demographic and Health contraception use by currently married women, antenatal Surveys (DHSs), in 2000, 2005, 2011, and 2014. These care (most recent birth), and skilled birth attendance (all population-based surveys target mainly women of child- births in last 5 years) from the questionnaire administered bearing age (15 to 49) but also collect some data about the to all women aged 15–49 years in the household. household and some from men in the same age range. The main survey concerns are fertility, family planning, infant Data analysis and child mortality, maternal and child health, and nutri- For a more complete picture, we considered a combin- tion. Since 2000, they have been conducted about every ation of the approaches often used in inequality studies 5 years. The first three surveys each sampled about 15,000 because each approach has some limitations that can lead women from about the same number of households. The to different conclusions [11, 12]. Our analysis started with Table 1 Maternal and child health indicators analyzed in this study Indicators Definition Stunting Percentage of children with a height-for-age z-score < −2 standard deviations from the reference median Wasting Percentage of children with a weight-for-height z-score < −2 standard deviations from the reference median Underweight Percentage of children with a weight-for-age z-score < −2 standard deviations from the reference median Neonatal mortality rate (NMR) The number of neonates dying before reaching 28 days of age per 1000 live births Infant mortality rate (IMR) The number of deaths among children under 12 months of age per 1000 live births Under-5 mortality rate The number of deaths among children under 5 years of age per 1000 live births (U5MR) Measles vaccination Percentage of children aged 12 to 23 months who received measlesa,b Full immunization Percentage of children aged 12 to 23 months who received BCG, measles, and three doses of polio and DPTb,a Contraceptive prevalence Percentage of currently married women aged 15 to 49 who currently use a modern method of contraception (modern method) Antenatal care visits, 4 Percentage of mothers aged 15 to 49 who had a live birth in the past 5 years who received at least 4 antenatal or more (ANC4+) care visits from any skilled personnel during pregnancy for the most recent birth Skilled birth attendant (SBA) Percentage of live births to mothers aged 15 to 49 in the past five years that were attended by skilled health attendant Notes:aImmunizations are either verified by card or based on recall of respondent. b Data not collected in 2014 mini-DHS survey Ambel et al. International Journal for Equity in Health (2017) 16:152 Page 3 of 16 how absolute and relative inequalities between the poor is plotted against the cumulative percentage of the popu- and the rich have widened or narrowed over time. Then, lation (x-axis). Thus if the bottom 5% of children by we looked at the concentration curve and concentration household wealth account for only 1% of the measles index, which capture inequality across a continuous vaccinations, the first point on the curve is (0.05, 0.01). spectrum of wealth, and what they reveal about the chan- Continuing: if the bottom 10% of children (cumulatively) ging pattern of inequality over time. Finally, we looked at account for 3% of measles vaccinations, the second point the decomposition of the concentration indexes to see the is (0.1, 0.03), and construction of the curve continues in changing role of various demographic and socio- the same way. The concentration curve is often plotted economic factors in the observed wealth-based inequality against the 45-degree line, the line of equality the con- in health services and outcomes. centration curve would follow if health outcomes were We computed absolute inequalities from rate differ- evenly distributed across the wealth rankings. ences between the poor and the rich, defined both as the We would expect the concentration curve for positive bottom 40% versus the top 60% and as the poorest health indicators (immunization, maternal health ser- quintile versus the richest. vices) to lie below the line of equity (poorer households The difference-in-differences comparison is as follows: account for a disproportionately low number of fully Let Ixt be the value of the indicator for group x (either vaccinated children or attended births). This is shown in r = rich or p = poor) in time period t (either t = 0 first the measles examples above, where both points plotted survey or t = T latest survey). We perform an F-test of lie below the 45-degree line. Conversely, we would ex- the hypothesis: pect the concentration curve for a negative health indi- cator (child mortality, malnutrition) to lie above the line I p0 −I r0 ¼ I pT −I rT ð1Þ of equality (poorer households account for a dispropor- tionately high number of child deaths). We expect that Second, instead of the difference between the value of the bottom 5% of children, ranked by wealth, account the indicator for rich and poor, the ratio of the values was for more than 5% of underweight children and thus the used. This emphasizes the difference between indicators point is above the 45-degree line. A concentration curve where both groups have very low values. The hypothesis that moves closer to the line of equality over time indi- tested was: cates decreasing inequality. I p0 I pT The concentration curve does more than offer a nice vis- ¼ ð2Þ ual summary of wealth-based inequality in an indicator; it I r 0 I rT is also useful to quantify the degree of inequality revealed. The analysis of inequality based on absolute and relative The concentration index (C) quantifies the degree of gaps was limited to binary distinctions: rich vs. poor. It inequality—twice the area between the concentration curve was thus somewhat sensitive to the definition of the bin- and the line of equality—which is analogous to how the ary distinction and also did not allow for analysis of in- Gini coefficient quantifies the degree of inequality in a Lo- equality across the whole range of wealth outcomes. For renz curve. We calculated C, for each indicator as follows: example, the comparison of the bottom quintile to top quintile entirely ignored any changes in the health indica- 2 C ¼ COV ðh; r Þ ð3Þ tor for the middle three-fifths of the population. If we see μ a decrease in inequality between the bottom 40% and the top 60%, we do not know whether this was due to im- where h is the health variable, μ is the mean, and r is the provements for the poorest of the poor or for those closer fractional rank of the individual in the wealth index. to the middle of the income distribution. Therefore, we When the outcome variable is binary, the concentration used concentration curves to illustrate the movement of index has some questionable properties, especially com- wealth inequalities in health across the entire range of paring populations that have significantly different means. wealth between the earliest and latest surveys. We also In particular, because it is mathematically bound between used concentration indices (C), which quantify the degree μ − 1 and 1- μ (where μ is the mean of the binary indica- of inequality in this analysis, and observed how they chan- tor), it tends to naturally fall (in absolute value) as the ged between the earliest and latest surveys [13]. value of μ increases. If only the richest 10% have access to A concentration curve plots the inequality of an out- a health service in the base year, the concentration index come variable against another factor, here household would be 0.9. If the richest 90% have access to the health wealth. It is constructed like a Lorenz curve, which illus- service in a subsequent year, the concentration index trates the degree of inequality in a certain variable such would be 0.1. It is therefore debatable whether that should as income. The population is sorted according to wealth, be considered a large decrease in inequality [14, 15]. For and the cumulative percentage of the indicator (y-axis) binary indicators, we calculated as alternative indicators Ambel et al. International Journal for Equity in Health (2017) 16:152 Page 4 of 16 the Wagstaff concentration index [W = 2/ μ (1- μ) COV(h, Table 2 Trends in MNCH outcomes in Ethiopia, 2000–2014 r)] and the Erryegers concentration index, [E = 8 COV(h, 2000 2005 2011 2014 Change r)], where h is a health outcome indicator and r is the frac- (Latest-Earliest) tional rank of the individual in the wealth index [14]. Stunting 57.0 49.5 44.1 40.6 −16.4 Our analysis also included decomposition of the con- Wasting 12.5 12.4 10.1 8.9 −3.6 centration index measure of wealth-related inequalities Underweight 41.9 34.1 29.1 26.6 −15.3 in selected health outcomes and services. This allowed NNMR 48.4 39.3 37.4 34.3 −14.1 us to see how differences in, e.g., family size, women’s IMR 95.9 77.6 58.9 59.3 −36.6 education, and access to safe water contribute to the ob- served wealth-related inequalities in health outcomes or U5MR 163.9 123.3 87.1 80.1 −83.8 health services, and how these patterns are changing Measles vaccination 27.0 36.5 56.7 - 29.7 over time. For example, the observed inequality in child- Full immunization 14.6 21.6 24.9 - 10.3 hood vaccination rates might be explained entirely by Contraceptive 7.2 15.7 30.0 45.2 38.0 differences in the mother’s education; decomposition of ANC4+ 10.4 11.9 15.9 24.2 13.8 the concentration index would reveal that. SBA 5.7 5.7 10.0 15.5 9.8 Decomposition of the concentration index is based on Source: Authors’ compilation from the EDHS (2000, 2005, 2011 & 2014) data the algebraic transformation, which for a linear model of Notes: The values are number of births per 1000 live births for NMR, IMR and a health indicator U5MR and percentages for the rest. Earliest survey year is 2000 for all X indicators. Latest survey year is 2014 for all but full immunization and measles y¼αþ β x þε k k k ð4Þ vaccination, for which our latest source of information is the 2011 DHS allows the concentration index to be written as health service delivery is among the least developed in X À low-income countries—modern services reach only a C¼  βk x k =μÞ C k þ GC ε =μ ð5Þ k small fraction of the population. For example, the results where μ is the mean of y, x  of the 2014 survey show that nationally modern contra- k is the mean of xk , Ck is the concentration index for xk , and CGε is the generalized ceptive use is 45%. The situation is much worse for ANC concentration index for the error term [15]. visits, where services coverage was 25%, and SBAs, where The dependent variables of interest in this study are coverage was 16%. Similarly, full immunization coverage binary indicators and thus best modeled using a non- in 2011 was 25%—among the lowest in similar countries linear model such as probit. Following previous work in Sub-Saharan Africa. Over the last two decades, how- [15, 16]), the linear approximation is given by the follow- ever, there has been considerable improvement in MNCH ing specification, outcomes in Ethiopia, though child undernutrition and X mortality rates are still high and coverage of maternal and h ¼ αm þ βm x þ u j j j ð6Þ health services is low. This holds true for all the health status and health service indicators analyzed in this study where h is the health variable of interest as defined earl- (Table 2): There is a consistent decline in ill health (under- ier, xj are the independent variables, αm is the constant nutrition and mortality) and an increase in health services term, βm j are the partial effects of each variable treated coverage (immunizations and maternal health services). as fixed parameters and evaluated as sample means, and To look at the trends by wealth status, we disaggregated u is the error term. The decomposed concentration the progress made. For each indicator, each line represents index (C) for a health outcome hi is therefore the value of the indicator for one group over time, with X the 95% confidence interval around each value. Figures 1 C¼ βm j =μÞC j þ GC u =μ ð7Þ j x to 4 present the results for child nutrition, child mortality, j immunization, and maternal health services. As expected, where Cj are the concentration indexes for xj, μ is the as a general pattern the lines for adverse outcomes slope j is the mean of xj, and mean of the health variables h, x downward for all groups and the lines for health services GCu/μ is the residual component that captures inequal- slope upward, showing MNCH is improving at all wealth ity that is not explained by systematic variation in the levels; the improvements in average national figures, how- regressors by income. ever, do not hide worsening results for the poor. The dis- tances between the curves and the slope of each curve in Results each graph show differing initial and final inequality for Trends the indicators. Table 2 presents the profile and trends of selected MNCH In Fig. 1, the lines in all graphs move downward basic- outcomes and services. As the table makes clear, Ethiopia’s ally in parallel, with perhaps a slight widening of the gap Ambel et al. International Journal for Equity in Health (2017) 16:152 Page 5 of 16 Fig. 1 Trends in child malnutrition by wealth ranking, 2000–2014. Source: DHS 2000 and mini-DHS 2014 data. Notes: Percent of children under 5 years old. Grey lines represent 95% confidence intervals between the different wealth groups. In the graph for between different groups, and the lines crisscross each stunting, we see that inequality is driven by the differ- other, with poorer households sometimes seeming to ence between the richest quintile and the rest, and that have lower child mortality (which might be true and re- this difference is increasing slightly. flect differences in practices like breastfeeding). It is not Figure 2 shows the trends for child mortality indica- clear whether it is actually true that there is no system- tors. The biggest improvement has been in the U5MR, atic difference in child mortality by household wealth with modest improvements in the IMR and essentially level or whether the figures are obscured by the data col- no change in neonatal mortality. No clear pattern of lection strategy (which only counts children whose wealth-related inequality is observable. The confidence mother is alive) or cultural traditions (unwillingness to intervals often overlap, showing no significant difference speak about an infant who died very young). Fig. 2 Trends in child mortality by wealth ranking, 2000–2014. Source: DHS 2000 and mini-DHS 2014 data. Notes: Deaths per 1000 live births. Grey lines represent 95% confidence intervals Ambel et al. International Journal for Equity in Health (2017) 16:152 Page 6 of 16 In Fig. 3, the trends show that immunization services section shows the relative inequality (the ratio of the indi- have expanded for all wealth groups, with the lines mov- cators) and the p-value for whether it has changed signifi- ing roughly in parallel. Again, the biggest gap is between cantly. A positive difference or ratio greater than 1 for an the richest quintile and all the others. ill-health outcome (child undernutrition or child mortal- As Fig. 4 shows, maternal health services have also im- ity) shows a pro-rich inequality—child undernutrition and proved for all groups, but there is a widening of the gap mortality rates were lower for children from better-off between rich and poor, with a greater increase in the use households. Likewise, a negative difference or ratio less of maternal health services by richer households. This than 1 in any of the immunization and maternal health trend is less pronounced for contraception use, and in fact service indicators (good health service utilization) implies there seems to be a slight closing of the gap in the last pro-rich inequality. These socioeconomic differences are year of data. Again, we see that wealth-based inequality is to be expected; our interest is in whether the differences primarily driven by the difference between the wealthiest are increasing or decreasing. quintile and the rest, especially for ANC and SBA. The results (Table 3) point to a widening of absolute pro- rich inequality in child nutritional outcomes between the Absolute and relative inequalities poor (bottom 40%) and the rich (top 60%)—an inequality The figures in the previous section give us a good indica- observed in all three child nutrition status indicators but tion of the general trends: fairly consistent improvements more significant for stunting and underweight. For the across all wealth groups, with some modest narrowing or poor, child stunting in the earliest survey was higher by widening of gaps depending on the indicator and the exact about 4.2 percentage points and in the latest the difference breakdown of households by wealth. To quantify and test rises to 10 percentage points. Similarly, the gap in under- the statistical significance of these observations, we report weight went up from 5 to 8 percentage points. In both the results from eqs. 1 and 2 in Tables 3 and 4. The results cases, these changes are statistically significant, which im- in Table 3 compare the bottom 40% of households and plies that pro-rich inequality is widening. The same conclu- the top 60%. Those in Table 4 compare the bottom quin- sions can be drawn when looking at relative inequalities: tile to the top quintile. In each table, the first section gives pro-rich inequality widened significantly during the period the value of the indicator for “poor” and “rich” households considered, at least for stunting and underweight. in both the first and the last year. The second shows the When we restrict our analysis to the poorest versus absolute inequality (the difference between the values of the richest quintile (Table 4), the trend recurs: an ever- the indicator for the two groups) in the earliest and the increasing gap between the rich and the poor in terms latest survey and the p-value for whether absolute inequal- of child nutrition. These changes are less statistically sig- ity changed significantly during that period. The third nificant, however, and only the relative inequality in Fig. 3 Trends in child immunization by wealth ranking, 2000–2011. Source: DHS 2000 and 2011 data. Notes: Percent of children 12–23 months old. Grey lines represent 95% confidence intervals Ambel et al. International Journal for Equity in Health (2017) 16:152 Page 7 of 16 Fig. 4 Trends in maternal health services by wealth ranking, 2000–2014. Source: DHS 2000 and mini-DHS 2014 data. Note: Percent of married women/births. Grey lines represent 95% confidence intervals stunting between the richest and the poorest changes the standard errors on the mortality rates for different significantly. subgroups are quite high, and that none of the changes The results for infant and child mortality appear to over time are statistically significant. show that mortality rates were in fact lower for poorer For all services except measles vaccination, in the first households in the first survey (the values for the differ- survey, utilization by poorer households was very low; ence between rich and poor are negative in all three in- thus while the difference between the rates increased dicators and in both ways of defining rich and poor). (significantly in the case of maternal health services), Then, by the final survey, the pro-poor gap appears to which suggests widening inequality, the ratio of the rates have disappeared for infant mortality and is reversed for decreased, suggesting that inequality had narrowed. For under-5 mortality, with both rates higher for poorer measles vaccination, both the variances and the ratio households. While these figures may represent real suggest decreasing inequality, although only the changes changes in mortality patterns, it should be noted that in the ratio are significant. Table 3 Trends in MNCH between poor and rich households, 2000–2014 Top 60% Bottom 40% Difference Ratio Earliest Latest Earliest Latest Earliest Latest p-value Earliest Latest p-value Survey Survey Survey Survey Survey Survey Survey Survey Stunting 55.2 36.2 59.4 45.9 −4.2 −9.8 0.064 0.9 0.8 0.013 Wasting 12.4 7.5 12.7 10.6 −0.3 −3.0 0.077 1.0 0.7 0.034 Underweight 39.9 23.1 44.7 30.9 −4.7 −7.8 0.189 0.9 0.7 0.011 NNMR 53.8 40.3 40.9 27.1 12.9 13.2 0.977 1.3 1.5 NA IMR 105.8 60.7 82.2 57.7 23.6 3.0 0.176 1.3 1.1 NA U5MR 171.9 79.0 152.7 81.7 19.2 −2.7 0.212 1.1 1.0 NA Measles vaccination 33.9 62.1 17.5 49.9 16.5 12.2 0.354 1.9 1.2 0.006 Full immunization 19.3 30.3 8.1 18.1 11.2 12.2 0.796 2.4 1.7 0.143 Contraceptive 10.1 50.8 3.2 36.6 6.9 14.2 0.033 3.2 1.4 0.002 ANC4+ 14.8 34.8 4.3 10.6 10.4 24.2 0.000 3.4 3.3 0.876 SBA 8.9 24.2 1.2 5.0 7.7 19.3 0.000 7.5 4.9 0.142 Source: Authors’ compilation from the EDHS (2000, 2005, 2011 & 2014) data Notes: Indicators are given in percentage points except for mortality rates, which are number of deaths per 1000 live births. The earliest year is 2000 for all indicators. The latest year is 2014 for all indicators except immunization (full and measles), for which it is 2011 Ambel et al. International Journal for Equity in Health (2017) 16:152 Page 8 of 16 Table 4 Trends in MNCH between poorest and richest households, 2000–2014 Richest Quintile Poorest Quintile Difference Ratio Earliest Latest Earliest Latest Earliest Latest p-val Earliest Latest p-val Survey Survey Survey Survey Survey Survey Survey Survey Stunting 47.7 27.4 59.3 47.1 −11.6 −19.7 0.057 0.8 0.6 0.002 Wasting 8.9 7.4 12.1 11.7 −3.2 −4.3 0.656 0.7 0.6 0.559 Underweight 31.0 16.6 43.8 33.4 −12.8 −16.8 0.237 0.7 0.5 0.003 NNMR 42.3 45.3 33.7 36.8 8.6 8.4 0.991 1.3 1.2 NA IMR 88.2 67.7 77.9 67.0 10.3 0.7 0.669 1.1 1.0 NA U5MR 137.6 78.4 136.1 89.7 1.5 −11.3 0.603 1.0 0.9 NA Measles vaccination 53.1 79.9 18.7 47.1 34.3 32.8 0.825 2.8 1.7 0.023 Full 34.8 50.9 7.2 17.5 27.6 33.4 0.419 4.8 2.9 0.153 immunization Contraceptive 26.1 57.2 3.2 30.9 23.0 26.2 0.538 8.3 1.8 0.000 ANC4+ 34.8 60.2 4.1 8.6 30.7 51.7 0.000 8.6 7.0 0.481 SBA 25.4 55.6 0.9 4.5 24.5 51.0 0.000 29.6 12.3 0.060 Source: Authors’ compilation from the EDHS (2000, 2005, 2011 & 2014) data Notes: Indicators are given in percentage points except for mortality rates, which are number of deaths per 1000 live births. The earliest year is 2000 for all indicators. The latest year is 2014 for all indicators except immunization (full and measles), for which it is 2011 Concentration curves and indexes services—richer households use health services more. The The concentration curves in Fig. 5 show child malnutrition values of the concentration indexes for mortality can be ei- as fairly even distributed across income categories, with the ther positive or negative; there is no clear trend. The abso- curves lying close to the line of equality, but the curves lute value of the concentration index can be taken as a moving away from the line indicate gradually increasing in- measure of the inequality present. The tables also show the equality. Overall, the results presented here are aligned with result of simple t-tests of whether the concentration index the findings already presented: the poor did not benefit as value changed significantly between the first and the last much from the improvements in child nutritional status. survey. Figure 6 shows the concentration curves for child The concentration indexes for all three undernutrition mortality indicators. Here the curves overlap and some indicators show pro-rich inequality heightening; in abso- cross the line of equity. This reflects the fact that there lute value the concentration index more than doubled for are no obvious patterns or trends when mortality rates all three, although from initially low levels of inequality. are broken down by household wealth level. Statistical The changes are significant for stunting and underweight. tests confirm that for these indicators there is no signifi- Using the Wagstaff or the Erryegers concentration index cant change in wealth-based inequality. yields the same result but with somewhat less significance. Figs. 7 and 8 illustrate the movement of wealth-related On the other hand, like the concentration curves (Fig. 5), inequalities in child immunizations and maternal health there is no clear pattern in the indexes for mortality indi- services. Here there is clearly significant inequality, with cators (Table 5). Table 5 also shows mixed progress in de- the concentration curves lying far below the line of creasing inequality in the use of MNCH services; over equality. However, we also see substantial movement time concentration index values decrease, but except for over time as the curves move closer to the line of equal- measles vaccination and contraceptive use the changes are ity, reflecting less wealth-based inequality in use of ser- not significant when the Wagstaff concentration index is vices. The figure for antenatal care shows clearly that used. However, the Erryegers concentration index leads to improvements are concentrated among richer house- the opposite conclusions: it shows increasing inequality in holds— middle-ranked households are catching up to four of the five indicators and is significant for maternal the wealthiest households, but the poorest households health services. This again reflects the pattern we see, par- are not catching up to those in the middle. ticularly in maternal health services: having expanded rap- For each indicator, Table 5 provides information on the idly from low levels tends to lead to a decrease in the concentration index and, for binary indicators, the alterna- standard concentration index that is not robust to using tive Wagstaff and Erryegers concentration indexes for the the Erryegers index. earliest and latest years. Note that, as expected, the index Despite progress then, as both the concentration values are negative for malnutrition indicators—richer curves and indexes illustrate, inequalities are still sub- households have less malnutrition—and positive for MNCH stantial in the case of ANC4+ and SBA. Ambel et al. International Journal for Equity in Health (2017) 16:152 Page 9 of 16 Fig. 5 Concentration curves of child nutritional status, 2000–2014. Source: DHS 2000 and mini-DHS 2014 data. Note: Percent of children under 5 years old Changes in inequality utilization by the rich and the poor has increased, but We used three different methods to assess changes over initially service use by poor households was extremely time in wealth-related inequality in indicators of health low and the gains they have made are substantial. Thus, status and health service utilization: rate differences, rate the rate ratio, the concentration curves, and two of the ratios, and concentration indices. Table 6 summarizes three concentration indexes analyzed all rate the changes the results. as a decrease in inequality. Although child malnutrition has been reduced for all income groups, wealth-based inequality worsened over Decomposition the period studied. This pattern is consistent for all three The analyses so far discussed demonstrate that, despite indicators and all methods of analysis. improvements, there are still considerable inequalities in The DHS data show no clear relationship between some MNCH outcomes and services. In this section, we household wealth and child mortality. In earlier years, examine what may be contributing to the wealth-related there may have been inequality in favor of poorer house- inequalities in certain indicators. holds, or possibly an inverse-U shaped pattern, with Tables 7 to 9 present the decomposition of concentra- households in the middle of the wealth distribution hav- tion indexes for selected indicators. Various controls that ing the highest child mortality. That pattern might be may be related to both wealth status and the value of explained by cultural factors, or it might be an artifact of the indicator are included to see how much of the in- the data collection strategy. The concentration index equality can be attributed to factors like the mother’s suggests there is some evidence that inequality in child education or adequate sanitation facilities. mortality has “improved,” moving from a pro-poor bias Table 7 summarizes the results for stunting: in each to a smaller pro-poor bias or a small pro-rich bias. year, even when controlling for other variables, the largest Inequality has narrowed slightly in child vaccination contribution comes from the wealth index. The next lar- services, particularly for measles vaccinations. For ma- gest comes from mother’s education. Comparing the pro- ternal health services, the results seem contradictory: gression of stunting between 2000 and 2014, we see that The absolute difference in maternal health service the increased inequality comes only from the contribution Ambel et al. International Journal for Equity in Health (2017) 16:152 Page 10 of 16 Fig. 6 Concentration curves of child mortality, 2000–2011. Source: DHS 2000 and mini-DHS 2014 data of the wealth index itself, not from any other factors. The Table 8 shows the decomposition for measles vaccin- decomposition for 2014 has a large positive residual com- ation (Panel a) and full immunization (Panel b). Com- ponent, however, suggesting that other factors not cap- paring results in 2000 and 2014, we see that for measles tured in the decomposition are in fact offsetting the rise in vaccination the contribution of the wealth index has wealth-related inequality in stunting. gone from being the largest to being one of the smallest Fig. 7 Concentration curves of child immunization coverage, 2000–2011. Source: DHS 2000 and 2011 data. Notes: Percent of children 12–23 months old Ambel et al. International Journal for Equity in Health (2017) 16:152 Page 11 of 16 Fig. 8 Concentration curves of selected maternal health services, 2000–2014. Source: DHS 2000 and mini-DHS 2014 data but when all immunizations are considered it was still entirely by differences in the education of the mother very important. This suggests that the pure wealth com- and the father. ponent of inequality in measles vaccinations almost dis- In Table 9 Panels a through c present the decompos- appeared when coverage was expanded, as it was for ition results for maternal health services. A serious limi- measles vaccination but not full immunization. In both tation is that in all specifications the residual component cases, the remaining inequalities can be explained almost is large. With that in mind, however, it is still possible to Table 5 Concentration indexes of selected child health status indicators in Ethiopia, 2000–2014 Concentration Index Wagstaff Concentration Index Erryegers Concentration Index Earliest Latest p-val Earliest Latest p-val Earliest Latest p-val Stunting −0.032 −0.086 0.001 −0.075 −0.145 0.018 −0.074 −0.140 0.021 Wasting −0.034 −0.102 0.105 −0.039 −0.112 0.116 −0.017 −0.036 0.240 Underweight −0.054 −0.119 0.003 −0.093 −0.162 0.025 −0.091 −0.127 0.163 NNMR 0.049 0.075 0.000 0.052 0.078 0.000 − − − IMR 0.038 0.008 0.000 0.042 0.009 0.000 − − − U5MR 0.016 −0.022 0.000 0.019 −0.024 0.000 − − − Measles vaccination 0.242 0.096 0.000 0.331 0.223 0.043 0.261 0.219 0.368 Full immunization 0.344 0.223 0.025 0.403 0.297 0.110 0.201 0.22 0.594 Contraceptive 0.496 0.119 0.000 0.534 0.218 0.000 0.142 0.216 0.009 ANC4+ 0.474 0.381 0.017 0.529 0.503 0.578 0.197 0.369 0.000 SBA 0.665 0.525 0.004 0.705 0.622 0.116 0.150 0.325 0.000 Source: Authors’ compilation from the EDHS (2000, 2005, 2011 & 2014) data Note: Latest survey for immunization indicators is 2011 Ambel et al. International Journal for Equity in Health (2017) 16:152 Page 12 of 16 Table 6 Summary of results of changes in income related MNCH inequalities, 2000–2014 Rate Differences Rate Ratio Conc. Test of Dominance: Concentration Curves Index Bottom 40% vs. Poorest (q1) vs. Bottom 40% vs. Poorest (q1) vs. mca rule iup rule Top 60% Richest (q5) Top 60% Richest (q5) Panel A: Health Status Stunting Worsened Worsened Worsened Worsened Worsened NS NS Wasting Worsened NS Worsened NS NS Worsened NS Underweight NS NS Worsened Worsened Worsened Worsened NS NMRa NS NS − − Improveda − − IMRa NS NS − − Improveda − − U5MR a NS NS − − Improveda − − Panel B: Health Services Measles vaccination NS NS Improved Improved Improved Improved NS b Full immunization NS NS NS NS Improved NS NS Contraceptive Worsened NS Improved Improved Improved Improved NS b ANC4+ Worsened Worsened NS NS Improved Improved NS SBA Worsened Worsened NS Improved Improvedb Improved Improved Notes: The table summarizes results of different approaches presented in the previous sections and the test of dominance. The test of dominance is based on [15]. The number of evenly spaced quintile points is 19 (from 5% to 95%) and the significance level is 5%. The dominance test rules, mca and iup, respectively denote the multiple comparison approach and the intersection union principle. “–” is test not applicable. NS is no significant change from the earliest survey a The improvement for mortality indicators is a progression of the distribution from a more pro-poor towards the line of equality. bImprovement is not significant if Wagstaff concentration index is used discern some common trends: there is only a slight Discussion decline in inequalities over the period, the prevailing In examining differential progress in health status and inequalities are still high, and the wealth index, edu- health services utilization in Ethiopia, we used a variety cation, and residence of the user all make important of methods to look at inequalities by household wealth contributions to the differences in utilization of these in selected MNCH indicators. We are able to make three services. main observations: Table 7 Decomposition of concentration index for stunting (2000–2014) 2000 2014 Elasticity Conc. index Contribution Elasticity Conc. index Contribution Child age 1.639 −0.003 −0.004 2.202 −0.011 −0.025 Child age squared −0.862 −0.004 0.003 −1.168 −0.016 0.018 Sex (Male) 0.028 −0.004 0.000 0.031 0.010 0.000 Birth order 0.010 −0.068 −0.001 −0.048 −0.024 0.001 Wealth index −0.192 0.142 −0.027 −0.310 0.180 −0.056 Child HH members 0.038 −0.016 −0.001 0.000 −0.033 0.000 HH size −0.008 −0.025 0.000 0.057 0.005 0.000 Education (mother) −0.020 0.531 −0.011 −0.037 0.460 −0.017 Education (HH head) −0.018 0.347 −0.006 −0.026 0.349 −0.009 Residence (urban) 0.010 0.848 0.009 −0.013 0.709 −0.009 Region . . −0.005 . . −0.001 Residual . . 0.010 . . 0.055 Total . . −0.033 . . −0.042 Notes: Child HH members are household members under 5 years of age. HH size is total household size. Regions are dummies for each region with one reference region (Tigray) omitted from the regressions Ambel et al. International Journal for Equity in Health (2017) 16:152 Page 13 of 16 Table 8 Decomposition of concentration index for immunization indicators (2000–2011) 2000 2011 8a. Measles Vaccination Elasticity Conc. ind Contribution Elasticity Conc. ind Contribution Age (mother’s) 0.124 −0.023 −0.003 0.304 0.000 0.000 Child HH members −0.011 −0.014 0.000 −0.073 −0.030 0.002 HH size −0.036 −0.027 0.001 −0.109 −0.005 0.001 Wealth index 1.437 0.098 0.141 0.856 0.104 0.089 Education (mother) 0.057 0.518 0.029 0.070 0.486 0.034 Education (father) 0.107 0.407 0.043 0.024 0.364 0.009 Residence (urban) −0.006 0.885 −0.005 −0.011 0.800 −0.009 Region . . 0.004 . . −0.001 Residual . . 0.031 . . −0.029 Total . . 0.242 . . 0.096 8b. Full Immunization Age (mother’s) 0.096 −0.023 −0.002 0.362 0.000 0.000 Child HH members 0.015 −0.014 0.000 −0.091 −0.030 0.003 HH size 0.065 −0.027 −0.002 −0.207 −0.005 0.001 Wealth Index 1.409 0.098 0.138 2.434 0.104 0.253 Education (mother) 0.110 0.525 0.058 0.042 0.486 0.021 Education (father) 0.112 0.412 0.046 0.000 0.363 0.000 Residence (urban) −0.021 0.906 −0.019 −0.065 0.800 −0.052 Region . . . . . −0.005 Residual . . . . . 0.002 Total . . 0.344 . . 0.223 Notes: Child HH members are household members under 5 years of age. HH size is total household size. Regions are dummies for each region with one reference region (Tigray) omitted from the regressions (1)According to trend analysis, there has been (measles and full vaccination, contraception, ANC4 substantial progress in MNCH services and +, and SBA). This is encouraging. It suggests that outcomes over the study period. DHS data for further expansion in their coverage could substan- 2000–14 show that child undernutrition and tially reduce the remaining inequalities. However, we mortality have declined considerably and health also find that for the poor, health outcomes have services coverage has increased. Indeed, Ethiopia is worsened, and the gap in child malnutrition between one of the few countries that achieved its MDG4 rich and poor households has widened. Patterns in (child mortality reduction) three years ahead of childhood mortality may also have shifted from a schedule, by three years. As in many other modest pro-poor bias to a modest pro-rich one. This countries, that may partly be due to its economic finding on the disconnect between health services performance during the past decade and the related and health outcomes agrees with previous studies improvements in living conditions. A recent poverty for a number of developing countries [4]. However, assessment study found that in 11 years head-count the considerable decline in health services inequality poverty dropped by 24 percentage points, from 55% could be attributable to the country’s flagship Health in 2000 to 31% in 2011 [17]. Ethiopia’s health system Extension Program, which may have helped signifi- may also have made an indispensable contribution. cantly to making services more available, particularly In the period from 1997 to 2015, the Health Sector to the poor. The program deployed over 38,000 Development Programs (HSDP) allocated resources health extension workers (HEWs) to local communi- to priority health outcomes and made services lo- ties for health promotion and basic service delivery cally available through power devolution and expan- [19]. As a lower-cadre alternative in their own or sion of infrastructure and human resources [18]. neighboring communities, these HEWs reached poor (2)The results of the analyses of wealth-related inequal- Ethiopians more effectively than medical doctors ities are mixed. Over time, there has been a narrow- and nurses. However, because improving health and ing of wealth-based inequalities in health services nutritional outcomes among the poor requires Ambel et al. International Journal for Equity in Health (2017) 16:152 Page 14 of 16 Table 9 Decomposition of concentration index for maternal health services (2000–2014) 2000 2011/14 9a. Contraceptive Elasticity Conc. index Contribution Elasticity Conc. index Contribution Age −0.481 −0.019 0.009 −0.768 −0.006 0.004 Wealth index 1.028 0.099 0.102 0.780 0.101 0.079 Education (years) 0.067 0.562 0.038 0.062 0.432 0.027 Residence (urban) 0.152 0.817 0.124 0.067 0.636 0.043 Protestant −0.014 −0.044 0.001 0.013 −0.019 0.000 Muslim 0.046 −0.018 −0.001 −0.073 −0.041 0.003 No more children 0.275 0.020 0.006 0.096 0.002 0.000 Region . . −0.009 . . 0.003 Residual . . 0.227 . . 0.224 Total . . 0.496 . . 0.382 9b. ANC4+ Age 0.102 −0.020 −0.002 −0.404 −0.005 0.002 Wealth index 1.430 0.100 0.143 1.764 0.130 0.230 Education (years) 0.103 0.566 0.058 0.167 0.449 0.075 Residence (urban) 0.078 0.851 0.066 0.064 0.708 0.046 Region . . 0.021 . . 0.022 Residual . . 0.186 . . 0.092 Total . . 0.474 . . 0.467 9c. SBA Age −0.085 −0.019 0.002 −0.46 −0.002 0.001 Child HH members −0.33 −0.023 0.008 −0.476 −0.037 0.018 HH size 0.124 −0.029 −0.004 −0.324 0.005 −0.002 Wealth index 1.278 0.095 0.122 2.3 0.125 0.287 Education (years) 0.096 0.53 0.051 0.161 0.441 0.071 Residence (Urban) 0.087 0.856 0.074 0.06 0.681 0.041 Region . . 0.023 . . 0.001 Residual . . 0.389 . . 0.108 Total . . 0.665 . . 0.526 Notes: Child HH members are household members under 5 years of age. HH size is total household size. The latest survey for ANC4+ and SBA is 2014. The latest survey used in this table for Contraceptive is 2011. We used 2011 to add more family planning variables that were not included in the 2014 mini-DHS. Regions are dummies for each region with one reference region (Tigray) omitted from the regressions coordinated action across the health, education, agri- The correlations observed between household wealth and culture, and water sectors [20–22], it may take time child health outcomes or MNCH service use are driven in to reduce the related inequality. part by such intermediate factors as women’s education level (3)Despite the narrowing trend, there is still substantial and access to services like water and sanitation. However, inequality in health services, especially in ANC4 and the decomposition of inequalities found that even once SBA. For example, in the 2014 survey, 35% of the these factors are accounted for, household wealth still has a top 60% made four or more ANC visits compared to direct effect on health service use and outcomes, and the 11% of the poorer 40%. Similarly, 24% of women changes in inequality observed are driven mainly by the dir- from the top 60% of households had SBA, compared ect effects rather than the intermediate factors [21, 22]. In- to just 5% of women from the bottom 40%. In creasing coverage of ANC4 and SBA goes beyond just general, differences in favor of the rich (top 60%) are making services available; there are also concerns about the large for most of the outcomes studied. The quality of services (e.g., qualified staff, equipment, water and differences are even more pronounced between the electricity for laboratories and delivery). The Government of poorest quintile (bottom 20%) and the richest Ethiopia has started taking action in this area, for example (top 20%). upgrading HEW qualifications so that they can be the Ambel et al. International Journal for Equity in Health (2017) 16:152 Page 15 of 16 equivalent of midwives. There have also been efforts to in- Additional file ventory facility readiness and address bottlenecks. Over the long term, a continuous quality improvement system could Additional file 1: Table S1a. Detailed Definition of Indicators. Table S1b.Definition of Terms. (DOCX 58 kb) help to narrow the inequalities in health outcomes. Abbreviations Limitations ANC4+: Antenatal care, four or more visits; CPR: Contraceptive prevalence How the DHS is designed affects how the indicators are rate; DHS: Demographic and Health Survey; HEW: Health extension worker; IMR: Infant mortality rate; IUD: interauterine device; iup: intersection union constructed in ways that may affect the analysis of in- principle (dominance test); mca: multiple comparison approach (dominance equality. For example, because data on child vaccinations test); MDGs: Millennium Development Goals; MNCH: Maternal, newborn, and are collected as part of the women’s questionnaire, they child health; NMR: Neonatal mortality rate; SBA: Skilled birth attendant; SDGs: Sustainable Development Goals; U5MR: Under-5 mortality rate; are only available for children whose biological mother VCHW: Voluntary community health worker is alive and in the household. Similarly, because infant and child mortality rates are calculated from information Acknowledgements We thank Anne Grant for editorial assistance. The views expressed in this collected from women about all their births in the previ- paper are entirely those of the authors and do not represent those of the ous five years, it does not take into account deceased World Bank or any member country. children whose mothers died giving birth or subse- quently. In contrast, because data on height and weight Ethical approval and consent to participate The study used data from Demographic and Health Surveys (DHS) collected were collected for all the children in the household, they by the DHS program and Central Statistical Agency. Those who conducted should be fully representative. the surveys reported that ethical considerations have been addressed. For In the analysis, we do see significantly worse nutrition the 2000, 2005, and 2011 surveys, please refer to the DHS program at: http:// dhsprogram.com/What-We-Do/Protecting-the-Privacy-of-DHS-Survey- outcomes for children whose biological mother is not Respondents.cfm. For the 2014 mini-DHS survey please see consents in the alive or not in the household, suggesting that mortality report: http://www.unicef.org/ethiopia/Mini_DHS_2014__Final_Report.pdf. rates may be underestimated and immunization rates Funding overestimated. These biases would be worse for sub- Not applicable. groups with higher rates of maternal mortality or (in the case of immunizations) fostering; if the biases are large, Availability of data and materials they may skew the inequality analysis. The study used four Demographic and Health Surveys, conducted in 2000, 2005, 2011, and 2014. The data from the first three surveys are available at Another limitation is that because data on ANC and https://dhsprogram.com. Data for 2014 were obtained from the Central SBA are collected in the women’s questionnaire, they do Statistical Agency, Addis Ababa, Ethiopia. not include pregnancies and births that ended in the Authors’ contributions death of mother. We would expect that poor antenatal AAA, CA, and QK conceived the research. AAA and EMF analyzed the data and or delivery care would be a risk factor for maternal mor- drafted the report. AMB, CA, QK and HW reviewed and edited the draft report. tality; we could be overestimating ANC and SBA, again AAA, EMF and HW revised the final document. All authors read and approved the final manuscript. with possibly differential biases based on the rates of maternal mortality in various subgroups. Consent for publication Not applicable. Conclusions Competing interests Ethiopia’s recent progress in MNCH was the starting The authors declare that they have no competing interests. point for fuller examination of the trends for rich and poor, and how wealth-related health inequality changed Publisher’s Note over the last two decades. The results obtained using Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. various approaches led to similar conclusions: We found pro-rich inequality in certain health status outcomes but Author details 1 in general pro-poor progress in services. In addition, in The World Bank, 1818 H Street, Washington, DC 20046, USA. 2The World Bank, Addis Ababa, Ethiopia. both health status and services, there is still substantial wealth-related inequality. The decomposition exercise Received: 30 October 2016 Accepted: 14 August 2017 shows how certain socioeconomic status indicators, such as the wealth index and education, may help to explain References existing inequalities. Ethiopia’s efforts to improve access 1. United Nations. Delivering as one in Ethiopia. Addis Ababa, Ethiopia, 2012. to health services have shown some positive results, but http://et.one.un.org/content/dam/unct/ethiopia/docs/UN%20in%20Ethiopia_ now it may be necessary, to change outcomes for the One%20UN%20Book%20(3).pdf. Accessed 5 Apr 2017. 2. MDG Progress Index: Gauging country-level achievements. Center for Global poorest, to focus on service quality and cross-sectoral Development (CGDEV), 2011 http://www.cgdev.org/page/mdg-progress- interventions. index-gauging-country-level-achievements. Accessed 31 Jan 2015. Ambel et al. International Journal for Equity in Health (2017) 16:152 Page 16 of 16 3. Central Statistical Agency [Ethiopia]. Ethiopia Mini Demographic and Health Survey 2014. Addis Ababa, Ethiopia. https://www.unicef.org/ethiopia/Mini_ DHS_2014__Final_Report.pdf. Accessed 20 Apr 2017. 4. Wagstaff A, Bredenkamp C, Buisman LR. Progress on global health goals: are the poor being left behind. World Bank Res Obs. 2014;29(2):137–62. 5. Onarheim KH, Tessema S, Johansson KA, Eide KT, Norheim OF, Miljeteig I. Prioritizing child health interventions in Ethiopia: modeling impact on child mortality, life expectancy and inequality in age at death. PLoS One. 2012; 7(8):e41521. 6. Wilunda C, Putoto G, Manenti F, Castiglioni M, Azzimonti G, Edessa W, Criel B. Measuring equity in utilization of emergency obstetric care at Wolisso Hospital in Oromiya, Ethiopia: a cross sectional study. Int J Equity Health. 2013;12:27. 7. World Bank. Health equity and financial protection datasheet: Ethiopia. Washington, DC: World Bank; 2012. http://documents.worldbank.org/ curated/en/942691468256149274/Health-equity-and-financial-protection- datasheet-Ethiopia. Accessed 20 Apr 2017. 8. Yesuf EA, Calderon-Margalit R. Disparities in the use of antenatal care service in Ethiopia over a period of fifteen years. BMC Pregnancy Childbirth. 2013;13:131. 9. Mirkuzie AH. Exploring inequities in skilled care at birth among migrant population in a metropolitan city, Addis Ababa, Ethiopia; a qualitative study. Int J Equity Health. 2014;13(1):110. 10. Rutstein SO, Rojas G. Guide to DHS statistics. Demographic and Health Surveys. Calverton, Maryland: ORC Macro; 2006. https://dhsprogram.com/ pubs/pdf/DHSG1/Guide_to_DHS_Statistics_29Oct2012_DHSG1.pdf. Accessed 20 Apr 2017. 11. McKinnon B, Harper S, Kaufman J, Bergevin Y. Socioeconomic inequality in neonatal mortality in countries of low and middle income: a multicountry analysis. Lancet Glob Health. 2004;2(3):e165–73. 12. Moser K, Frost C, Leon DA. Comparing health inequalities across time and place raute ratios and rate differences lead to different conclusions: analysis of cross-sectional data from 22 countries 1991-2001. Int J Epidemiol. 2001; 36:1285–91. 13. Kakwani N, Wagstaff A, van Doorslaer E. Socioeconomic inequalities in health: measurement, computation and statistical inference. J Econ. 1997;77: 87–103. 14. Kjellsson G, Gerdtham UG. On correcting the concentration index for binary variables. J Health Econ. 2012;32:659–70. 15. O’Donnell O, van Doorslaer E, Wagstaff A, Lindelow M. Analyzing health equity using household survey data: a guide to techniques and their implementation. Washington, DC: World Bank; 2008. 16. Wagstaff A, van Doorslaer E, Watanabe N. On decomposing health sector inequalities, with an application to malnutrition inequalities in Vietnam. J Econometrics. 2003;112:207–23. 17. Khan Q, Faguet JP, Ambel A. Blending top-down federalism with bottom- up engagement to reduce inequality in Ethiopia. World Dev. 2017;96:326– 42. 18. World Bank Group. Ethiopia Poverty Assessment 2014. Washington, DC: World Bank. License: CC BY 3.0 IGO. 2016. https://openknowledge. worldbank.org/handle/10986/21323. Accessed 5 Apr 2017. 19. Wang H, Tesfaye R, Gandham NVR, Chekagn CT. Ethiopia health extension program: an institutionalized community approach for universal health coverage. Washington, DC: World Bank; 2016. 20. Ruel MT, Alderman H. Nutrition-sensitive interventions and programmes: how can they help to accelerate progress in improving maternal and child nutrition. Lancet. 2013;382(9891):536–51. 21. Amate-Fortes I, Guarnido-Rueda A, Molina-Morales A. Determinants of child Submit your next manuscript to BioMed Central health inequalities in developing countries: a new perspective. Soc. 2016;53: 641. doi:https://doi.org/10.1007/s12115-016-0072-y. and we will help you at every step: 22. Marmot M. Achieving health equity: from root causes to fair outcomes. • We accept pre-submission inquiries Lancet. 2007; 370:1153–63. doi: https://doi.org/10.1016/S0140- 6736%2807%2961385-3. • Our selector tool helps you to find the most relevant journal • We provide round the clock customer support • Convenient online submission • Thorough peer review • Inclusion in PubMed and all major indexing services • Maximum visibility for your research Submit your manuscript at www.biomedcentral.com/submit