Inequalities and trends in Neonatal Mortality Rate (NMR) in Ethiopia: Evidence from the Ethiopia Demographic and Health Surveys, 2000–2016

Background Substantial inequality in neonatal mortality rates (NMR) remains in low- and middle-income countries to the detriment of disadvantaged subpopulations. In Ethiopia, there is a dearth of evidence on the extent and trends of disparity in NMR. This study assessed the socioeconomic, residence and sex-based inequalities in NMR, as well as examined its change over a sixteen year period in Ethiopia. Methods Using the World Health Organization’s (WHO) Health Equity Assessment Toolkit (HEAT) software, data from the Ethiopia Demographic and Health Surveys (EDHS) were analyzed between 2000 and 2016. NMR was disaggregated by four equity stratifiers: education, wealth, residence and sex. In addition, absolute and relative inequality measures, namely Difference, Population Attributable Risk (PAR), Ratio, Relative Concentration Index (RCI) and Slope Index of Inequality (SII) were calculated to understand inequalities from different perspectives. Corresponding 95% Uncertainty Intervals (UIs) were computed to measure statistical significance. Findings Large educational inequalities in NMR were found in 2000, 2005, and 2011, while wealth-driven inequality occurred in 2011. Sex disparity was noted in all the surveys, and urban-rural differentials remained in all the surveys except in 2016. While socioeconomic and area-related inequalities decreased over time, sex related inequality did not change during the period of study. Conclusions NMR appeared to be concentrated among male newborns, neonates born to illiterate and poor women and those living in rural settings. However, the inequality narrowed over time. Interventions appropriate for different subpopulations need to be designed.

This study aims to assess socioeconomic, residence and sex-based inequalities in NMR, and the change over the course of roughly two decades in Ethiopia.  (1) where interventions needed to avoid neonatal mortality are still low (2)(3)(4). Expansions of effective interventions to all who need it improve survivals of neonates, however, launching strategies necessary to meet this aim remain a challenge (5).

Methods
Literature shows that socioeconomic inequity in neonatal deaths tend to occur in disadvantaged subpopulations with in a country (5). The majority of neonatal deaths tend to take place in Low and Middle Income Countries (LMIC) (6), disfavoring the poor, illiterate and rural communities (5, 7); neonates born to women in higher socioeconomic positions have better survival rates. However, in most of the LMICs, the inequality in NMR had decreased over the last 20 years (5). As the global community works toward an equity-oriented international agenda to meet the United Nations' Sustainable Development Goals (SDG), it is imperative that interventions reach these subpopulations in order to remove existing NMR disparities. Evidence is necessary to outline the NMR, who is affected most, and how it is changing over time. This, in turn, will inform policies and strategies that target sub-groups of populations who suffer most from high NMR.
In 2018, nearly 100,000 babies died within the first 28 days following birth in Ethiopia wealth, education, residence and sex) and how the disparity has changed over the last two decades. In fact, few studies have attempted to assess NMR inequality in Ethiopia (5,7,8). This study enriched the available evidence in a number of areas. First, it investigated NMR inequality through the rigorous method recommended by the World Health Organization (WHO) Handbook on health inequality monitoring. This method considers different dimensions of inequality measures in the analysis to gain a better understanding of the NMR disparity (9). Commonly used absolute (difference) and relative (ratio) simple measures of inequality were calculated to reflect the magnitude of the difference in NMR between studied subgroups and how inequality changed over time (9). In addition to the simple measures, which allow comparison between two subgroups only, the study also estimated the inequality using measures that account for all sub-groups of an equity stratifier (9). Secondly, it uses the common equity stratifiers recommended by the WHO (9). This is in contrast to most prior studies that limited NMR inequality analysis to just socioeconomic dimensions such as wealth and education (5,7). Third, most previous studies used traditional regression methods not suitable for inequality studies (8,10,11) or were limited to certain specific areas in the country without reflecting aspects of the inequality at the national context (11,12).

Study Setting
With 109 million people, Ethiopia is the most populous country in Africa second only to Nigeria (14). Evidence suggests that Ethiopia experienced consistent economic growth over the last decade, and the agricultural sector has been a major driver of that growth were used. The methodology DHS follows has been described elsewhere (23)(24)(25)(26).
Briefly, a two-stage cluster design was employed to select women aged 15 to 49 years.

Statistical analysis
The inequality in NMR was examined in two steps. First, the NMR was disaggregated by the equity stratifiers mentioned above. Secondly, the inequality was assessed using Whereas Ratio is a relative measure, the remaining are absolute summary measures.
The choice of summary measures is in compliance with evidence suggesting the scientific importance of adopting both absolute and relative summary measures in a single health inequality study (9). The main reason being that, relative and absolute inequality measures could potentially lead to different, even contrasting conclusions (9), and failing to showcase these different scenarios can bias informed decisions. Complex measures account for size of categories of a sub-population, unlike simple measures.
When a population shift is likely to occur, especially when trend analysis is an aim in the study, complex measures are likely to reflect the true change in equality over time (9).
On the other hand, simple measures are easy for interpretation and understanding.
Therefore, an inequality study should combine both simple and complex, as well as relative and absolute measures to provide a more comprehensive analysis.
The WHO's HEAT version 3.1 software was used for the analysis (28). The software is available at:https://www.who.int/gho/health_equity/assessment_toolkit/en/]. Interested researchers can download and use the software. The procedures followed for calculating summary measures are discussed in the HEAT software technical notes (28), and in the WHO handbook on the health inequality monitoring (9). Therefore, only a short account is provided here. For education and economic status, Difference was calculated as NMR in the poorest group minus and NMR in the richest group. For educational status, Difference in NMR between the no education group and secondary education group was conducted. Similarly, for the place of residence, Difference pertains to that between rural and urban populations. PAR was computed as the difference between NMR estimate for the reference subgroup, yref, and the national average of NMR. In this study, yref refers to the following to calculate NMR inequality for PAR: rural setting for a place of residence, female for sex, secondary education for education and richest sub-groups for economic status. While zero indicates absence of inequality, the greater absolute value of PAR indicates a higher level of inequality. The SII was computed through a generalized linear regression model. The computation was restricted to two dimensions (education and economic status) and required ranking of a weighted sample of the whole population in order from the most disadvantaged (rank 0) to the most advantaged (rank 1) subgroups. The poorest and uneducated individuals were considered the most disadvantaged, while those that had completed secondary education and the richest subgroups were deemed most advantaged. NMR was

Ethical Consideration
The analyses were completed using the publicly available data from DHS. Ethical procedures were the responsibility of the institutions that commissioned, funded, or managed the surveys. All DHS surveys are approved by ICF international as well as an Institutional Review Board (IRB) in the respective country to ensure that the protocols are in compliance with the U.S. Department of Health and Human Services regulations for the protection of human subjects. Table 1 presents NMR disaggregated by the four dimensions of inequality for each of the four EDHS. In 2000, the highest NMR was recorded among the poor, middle and rich sub-groups, whereas the lowest was observed among the poorest and richest. This uneven distribution of NMR resulted in an inverted U shape of wealth-related inequality.

Results
In 2016, NMR was disproportionately high amongst the rich, with no statistically significant differences between the other four sub-categories. Those who completed secondary education experienced the lowest NMR in 2000 and 2005. However, in 2016, NMR was more evenly distributed across all the three sub-groups of education. Across all surveyed years, a significant gender related disparity persisted with the disadvantage for female neonates. We did not identify a significant difference in NMR between urban and rural areas for all the survey periods. See table 1 for details.   (Table 2).
Both the relative and absolute measures indicated that NMR appeared to be largely concentrated among infants of women with no formal education. R showed, for instance, that in 2011, NMR among the uneducated women was about 2 times higher than that of women who completed secondary education. PAR and RCI also showed clear disparity in NMR, again disfavoring the illiterate sub-group. By contrast, estimates of the SII revealed that NMR was largely prevalent among women with higher educational attainment in both 2000 and 2011. Overall, educational inequality in NMR had been persistently falling across all measures over time ( Table 2).
While no urban-rural differentials of NMR was documented using the simple measures, estimates of the PAR suggest that NMR was more likely to dominate rural settings in the first three time points but disappears in 2016. Other notable finding was the significant sex disparities; NMR was more prevalent among male children in all the surveyed time periods by all the measures (Table 2).

Discussion
To our knowledge, this is the first attempt to comprehensively evaluate the nature of  births and endured about twice as much deaths when compared with female children.
Unlike the other equity stratifiers based inequality, sex based disparity did not improve with time. Other studies have found similar NMR inequality results with an advantage for female as compared to male neonates (31). Why, then, are male neonates at higher risk of dying than female neonate? Yet another aspect to explore in future studies.
The strength of this study is two-fold. First, inequality in NMR was assessed through different summary measures; adopting a number of inequality measures in the study would help showcase all possible patterns of NMR inequality. Second, the findings use an established high quality WHO monitor database. The database has been prepared by experts proficient in the area and we believe this contributed to the quality of evidence analyzed and reported in this paper.
However, the study also suffered some limitations. It did not identify factors that contribute to the measured NMR inequality. Future studies consider using a decomposition approach to study factors that contribute to the NMR inequality, and to see whether their contribution changes with time. Furthermore, inequality in NMR should be conducted in smaller areas such as villages, towns, districts and zones as the EDHS dataset-based findings cannot be applied to areas smaller than the subnational regions and city administrations.

Conclusions
The study found substantial amount of inequalities in NMR across the different equity stratifiers. Encouragingly, however, NMR inequality in Ethiopia has been on the decline according to almost 20 years of studied data. The improvement in inequality of NMR overtime can be associated with many factors such as Ethiopia's dedication to meet the NMR SDG of 12 deaths per 1000 newborns by 2030, improved economic growth, and improved coverage of maternal and child health services. Inequality studies such as this are useful to inform equity-oriented interventions aimed at eliminating inequality; they signal affected sub groups to strategically target the plan and alleviate the problem.
Policy makers should prioritize the subpopulations experiencing the highest NMR without ignoring the whole population. Despite gains made so far, much work is still needed to minimize, if possible eliminate, NMR disparity in across different equity stratifiers.