Trends in inequality in the coverage of vitamin A supplementation among children 6–59 months of age over two decades in Ethiopia: Evidence from demographic and health surveys

Objectives: There is a dearth of evidence on inequalities in vitamin A supplementation in Ethiopia. The goal of this study was to assess the magnitude and overtime changes of inequalities in vitamin A supplementation among children aged 6–59 months in Ethiopia. Methods: We extracted data from four waves of the Ethiopia Demographic and Health Surveys (2000, 2005, 2011, and 2016). The analysis was carried out using the 2019 updated World Health Organization’s Health Equity Assessment Toolkit software that facilitates the use of stored data from World Health Organization’s Health Equity Monitor Database. We conducted analysis of inequality in vitamin A supplementation by five equity stratifiers: household economic status, educational status, place of residence, child’s sex, and subnational region. Four summary measures—population attributable fraction, ratio, difference, and population attributable risk—were assessed. We computed 95% uncertainty intervals for each point estimate to ascertain statistical significance of the observed vitamin A supplementation inequalities and overtime disparities. Results: The findings suggest marked absolute and relative pro-rich (population attributable fraction = 29.51, 95% uncertainty interval; 25.49–33.53, population attributable risk = 13.18, 95% uncertainty intervals; 11.38–14.98) and pro-urban (difference = 16.55, 95% uncertainty intervals; 11.23–21.87, population attributable fraction = 32.95, 95% uncertainty intervals; 32.12–33.78) inequalities. In addition, we found education-related (population attributable risk = 18.95, 95% uncertainty intervals; 18.22–19.67, ratio = 1.54, 95% uncertainty intervals; 1.37–1.71), and subnational regional (difference = 38.56, 95% uncertainty intervals; 29.57–47.54, ratio = 2.10, 95% uncertainty intervals; 1.66–2.54) inequalities that favored children from educated subgroups and those living in some regions such as Tigray. However, no sex-based inequalities were observed. While constant pattern was observed in subnational regional disparities, mixed but increasing patterns of socioeconomic and urban–rural inequalities were observed in the most recent surveys (2011–2016). Conclusion: In this study, we found extensive socioeconomic and geographic-based disparities that favored children from advantaged subgroups such as those whose mothers were educated, lived in the richest/richer households, resided in urban areas, and from regions like Tigray. Government policies and programs should prioritize underprivileged subpopulations and empower women as a means to increase national coverage and achieve universal accessibility of vitamin A supplementation.

It is a descriptive cross-sectional study and we used data from four waves of EDHSs (2000, 2005, 2011, and 2016). The EDHS is conducted with the financial support of the United State Agency for International Development (USAID) and technical assistance of United Nation International Children’s Emergency Fund (UNICEF). The EDHS is a nationally representative survey designed to collect data on various health topics such as nutrition, domestic violence and female genital mutilation, access to mass media, fertility, young child development, breastfeeding and food intake, vaccinations, and treatment of diseases. By providing the government of Ethiopia with valid and up-to-date health indicators on reproductive-aged women (15–49 years of age), men 15–59 years old, and children under 5, the survey aims to monitor and assess the health situation of the population. The sample for all the four EDHS was designed to provide population and health indicators at national (urban and rural) and regional levels. The sample design permitted for specific indicators, such as VAS, to be calculated for each of Ethiopia’s 11 geographic/administrative regions (the nine regional states and two city administrations).1,24–26 The 1994 population and housing census, conducted by the Central Statistical Agency (CSA), was provided the sampling frame for the 2000 and 2005 EDHSs, while the 2007 population and housing census used for the 2011 and 2016 EDHSs. Administratively, regions in Ethiopia are divided into zones, and zones into administrative units called weredas. Each wereda is further subdivided into the lowest administrative unit, called kebele. Each kebele was subdivided into census enumeration areas (EAs), which were convenient for the implementation of the census. An EA is a geographic area that covers an average of 181 households. All four EDHS samples used a stratified, two-stage cluster sampling design, and EAs were the sampling units for the first stage. Thus, 2348 clusters (672 urban and 1676 rural) were selected from the list of EA using proportional probability sampling (PPS) technique. In the second stage, households from each cluster were then systematically selected for participation in the survey. In all four surveys, a total of 65,112 households were selected for the sample, of which 62,180 were occupied. Of the occupied households, 61,145 were successfully interviewed. A total of 37,625 children aged 6–59 months were included. Detailed description of the sampling design and overall methodology of EDHSs is explained elsewhere.1,24–26 The EDHS data are collected usually every 5 years with the use of pretested validated quantitative tools and structured methodologies. Specifically for this study, data were collected using a questionnaire that included information on children’s identity (age, sex, and relationship to the main caregiver) and VAS coverage during the preceding 6 months. Caregivers were shown samples of VA capsules and were asked about their sociodemographic characteristics and their sources of information on vitamin A. The information is based on mother’s recall, health facility information (where available), and the vaccination card (where available). In case of missing information on the day of the supplementation date, we imputed 15 as the probable day. The rationale for selection of day 15 refers to its position in the middle of the month; therefore, it would minimize potential bias referring to the child’s age in the date of the supplementation.1,24–26 We measured inequality in VAS coverage, which is the primary outcome variable of interest for the study. VAS was assessed among living children aged 6–59 months who received vitamin A capsule or supplement 6 months before the interview. 27 In the EDHS, the mothers or caregivers are questioned on whether their child had received vitamin A capsule or not. If the child received the supplement, we coded the variable as “yes” and if not, we coded it as “no.” 27 We measured inequality in VAS coverage using five equity stratifiers, namely, household’s economic status, educational status, child’s sex, place of residence, and subnational region. Household economic status was measured using the Demographic and Health Survey (DHS) wealth index, which is customarily calculated by considering the possession of durable goods, household characteristics and availability of basic household facilities following the methodology explained elsewhere. 28 For all the surveys, the commonly used variables took account of possession of a car, motorcycle, bicycle, electricity, television, radio, and material used for constructing wall, roof and floor of household house, water and hygiene and sanitation facilities.28,29 The constructed wealth index is then divided into five quintiles, namely, wealth quintile 1 (poorest), wealth quintile 2 (poorer), wealth quintile 3 (middle), wealth quintile 4 (richer), and wealth quintile 5 (richest). The other equity stratifiers are maternal education (coded as no education, primary school, secondary school and above), place of residence (coded as rural and urban), child’s sex (coded as male and female), and subnational region (coded into nine regions and two city administration: Tigray, Afar, Amhara, Oromiya, Somali, Benishangul-Gumuz, South Nation and Nationalities People (SNNP), Gambela, Harari, Addis Ababa, and Dire Dawa). The analysis was done using HEAT software that is recommended by World Health Organization (WHO) for investigation of health inequalities and it is available offline after installation. The detailed description of the software is available elsewhere.30,31 In brief, the HEAT is software that enables the examination and analysis of health inequalities across and within countries. The software is valuable for exploring the health disparity situation in a systematic manner. The HEAT software application comprises the WHO Health Equity Monitor (HEM) database. 32 The database has large sets of data from Multiple Indicator Cluster Survey (MICS) and DHS which are carried out in several LMICs including Ethiopia. Currently, the HEM database comprises more than 30 maternal, neonatal, child and reproductive health indicators. The analysis included two key steps. First, disaggregation of VAS was made using the above-mentioned five equity stratifiers. Following the disaggregation, VAS inequality was further analyzed using the four summary measures, namely, population attributable fraction (PAF), population attributable risk (PAR), difference (D), and ratio (R). The selection of the summary measures for an inequality study should be based on the fact that, the selected summary measures need to be of simple and complex measures. 33 At the same time, summary measures need to be relative and absolute measures to be able to examine inequality from different angles. For our study, we selected measures of inequality by these recommendations. PAF and PAR are complex measures, whereas R and D are simple measures. 33 In addition, PAF and R are relative summary measures, whereas PAR and D are absolute summary measures. Simple measures make pairwise comparisons of health between two subgroups, such as the most and least wealthy. 33 Simple pairwise comparisons have historically been the dominant type of measurement used in inequality monitoring, as their simplicity makes them intuitive and easily understood. Complex measurements, on the contrary, make use of data from all subgroups to assess inequality. 33 When describing the inequality in a health indicator by region, for instance, pairwise comparisons can be used to describe the inequality between two selected regions—such as worst versus best—whereas complex measures could describe the inequality that exists among all regions. While pairwise comparisons of inequality have certain limitations that complex measures overcome, they will be described here at length as they play an important role in inequality monitoring. Because they are straightforward, and they are preferable over complex measures in situations where complex measures do not present a substantially improved picture of inequality. 33 Detailed description of the calculation of each of the summary measures used in this study is explained elsewhere,33,34 but we have highlighted here a summary. D is calculated by subtracting two subgroups (D = Yhigh − Ylow), where Yhigh represents richest, secondary school and above, female, and urban and Ylow represents poorest, no education, male, and rural for economic, education, sex, and place of residence dimensions, respectively. Similar calculation was applied for subregions (region with highest VAS coverage minus lowest VAS coverage). The calculation of R in all five dimensions of inequalities is similar with D except that R is dividing subgroups with highest VAS coverage with subgroups with lowest VAS coverage as follows: R is calculated by dividing two subgroups (R = Yhigh/Ylow), where Yhigh represents richest, secondary school and above, female, and urban and Ylow represents poorest, no education, male, and rural for economic, education, sex and place of residence dimensions, respectively. PAR was calculated by subtracting the national average of VAS coverage from the reference subgroups. The reference subgroups for economic, education, sex, and place of residence are richest, secondary school and above, female, and urban, respectively. For subnational, the reference subgroups are region with the highest VAS coverage in each survey. PAF is computed by dividing the PAR with the national average of VAS coverage (μ) and multiplied by 100. An inequality in VAS coverage is nonexistent if the D included zero and R included 1. PAF and PAR take the value of zero if no inequalities or same levels are recorded across subgroups. The greater the value of PAF and PAR from zero, the higher the inequality. The value of PAF and PAR indicates the potential improvement in the national coverage of VAS coverage if the VAS coverage reached the same level across subgroups or no inequality across subgroups. A positive value of PAR and PAF indicates higher concentration of coverage of VAS coverage among advantaged subpopulations such as richest, secondary school and above, female children and urban residents as well as regions with the highest VAS coverage. Inequality trends were assessed in caution and by referring to the uncertainty intervals (UI) of each summary measure of different surveys. That means if the UIs did not overlap, there were increasing or decreasing changes, but the overlapping of UIs was considered a constant pattern. However, the small and large overlapping was not treated equally and authors considered this important concept during interpretations of trends. For enabling this study’s quality of evidence, we followed the guideline indicated in strengthening Reporting of Observational Studies in Epidemiology (STROBE). 35 For the analysis of this study, ethical approval was not sought for the present study because authors utilized already existing secondary data. All ethical procedures that were followed by the custodians of the data have been reported in the manuscript. In addition, the University of Ottawa’s Office of Research Ethics and Integrity stated that “no ethics review is required for the use of previously collected, publicly available, anonymously collected data” (https://research.uottawa.ca/ethics/submission-and-review/types-review).

N/A

• zegeye, betregiorgis
• olorunsaiye, comfort z.
• ahinkorah, bright opoku
• ameyaw, edward kwabena
• seidu, abdul aziz
• budu, eugene
• yaya, sanni