Page 21 - IJPS-9-2

International Journal of
Population Studies Dominant drivers of inequalities in child survival

and the review of literature and model building procedures. Furthermore, to determine the relative importance of
The review of most recent literatures (Bras & Mandemakers, each predictor to the outcome, we used the dominance
2022; Rebouças et al., 2022) on the subject indicated that analysis (Azen & Traxel, 2009) by employing a user-
these variables are most frequently reported predictors of developed STATA command “domin” (Luchman, 2021).
child survival in Ethiopia. The five internationally accepted Dominance analysis is a technique developed to estimate
dimensions of child survival inequalities were used as relative importance of all predictors in a logistic regression
predictor variables in this paper to approximately measure model in relation to an outcome variable (Tonidandel &
inequality drivers (or as proxy measures of inequality LeBreton, 2010). It relies on estimating the regression
drivers). Henceforth, the predictor variables are referred values of all possible combinations of predictors and
to as inequality drivers. Table 1 presents the coding for the measures relative importance by doing comparisons of
predictor variables. Household asset-based wealth index was all predictors in the model as they relate to an outcome
categorized into five quintiles and regrouped as poor (the variable (Tighe and Schatschneider, 2014). The method
first two quintiles: poorest and poorer) and non-poor (the also allows us in identifying the relative dominance of the
last three quintiles: middle, richer and richest). Education predictors (Azen & Traxel, 2009; Lee & Dahinten, 2021).
was used to reflect the level of education attained by a child’s We also conducted sensitivity analysis to predict the
mother, and grouped into two subgroups: no education and outcome of a decision given a certain range of variables
primary and above. Each of the place of residence (rural or (Appendix B). All analyses were weighted and conducted
urban) and child sex (female or male) was classified into using STATA version 15.
two subgroups. The old nine regional states and the two city
administrations were regrouped into three as emerging (Afar, 3. Results
Somali, Benishangul-Gumuz, and Gambela), established 3.1. Background characteristics of the study
(Amhara, Oromia, SNNP, Tigray and Harari), and central participants
(Addis Ababa and Dire Dawa city administrations) (Bareke
et al., 2022) (Table 1). Table 2 presents the background characteristics of the
study participants by inequality dimensions. More than
2.3. Statistical analysis half (52.64%) of children were born to mothers residing in
established regions, and majority (82.88%) of children were
To reduce bias, about 3.2% of children with missing height/ born in rural areas. More than half (51.21%) of children
length, weight, and unknown responses were excluded
from the analysis of undernutrition. For childhood were born in households grouped as poor wealth index.
anemia analysis, data from children aged 6 – 59 months Table 2 also shows that most (70.63%) of the children were
were used. Descriptive statistics were used to describe the born to uneducated mothers, and a little more than half
background characteristics of the study participants and (51.25%) of the children were male. About 51% and 55.06%
the key variables. A correlation-based assessment was used of children were undernourished and anemic, respectively,
and more than 8% of children were reported to have died
to detect multicollinearity, and an absolute correlation (Table 2).
coefficient of less than 0.6 was observed among predictors
indicating the absence of multicollinearity (Senaviratna 3.2. Bivariate logistic regression results
& Cooray, 2019). Logistic regression was used to identify
significant drivers of inequality in child survival. Bivariate Table 3 presents results from the bivariate logistic
logistic regression analysis was used to establish the regression analysis (with adjusted odds ratio) of the
strength of the relationship between inequality drivers association between the outcome variables and inequality
and outcome variables. Chi-square tests were computed to drivers. Table 3 shows that all socioeconomic (place
verify the significant association. of residence, household wealth index and maternal
education), geographic (region), and biological (sex of
child) inequality drivers were significantly associated
Table 1. Coding of proxy measures for the inequality drivers
with childhood undernutrition (at p < 0.001) and U5M
Inequality drivers Description (at p < 0.05). Likewise, the region, place of residence,
Sex of child 1=Female; 2=Male household wealth index and maternal education status had
Maternal education 1=No education; 2=Primary+ statistically significant association with childhood anemia.
Household wealth index 1=Poor; 2=Non-poor The bivariate logistic regression analysis finding revealed
that regional category, place of residence, household
Place of residence 1=Rural; 2=Urban
wealth index maternal education, and child sex are
Regional category 1=Emerging; 2=Established; 3=Central potential and significant drivers of inequality in childhood

Volume 9 Issue 2 (2023) 15 https://doi.org/10.36922/ijps.427