Data source
Data used for this study emerged from the 1998, 2003, 2008, and 2014 Ghana Demographic and Health Surveys (DHS) which were executed by the Measure DHS Program. The DHS Program conducts these surveys for several low- and middle-income countries. The DHS generally collates information on child and maternal health, including issues on stunting. Surveys of the DHS Program such as the ones included in this study utilize a twofold sampling procedure. In Ghana, the initial stage constituted the systematic selection of clusters/enumeration areas within rural and urban settings of Ghana. The second stage was characterized by household selection within the enumeration areas that were selected in stage one. Eligible women (permanent residents and those who joined the households the night before the survey) were subsequently interviewed [7]. A total of 10,826 (2,529 in 1998; 2,874 in 2003; 2,529 in 2008; 2,894 in 2014) women were included in this present study.
Variables
Dependent variable
Stunting among children below five years was our dependent variable. The dependent variable, stunting, was measured by the DHS by using the WHO Child Growth Standards [15]. As such, the DHS collected data on each child’s height/length, sex, and age to compute the number of standard deviations (i.e. z-score) to determine whether the child’s height is above or below the median of the 2006 WHO growth reference population. Stunting, therefore, was defined as a z-score less than -2 and not stunted if otherwise [15].
Independent variables
A total of four inequality stratifiers served as the independent variables for this study. The first is economic status, which was computed using the Principal Component Analysis (PCA). It was derived by segregating households into five levels based on ownership of some cardinal assets (e.g. television and bicycle) and dwelling characteristics (e.g. floor and roofing material). The second is education, measured as no education, primary education, and secondary education or higher. The third is the place of residence (rural or urban) whilst the sex of the child was either male or female. Finally, the region of residence comprised all the then ten administrative regions of Ghana.
Analysis
We analysed the data with the assistance of the 2019 updated version of WHO’s HEAT software via the WHO Health Equity Monitor database [16]. Four principal summary measures were employed in our analysis. These four measures are Ratio (R), Difference (D), Population Attributable Fraction (PAF), and Population Attributable Risk (PAR). These four distinct summary measures were employed due to the recommendation by WHO that the usage of different summary measures helps to generate outcomes that are sensitive to policy formulation [17].
In deriving our summary measures, the following procedures were followed. Concerning economic status, D was computed by subtracting stunting prevalence among children of poorest women (quintile 1) from the stunting prevalence among children of richest women (quintile 5). On education, stunting prevalence among the children of women without education was subtracted from the prevalence among children of women who had secondary/higher education. Similarly, prevalence among children of rural residents was subtracted from the prevalence among children of urban residents. The D for the region variable was calculated as the region with the maximum prevalence of stunting minus the region with the minimum prevalence across the respective surveys.
In computing R, it was calculated as the ratio of two different populations, i.e. R=Yhigh/Ylow. In the case of a residence, R denoted a place of residence where Yhigh stood for urban and Ylow stood for the rural population. With education, Yhigh implied children of women with secondary or higher education where Ylow represented children of women with no education. For wealth quintile, Yhigh was the richest quintile whilst Ylow was the poorest quintile. In the same manner, Yhigh represented males or females, depending on the category with the highest prevalence in a particular survey. The PAR was derived by ascertaining the difference between estimates of the reference sub-populations (known as yref) of the national mean of stunting among children under 5. With this, (μ): PAR = yref-μ, with μ being the national mean of stunting. Likewise, PAF denoted the relative inequality dimension of PAR and it was derived as PAF = (PAR/ μ)*100.
Ethical issues
Data used for this study is a publicly available de-identified data. Ethical approval for Ghana DHS was granted by the Ghana Health Service and the Ethical Review Board of the Measure DHS. All participants consented either in writing or verbally before participating in the surveys.