Sources of Data
Zimbabwe Demographic and Health Surveys (ZDHS) of 2010\11 and 2015 were used for analyses. Both data sets had population samples of 2,666 and 2,708 under-five children respectively, aged 0-59 months. Both 2010/11 and 2015, ZDHS samples were nationally representative composed of more than 11,000 households [35, 36]. The data sets gave representative information for most indicators in Zimbabwe for urban and rural areas [35–37].
The samples were representative of each of Zimbabwe’s ten provinces: Manicaland, Mashonaland Central, Mashonaland East, Mashonaland West, Matabeleland North, Matabeleland South, Midlands, Masvingo, Harare, and Bulawayo. The sampling frame for the 2002 and 2012 population census were used in both data sets [35–37].
Outcome variables
The study assessed two indicators of child health in under-fives. The two indicators were endorsed by countries represented at the UN Statistical Commission to monitor target 2.1. Prevalence of undernourishment (malnutrition) and prevalence of severe food insecurity in the population were outcome variables [33].
Food insecurity among under-five children was determined using the WHO dietary diversity score, which is based on the Infant And Young Child Feeding (IYCF) practices. Dietary diversity is the number of different foods or food groups consumed over a given reference period [38].
For this study 13 food groups were considered, namely food from grains, food from tubers, eggs, meat, pumpkin & carrots, green leafy vegetables, vitamin A fruits, other fruits, liver & heart, fish, (beans, peas, lentils, nuts), other milk products and yogurt. The IYCF tool defines minimum dietary diversity as indicator for food security by a cut- off point of >4 [38], in this study children with less than 3 of the 13 food groups were defined as food insecure. Children feeding responses in both surveys solemnly rely on the 24-hour recall method, hence results on food security are prone to recall bias.
Malnutrition was assessed using the child anthropometric measure of weight-for-age. Weight-for-age is a composite index of height-for-age and weight-for-height hence takes account of both acute and chronic under-nutrition. Children whose weight-for-age z-score was below minus two standard deviations (-2 SD) from the median were considered malnourished. Chi-square tests were used to assess the difference between food security status; nutritional status and socioeconomic classes, residence status, child age, and other background characteristics.
Socioeconomic status
Socioeconomic status was adapted from the wealth index of households in the original surveys (ZDHS) [35, 36]. In both ZDHS’s, wealth index was reported as scores based on the number and kinds of consumer goods owned, ranging from a television to a bicycle or a car, plus housing characteristics such as source of drinking water, toilet facilities, and flooring materials [39]. The latter scores were derived using principal component analysis [39].
National wealth quintiles were compiled by assigning household scores, then each person was ranked in the household population by their score, and lastly the distribution was divided into five equal categories, each with 20 percent of the population in the original studies [35, 36]. For this study socioeconomic status was then re-categorised from 5 (poorer, poor, middle, richer, richest) groups into 3 groups thus poor, middle and rich (Table 1).
Child age
The study focused on children under-five years of age (0-59months). Children’s age was then recoded into 3 groups based on the South Africa’s department of health age definitions [40]. The 3 child age groups were defined as; Neonates (1 day-1 month), Infants (1 month-24 months) and Young children (24 months-59 months) (Table 1).
Mother’s education
Zimbabwe’s education system is composed of 3 levels; primary education, secondary education and tertiary education [41]. The primary level is a seven-year cycle with an official entry age of six years running from Grade 1 to 7. However, prior to Grade 1 children are enrolled for early childhood education and care (preschool) for a year, but the latter is not formally considered as part of primary education.
Tertiary education in Zimbabwe covers all universities, technical colleges, polytechnic colleges, teacher’s training colleges and other vocational skills training canters [41]. Mother’s education was recoded into 3 categories thus; no education, primary and tertiary educated (Table 1).
Erreygers Normalised Concentration index
The study used Erreygers normalised concentration indices [42], in determining socio-economic inequalities in child nutrition and food security. The study adopted the latter approach as the concentration index approach does not entirely measure inequalities in ordinal health variables [42, 43]. The latter index is expressed as a value of a health variable which would have been assigned to an individual as a function of a socioeconomic category to which the individual belongs [44].
Concentration index is a mathematical derivative of the concentration curve. On the concentration curve, the x-axis represents cumulative proportion of individuals by socioeconomic class starting with the lowest socioeconomic class (poorest) and ending with highest socioeconomic class (richest), while the y-axis is the cumulative total proportion of health in these individuals [45].
The Concentration curve identifies the existence of socioeconomic inequalities in health sector/outcome variables, and is only sensitive to relative inequality [46]. The bounds of this measure are -1 and 1 with a negative (positive) value representing inequality favouring the worse-off (better-off).
Erreygers normalised index can be expressed algebraically as;
Where;
Zi represents number of individuals in a given population
i denotes the socioeconomic rank of the individual ranging from the richest to the poorest
h represents the health situation of the whole population
Theil index
The study used a generalized entropy measure known as the Theil index, mainly because of its decomposability [47]. The entropy measure is well-suited for estimating the contribution of different groups to total inequality [1]. Unlike other measures of inequalities like Gini index or The index of dissimilarity, generalised entropy class measures satisfy the five standard criteria for measuring inequalities including the attractive property of being easily decomposable by subgroups [42]. The Theil index is argued to be a measure of inequalities with distinctive properties, hence making it a powerful instrument in analysing patterns and dynamics of inequalities [48].
Generalised Entropy (GE) measures are cited to be based on the idea of divergence between probability distributions derived from information theory [47, 49–51]. Inequality decomposition was done by population subgroup to separate total inequality in the distribution into components of inequalities between the selected groups and the remaining within-group inequality. The study focused on decomposition by place of residence (rural/urban) and socioeconomic status (poor, middle and rich).
For this study we used the syntax ineqdec0, which is a stripped-down version the syntax ineqdeco in Stata version 13.1. The study used ineqdec0 syntax so as to include zeros and negative incomes in calculations. Theoretically, Theil index ranges from 0 to infinity, with 0 being a state of equal distribution and values greater than 0 representing increasing levels of inequality [52, 53]. Data analysis was done using Stata version 13.1 (Stata Corp, Texas, United States).