Menu
Menu
Menu
Menu
Background: A single anthropometric index such as stunting, wasting, or underweight does not show the holistic picture of under-five children’s undernutrition status. To alleviate this problem, we adopted a multifaceted single index known as the composite index for anthropometric failure (CIAF). Using this undernutrition index, we investigated the disparities of Ethiopian under-five children’s undernutrition status in space and time. Methods: Data for analysis were extracted from the Ethiopian Demographic and Health Surveys (EDHSs). The space–time dynamics models were formulated to explore the effects of different covariates on undernutrition among children under five in 72 administrative zones in Ethiopia. Results: The general nested spatial–temporal dynamic model with spatial and temporal lags autoregressive components was found to be the most adequate (AIC = -409.33, R2 = 96.01) model. According to the model results, the increase in the percentage of breastfeeding mothers in the zone decreases the CIAF rates of children in the zone. Similarly, the increase in the percentages of parental education, and mothers’ nutritional status in the zones decreases the CIAF rate in the zone. On the hand, increased percentages of households with unimproved water access, unimproved sanitation facilities, deprivation of women’s autonomy, unemployment of women, and lower wealth index contributed to the increased CIAF rate in the zone. Conclusion: The CIAF risk factors are spatially and temporally correlated across 72 administrative zones in Ethiopia. There exist geographical differences in CIAF among the zones, which are influenced by spatial neighborhoods of the zone and temporal lags within the zone. Hence these findings emphasize the need to take the spatial neighborhood and historical/temporal contexts into account when planning CIAF prevention.
Data for the analysis was drawn from 72 administrative zones in Ethiopia. Ethiopia is located in East Africa (Fig. 1), with a total land area of 1.1 million km2. The country has 11 national regions and 72 administrative divisions (zones). Locations of the 72 administrative divisions (zones) of Ethiopia: a Regions; b administrative zones of the study area (Source: Authors) The country has undertaken several economic development programs across regions and zones for eradicating undernutrition, poverty, hunger, illiteracy, and infant and maternal mortality, among others. Despite all these efforts by the concerned bodies, there are economic or poverty disparities and inequalities between the different administrative zones of Ethiopia [34]. We used the secondary Ethiopian Demographic and Health Survey (EDHS). There are several EDHS datasets and for this study, we used birth history records. A total of 30,791 children consisting of 8,765 from 2016, 9,611 from 2011, 3,850 from 2005, and 8,565 from the 2000 EDHS respectively were plausible for analysis. In this study, the zones are the spatial unit of analysis [13]. The outcome variable in this study was the proportion of CIAF for the zones [34]. Most of the previous studies on the prevalence of undernutrition in Ethiopia have focused on a single conventional anthropometric index of stunting, underweight, or wasting [4–8, 12, 19–21], separately proposed by the World Health Organization (WHO) [10]. However, these conventional indices of undernutrition may overlap so that the same child could show signs of having two or more of the indicators simultaneously; insufficient for determining the overall real burden of undernutrition situations among under-five children [5–7, 11–18]. The CIAF is computed by grouping those children whose height and weight were above the age-specific norm (above -2 z-scores) and those children whose height and weight for their age are below the norm and those who are experiencing one or more forms of anthropometric failure as express as B-wasting only, C-wasting and underweight, D- wasting, stunting and underweight, E- stunting and underweight, F-stunting only and Y- underweight only. The CIAF is then calculated by aggregating these six (B-Y) categories [16, 18, 27–29]. The choice of the covariates is guided by existing literature to study the determinants of child undernutrition in developing countries [4, 8, 10, 35]. In this paper, these explanatory variables considered in this study are also measured at the zone level. The zone-specific information on children, and households, such as the availability of improved drinking water, the percentage of literate mothers, the proportion of working mothers, and the percentage of households having access to drainage and sanitation facilities in the zones, was modeled with CIAF. The variables have been classified into the following categories: child, maternal, household, and geographic variables (Table (Table11). The description of the covariates included in the model Different studies [1–5] showed that children from “arid” geographic areas were associated with undernutrition. In Ethiopia, we wanted to see the impacts of the change of geographical covariates on undernutrition [3–5]. This is because of frequent and severe shortfalls in precipitation, and continuous rises in temperature, which may result in food insecurity, droughts, and undernutrition. Furthermore, more than three-quarters of Ethiopians depend on subsistence and rain-fed farming, livestock production that is historically linked to low crop production, and less diversified and commercial foods. Therefore we have extracted the geospatial covariates from the GPS dataset of the demographic and health survey data and this is joined with the DHS row dataset. Finally, we successfully modeled the CIAF at the zonal level by using both the EDHS and geospatial covariates. The classical linear models estimated by ordinary least squares methods cannot take into account the fact that data collected based upon spatial and time specifications is not independent of its spatial location across different periods. If the spatial and temporal effects are neglected in the model, the estimated values will be biased [4–11, 36–40]. Observations available across space (N spaces) and time (T time points), a range of different model specifications need to be considered to allow different combinations of the two cases. Let yt denote an NT × 1 column vector of observations on the dependent variable with spatial units (i = 1,2,…, N) and temporal units (t = 1,2,…, T), X be an NT × k matrix of observations on the covariates, and the spatial weight matrix W, which is constant over time, is the N × N positive matrix describing the spatial arrangement of the n units whose diagonal elements are set to be zero. Each entry wij∈W represents the spatial weight matrix associated with units i and j [38–42]. The elements of wij is (i, j), which is the neighborhood matrix of the row standardized matrix with a dimension of 72 × 72. Hence, the non-zero elements of the matrix indicate whether the two locations are neighbours. This weighted matrix is commonly expressed as: The existence of spatial autocorrelation in the dataset is checked by using Moran’s I. The Moran’s I is used to associate weight (wij) to each of the pairs [261–265], which quantifies the spatial pattern. The test is given as follows, where n is the number of investigated points, xi,xj the observed value of two points of interest, μ the expected value of x, and wij the elements of the spatial weight matrix. In Moran’s I ranges [-1, 1] the value of 1 signifies that clusters with high values of the variable of interest are close to clusters with similar high values, while -1 indicates that high values are near to low values. In this paper, the four basic spatial time dynamics models (spatial Durbin model, spatial autoregressive model, spatial error model, and general nested model with space–time), were adopted [14, 42, 43]. Let the WX be the interaction effects among the covariates with the spatial components, and the Wu the interaction effects among the error terms of different observations, [Wyt]i is the ith element of the spatial lag vector in the same period. The [Wyt-1]i is the ith element of the spatial lag vector of observations on the response variable in the previous time. When the response variable is related to the same locations as well as the neighboring locations in another period, the model is called a space–time recursive model. The yit-1 is the observations on the dependent variable in the previous period. Moreover, let ρ be the spatial dependence parameter, θ the spatio-temporal diffusion parameter, and ϕ the autoregressive time dependence parameter [4–11, 36–40] (Fig. 2). The space–time dynamic models. GNS: General Nesting Spatial model; SDM: Spatial Durbin Model; SAR: Spatial Autoregressive model; SEM: Spatial Error Model. When the response variable is related to the same locations as well as the neighboring locations in another period, the model is called the space–time recursive model. The yit-1 is the observations on the dependent variable in the previous period. The standard assumptions that εij∼N(0,σ2) and Eεitεjs=0 for i≠j or t≠s apply in any case [12, 14, 36, 42, 43].
N/A
