Background and aims: Childhood acute malnutrition, in the form of wasting defined by a severe weight loss as a result of acute food shortage and/or illness. It is a critical public health problem that needs urgent attention in developing countries, like Ethiopia. Despite its variation between localities, the risk factors and its geospatial variation were not addressed enough across the various corner of the country. Therefore, the current study was undertaken to assess spatial variation and factors associated with acute malnutrition among under-five children in Ethiopia. Methods: A total weighted sample of 4 955 under-five children were included from the 2019 Demographic and Health Survey. Getis-Ord spatial statistical tool used to identify the hot and cold spot areas of severe and acute malnutrition. A multilevel multivariable logistic regression model using was used to examine predictors of acute malnutrition. In the multivariable multilevel analysis, Adjusted Odds Ratio with 95% CI was used to declare significant determinants of acute malnutrition among children. Result: Among 4 955 under-five children, 7% of them were wasted and 1% of them were severely wasted in Ethiopia during the 2019 national demographic survey. The distribution was followed some spatial geo-locations where most parts of Somali were severely affected (RR = 1.46, P37 value <0.001), and the distribution affected few areas of Afar, Gambella, and Benishangul Gumz regions. Factors that significantly associated with childhood wasting were: gender(male)1.9 (1.3–2.7), age (above 36 months) 0.5 (0.2–0.9), wealth index(richest) 0.5 (0.2–0.8), and water source (unimproved source) 1.5 (1.0–2.3). Conclusions: Our finding implies, the distribution of childhood wasting was not random. Regions like Afar, Somali, and pocket areas in Gambella and SNNP should be considered as priority areas nutritional interventions for reducing acute malnutrition. The established socio-demographic and economic characteristics can be also used to develop strategies.
We have used the Ethiopian Demographic Health Information Survey (EDHS) of 2019 to identify factors associated with wasting which were community-based cross-sectional surveys conducted across the country. Ethiopia, the most populous country in Africa, is situated in the Horn of Africa between 3 and 15 degrees north latitude and 33 and 48 degrees east longitude (3°–15° N and 33°–48°E). It has an administrative structure of nine regional states (Tigray, Afar, Amhara, Oromiya, Somali, Benishangul-Gumuz, Southern Nations Nationalities and People [SNNP], Gambela, and Harari) and two city administrations (Addis Ababa and Dire Dawa). These are subdivided into 68 zones, 817 administrative districts which are further divided into 16 253 Kebeles, the smallest administrative units of the country. It has an estimated population of 114.96 million in 2020, which makes it second in Africa and 12th in the world’s most populous country. Our data source was the EDHS survey which was collected in 2019. EDHS is collected every five years by the Ethiopian Central Statistical Agency (CSA) along with ICF International and funded by USAID. The data sets for EDHSs were downloaded in SPSS format with permission from the Measure DHS website (http://www.dhsprogram.com). The shapefile of the map of Ethiopia has been accessed as an open-source without restriction from Open Africa website (https://africaopendata.org/dataset/ethiopia-shapefiles). The EDHSs samples were collected using stratified in a two-stage cluster sampling technique. In the first stage, each region was stratified into urban and rural areas. In the second stage of selection, a fixed number of households per cluster were selected with an equal probability of systematic selection from the newly created household listing. All women age 15–49, who were either permanent residents of the selected households or visitors who slept in the household the night before the survey, were eligible to be interviewed. All under-five children within five years during the surveys in Ethiopia were the source of the population for this study, whereas all under-five children in the selected enumeration areas (EAs) within five years during the survey were the study population. Ultimately, a total representative sample of 5 057 under-five children was included in the 2019 survey [18,20,21]. Geographic coordinates of each survey cluster were also collected using Global Positioning System (GPS) receivers. To ensure confidentiality, GPS latitude/longitude positions for all surveys were randomly displaced before public release. The detailed procedure has been presented in each EDHSs report [18,20,21]. In this study, the dependent variable was under-five wasting which is defined as the percentage of under-five children whose weight-for-height z-score (WHZ) is below –2 SD in the national center for health statistics (NCHS) growth curve. Therefore, we consider under-five wasting (wasted = 1 or not wasted = 0) as the outcome variable [14]. When Y is the outcome variable (wasting), while i is for the individual-level factors, while j is for the community factors. The independent variables included: socio-demographic and economic factors: age, sex, occupation, educational status, head of household, wealth index, and religion, geographical factors (region, residence, and temperature), maternal health service utilization factors (antenatal care, place of delivery, and postnatal care), nutritional status of mother (BMI and HFA), birth weight, the timing of breastfeeding, clinical factors (anemic status of the mother, anemic status of the child), drinking safe water, latrine use and media exposure of respondents. Early initiation of breastfeeding –infants who are sucking the breast milk within one hour of birth. Introduction of solid, semi-solid, or soft foods (6–8 months), birth interval were factors for childhood wasting [3,4,7,24,25,26,30,32,36]. After downloading EDHS data, sample weights were applied to compensate for the unequal probability of selection between each stratum, data cleaning and recording were carried out in SPSS statistical software version 24. The EDHS datasets were joined to Global Positioning System (GPS) coordinates of EDHS using the joining variable as recommended by DHS measure. The data was exported into Arc GIS 10.8 to visualize key estimation, clusters, and regional variation among wasting. For the spatial analysis, ArcGIS version 10.8 and Sat Scan version 9.6 statistical software were used for exploring the spatial distribution, global spatial autocorrelation, spatial interpolation, and for identifying significance. The spatial autocorrelation (Global Moran’s I) statistic measure was used to evaluate whether the spatial distribution of wasting was random or not. Moran’s I is a spatial statistic used to measure spatial autocorrelation by taking the entire data set and produce a single value that ranges from –1 to + 1. Moran’s I values close to –1, 1, and 0 indicate wasting was dispersed, wasting was clustered, and wasting was distributed randomly, respectively. A statistically significant Moran’s I (P < 0:05) leads to rejection of the null hypothesis (wasting is randomly distributed) and indicates the presence of spatial autocorrelation. The local Getis-Ord G index (LGi) was used to analyze causality autocorrelation into positive and negative. If the prevalence rates had similar attributes of high or low values (high-high or low-low autocorrelation), they were defined as positive autocorrelation whereas if the attributes had opposing values (high-low or low-high autocorrelation) they were defined as negative autocorrelation. Moreover, the spatial interpolation technique was applied to predict the un-sampled/unmeasured value from sampled measurements. Autocorrelation can be classified into positive and negative correlations through the local Getis-Ord G positive autocorrelation occurs when similar values are clustered together on a map (high rates surrounded by nearby high rates or low rates surrounded by nearby low rates). Negative autocorrelation indicates different values clustered together on a map, that is, high values surrounded by nearby low values or low values surrounded by nearby high values. Statistical significance of autocorrelation was determined by z-scores and p-value with a 95% level of confidence. The distribution and variations of wasting prevalence rates among children across the country were displayed on the map. Using Kuldorff’s SaTScan version 9.6 program, spatial scan statistical analysis was used to classify statistically important hotspot areas. To fit the Bernoulli model, we used wasting under-five children as cases and not wasted children as controls. The numbers of cases in each location have Bernoulli distribution and a maximum spatial cluster size of < 50% of the population was used as an upper limit. Z-score is computed to determine the statistical significance of clustering, and the P-value was used to determine if the number of observed 6 to 59 months aged children who were within the potential cluster was significant or not. The null hypothesis of no clusters was rejected when the P-value ≤ 0.05. Based on 999 Monte Carlo replications the significant clusters were identified and ranked based on their likelihood ratio test [37,38]. The multivariable multilevel logistic regression model was used to determine the effect of different factors on wasting. For this multilevel analysis, four models were constructed. Those are the null model without predictors (Model I), model II with only individual-level variables, model III with only community-level variables, and model IV both individual-level and community-level variables. For model comparison, we used the log-likelihood ratio (LLR) and deviance. The highest log-likelihood or the smallest deviance wins the best-fitted model. Therefore, model III which includes both individual and community-level variables was selected as the best fit model for the data. An adjusted OR (AOR) with 95% CIs was computed to identify the independent factors of under-five wasting at p value<0.05. A multicollinearity test was done in order to rule out a significant correlation between variables. If the values of variance inflation factor (VIF) were lower than 10, then the collinearity problem was considered less likely. Correlation coefficient (ICC), a proportional change in community variance (PCV), and median odds ratio (MOR) were used for measuring variation or random effect [35]. The intra-class correlation coefficient is a measure of within-cluster variation (i.e. the variation between individuals within the same cluster). The PCV is a measurement of the total variation attributed to individual and/or community-level factors at each model. The MOR is the median odds ratio between the individual of higher propensity and the individual of lower propensity when comparing two individuals from two different randomly chosen clusters and it measures the unexplained cluster heterogeneity (the variation between clusters) by comparing two persons from two randomly chosen different clusters. The MOR measure is always greater than or equal to “1.” If The MOR measure is “1,” there is no variation between clusters. The within-cluster correlation was measured using intra-cluster correlation (ICC) which is expected to be 10% to use the model. The ICC, PCV, and MOR were determined using the estimated variance of clusters using the following formula: ICC=VV+π2/3, Where V is a variance of estimated clusters, MOR=Exp2×V×0.6745 PCV=VA−VBVB×100, were VA = variance of the initial model; VB = variance of the model with more terms. The multilevel analysis model is one of the analysis methods that can correctly handle the correlated data. A multilevel model evaluates how factors at different levels affect the dependent variable. A multilevel model provides correct parameter estimates by correcting the biases introduced from clustering by producing correct SEs, thus producing correct CI and significance tests. Publicly available EDHSs data were used for this study. Ethical approval of EDHS was obtained from the ICF Institutional Review Board (IRB), Ethiopia Health and Nutrition Research Institute Review Board, and the Ministry of Science and Technology. For this particular study, a brief description of the protocol was submitted to the MEASURE DHS program to access and analyze the data. Permission was obtained from the program to access and analyze the data. During EDHS data collection, Informed consent was taken from each participant, and all identifiers were removed and the confidentiality of the information was maintained.
N/A
DIMA AI Care
