Menu
Menu
Menu
Menu
Background: Undernutrition is a significant public health challenge and one of the leading causes of child mortality in a wide range of developing countries, including Ethiopia. Poor access to water, sanitation, and hygiene (WASH) facilities commonly contributes to child growth failure. There is a paucity of information on the interrelationship between WASH and child undernutrition (stunting and wasting). This study aimed to assess the association between WASH and undernutrition among under-five-year-old children in Ethiopia. Methods: A secondary data analysis was undertaken based on the Ethiopian Demographic and Health Surveys (EDHS) conducted from 2000 to 2016. A total of 33,763 recent live births extracted from the EDHS reports were included in the current analysis. Multilevel logistic regression models were used to investigate the association between WASH and child undernutrition. Relevant factors from EDHS data were identified after extensive literature review. Results: The overall prevalences of stunting and wasting were 47.29% [95% CI: (46.75, 47.82%)] and 10.98% [95% CI: (10.65, 11.32%)], respectively. Children from households having unimproved toilet facilities [AOR: 1.20, 95% CI: (1.05,1.39)], practicing open defecation [AOR: 1.29, 95% CI: (1.11,1.51)], and living in households with dirt floors [AOR: 1.32, 95% CI: (1.12,1.57)] were associated with higher odds of being stunted. Children from households having unimproved drinking water sources were significantly less likely to be wasted [AOR: 0.85, 95% CI: (0.76,0.95)] and stunted [AOR: 0.91, 95% CI: (0.83, 0.99)]. We found no statistical differences between improved sanitation, safe disposal of a child’s stool, or improved household flooring and child wasting. Conclusion: The present study confirms that the quality of access to sanitation and housing conditions affects child linear growth indicators. Besides, household sources of drinking water did not predict the occurrence of either wasting or stunting. Further longitudinal and interventional studies are needed to determine whether individual and joint access to WASH facilities was strongly associated with child stunting and wasting.
Ethiopia is Africa’s second-most populated country, after Nigeria, with a population of over a hundred million people. Ethiopia, with a federal system of government has 10 regions (i.e., Afar, Amhara, Benishangul-Gumuz, Gambella, Harari, Oromia, Somali, Sidama, Southern Nations and Nationalities and People (SNNP), and Tigray) and two chartered cities (i.e., Addis Ababa and Dire Dawa). Ethiopia shares borders with Eritrea in the north, Kenya and Somalia in the south, South Sudan and North Sudan in the west, and Djibouti and Somalia in the East [29]. The datasets from the four rounds of the Ethiopian Demography and Health Surveys (EDHS) conducted from 2000 to 2016 were used in this study [29–32]. The EDHS is a nationally representative survey collected every five years, providing population and health indicators at the regional and national levels. The EDHS used a multistage cluster sampling technique, whereby data are hierarchical (i.e., children and mothers were nested within households, and households were nested within clusters). For this reason, we employed a multilevel logistic regression model, which has many advantages over the classical logistic regression model and is appropriate for analysing factors from different levels. A detailed description of analysis is presented in the data analysis section. The datasets of each survey were obtained from the following EDHS data repository https://dhsprogram.com. In brief, the 2000 and 2005 data were collected based on the 1994 population and housing census frame, while the 2011 and 2016 data were collected based on the 2007 population and housing census frame [29–32]. EDHS data were collected using a stratified two stage cluster sampling technique. In the first stage, a total of 539 enumeration areas (EAs) or clusters (138 in urban areas and 401 in rural areas), 540 EAs (145 urban and 395 rural), 624 clusters (187 in urban areas and 437 in rural areas), and 645 clusters (202 in urban areas and 443 in rural areas) were selected using systematic sampling with probability proportional to size, respectively the 2000, 2005, 2011 and 2016 EDHS surveys. At the second sampling stage, a systematic sample of households per EA was selected in all the regions to provide statistically reliable estimates of key demographic and health variables. The EDHS used a questionnaire that was adapted from model survey tools developed for the DHS Program project. Mothers or caregivers provided all information related to children and mothers or caregivers through face-to-face interviews which were held at their homes. Water, Sanitation and Hygiene (WASH) indicators were also collected through face-to-face interviews and observation methods. The EDHS collected data on children’s nutritional status by measuring the weight and height of under-fives in all sampled households. Weight was measured with an electronic mother-infant scale (SECA 878 flat) designed for mobile use. Height was measured with a measuring board (Shorr Board). Children younger than 24 months were measured lying down on the board (recumbent length), while standing height was measured for older children, in conformity with previous studies[29–32]. The prevalence of stunting and wasting, defined by the World Health Organization (WHO), were the primary outcome variables of interest [33]. Height-for-age is a measure of linear growth retardation and cumulative growth deficits. Children, whose height-for-age Z-scores were below minus two standard deviations (-2 SD) from the median of the reference population, were considered short for their age (stunted) or chronically undernourished [33, 34].The weight-for-height index measures body mass in relation to body height or length and describes current nutritional status. Children, whose Z-scores below minus two standard deviations (-2 SD) from the median of the reference population, were considered thin (wasted) or acutely undernourished [33]. The key exposure variables examined were all variables related to WASH, and specifically, sanitation facility (improved/unimproved), sources of drinking water (improved/unimproved), time to obtain drinking water (round trip) were classified as ‘water on premise’, ‘≤ 30 minutes round-trip fetching times’, ‘31–60 minutes round-trip fetching times’, ‘and > 60 minutes round-trip fetching times’, child stool disposal (safe/unsafe), and housing floor (improved/unimproved). A household floor was considered as improved only if households were without dirt floors. The World Health Organization (WHO)/ United Nations Children’s Fund (UNICEF)- Joint Monitoring Programme (JMP) for water improved supply and sanitation definition was taken into consideration in this study [35]. Unsafe disposal of children’s stool was defined as the disposal of faeces in any site other than a latrine, whereas other methods such as “child used latrine or latrine” and “put/rinsed into latrine or latrine” were considered as “safe disposal” [36] (Table 1). Exposure variable description and survey question What kind of toilet facility do members of your household usually use? (verify by observation) The last time (NAME OF YOUNGEST CHILD living with the respondent) passed stool, what was done to dispose of the stool? Observe the main material of the floor of the dwelling. Record observation How long does it take to go there, get water, and come back? As undernutrition results from a combination of factors, several control variables were considered in this study. We classified the control variables as child-related, parental-related, household-related, and community-related. As a result, the following factors were considered in the analysis. Child-related variables include: diarrhea, fever, symptoms of acute respiratory infection (ARI), sex, age (months), birth order, birth interval, size of child at birth (mother’s perceived baby size at birth), currently breastfeeding, early initiation of breastfeeding (children born in the past 2 years who started breastfeeding within one hour of birth), received all basic vaccination (i.e., child received a Bacillus Calmette–Guérin [BCG] vaccination against tuberculosis, 3 doses of Diphtheria, pertussis, and tetanus vaccine [DPT], ≥ 3 doses of polio vaccine [OPV], and 1 dose of measles vaccine). Parental-related factors included: mother’s age, mother’s educational level (no education, primary, secondary, and higher), mother’s occupation (not working, non-agriculture, or agriculture), antenatal care visits (ANC) (none, 1–3, or 4+), maternal body mass index (BMI), husband’s educational level, husband’s occupation (not working, non-agriculture, or agriculture), listening to the radio, and watching television. Household-level factors include: wealth index categorized (poor, middle, or rich) and household size (1–4 or ≥ 5). The wealth index is categorised into five wealth quintiles: ‘very poor’, ‘poor’, ‘middle’, ‘rich’ and ‘very rich. For this analysis, we re-coded the wealth index into three categories for adequate sampling in each category: ‘poor’ (poor and very poor), ‘middle’ and ‘rich’ (rich and very rich). Community-level factors include: ecological zone (tropical zone, subtropical zone, and cool zone), place of residence (urban and rural), and region (agrarian, pastoralist, and city-dweller). All statistical analyses were conducted using Stata™ software version 15.1 (Stata Corp, College Station, TX, USA). Descriptive statistics were used to describe the socio-demographic and economic characteristics of children included in the study. Differences in the two outcome variables “stunting” and “wasting” were presented across socio-demographic characteristics of interest using frequencies and percentages. A multilevel logistics regression analysis was performed using a stage modelling approach for each outcome (i.e., stunting and wasting). This means that each of the five-level factors (i.e., WASH, child-related factors, parental-related factors, household-related factors, and community-level factors) were examined using a series of multilevel logistic regression models, adjusting for selected potential confounders. A multilevel logistic regression model was used because of the nested structure of the EDHS data (i.e., individuals nested within households and households nested within clusters). Sampling weight was used during data analysis to adjust for non-proportional allocation of sample and possible differences in response rates across regions included in the survey. A detailed explanation of the weighting procedure has described in the EDHS methodology report [29–32]. Hierarchical multilevel models were run following the recommendations of a previous study that suggest complex hierarchical relationships of different determinants at different levels [37]. This approach allowed distal factors to be adequately investigated without interference from proximal factors [38]. A similar approach was also used to identify previous related literature [39]. In brief, a multilevel bivariable logistic regression model (Model 0- maximum model) was fitted with each explanatory variable to select candidates with p-value a < 0.20 for the stage multivariable models. Accordingly, Model 1 incorporated WASH variables only. Model 2 incorporated WASH plus child-related variables (all child-related explanatory variables with p-values < 0.2 from Model 0 were entered into the Model1). Model 3 incorporated WASH + child-related variables + parental-related factors (all parental-related variable with p-values < 0.2 from Model 0 were entered into Model 3). Model 4 incorporated WASH + child-related factors + parental-related factors + household-related factors (all household-related variables with p-values < 0.2 from Model 0 were entered into the model 4). Model 5 incorporated WASH + child-related variables + parental factors + household factors + community-level factors. Model 6 was the final model that included only variables with a p-value < 0.2 from Model 5. Both crude odds ratio (COR) and adjusted odds ratio (AOR) ,along with 95% confidence intervals (CI), were used to estimate the strength of the association between explanatory and response variables.
