Factors affecting under-ve diarrhea in households with dirt oors: a multilevel mixed-effects analysis from a national survey of Ethiopia

Introduction: Diarrhea remains the second leading cause of mortality among under-ve children in Ethiopia, spreads more easily in homes with dirt oors. Even though the determinants of diarrheal illness is widely recognized at the global level, there is a limited number of studies that identied the individual and community-level factors solely among children residing in households with sub-optimal ooring (i.e. earth, mud or sand, and dung). Therefore, this study aims to identify individual and community-level factors associated with childhood diarrhea in households with sub-optimal ooring in Ethiopia. Methods: Data from the fourth round of the Ethiopian Demographic and Health Survey (EDHS) conducted in 2016 is used to carry out the analysis. The EDHS was large, cross-sectional by design, and nationally representative. A total of 9,097 under-ve children living in households with sub-optimal oors and 645 community clusters were included in the current analysis. A multi-level logistic regression analysis was used to examine both individual and community level factors while accounting for the hierarchal structure of the data. Results: The overall prevalence of childhood diarrhea in households with suboptimal ooring in Ethiopia was 11.9% (95% CI:11.2-12.6). Children aged 6-11 months (adjusted OR [AOR]=2.68, 95%CI: 1.89-3.82), and ≥ 12 months (AOR=1.65, 95%CI: 1.22-2.24) were more likely to have diarrhea compared with children aged 0-5 months. Having cough (AOR=6.09, 95% CI=4.93-7.52) and fever (AOR=1.62, 95%CI: 1.30-2.01) were found to increase the risk for the development of diarrhea in children. Maternal age (25-34 years) and paternal lack of educated were found to be associated with lower odds of childhood diarrheal prevalence rates. Further, children from households with higher family size ( ≥ 5 family size), low birth order (rstborn), and the number of under-ve children (2-3 children) were reported decreased odds of diarrhea. At the community level, children living in pastoralist regions (such as Afar and Somali region) signicantly less likely to have diarrhea vulnerability (AOR: 0.61, 95%CI: 0.40-0.91) compared to those who reside in cities.


Introduction
Diarrhea causes more than half a million childhood mortality across low-income and middle-income countries (LMICs) [1]. According to the Global Burden of Disease Study 2017, more than 910 million childhood diarrheal cases were reported each year [2]. Globally, diarrhea remains a leading cause of under-ve mortality-account for 1 in 9 child deaths [3]. In addition to this staggering loss of under-ve life, it can have a detrimental impact on childhood growth and cognitive development [4,5]. It is also estimated that the odds of stunting at 24 months increased by 5% with each diarrheal episode [6].
Furthermore, the proportion of reported antibiotic use among sick children under 5 years of age with fever, diarrhea, or cough was less than fty percent, 43.1% [7].
Diarrhea is preventable with the application of hand hygiene, basic sanitation, and the provision of safe drinking water [8]. Almost 88% of diarrhea-associated deaths are attributable to unsafe water, inadequate sanitation, and insu cient hygiene [9]. Several studies explore the effect of single and combined water, sanitation, and hygiene (WASH) intervention in the reduce risk of diarrheal morbidity [8, [10][11][12][13][14]. For example, a systematic review for the effect of WASH interventions for acute childhood diarrhea found that various WASH interventions show diarrhea risk reductions between 27% and 53% in children under ve years old, depending on intervention type [11]. A recent updated meta-analysis showed that promoting handwashing with soap associated with reduced risk of diarrhea by 30% [14].
In Ethiopia, where there is a lack of safe excreta disposal facility, safe drinking water, and poor hygiene practice [8, 15,16], diarrhea remains is an important public health problem. Over the past two decades, the prevalence of childhood diarrhea has shown a declining trend-from 24% in 2000 to 12% in 2016 [17,18].
Despite progress in the reduction of childhood diarrhea in Ethiopia, still it is the second leading cause of morbidity and mortality among under-ve children next to pneumonia [15,18]. A recent systematic review nding revealed that the pooled prevalence of diarrhea among under-ve children in Ethiopia was 22% [19].
Different studies showed that household-level risk factors such as lack of safe water and improved sanitation facility, poor maternal hygiene, household water treatment, methods of complementary feeding, types of water storage equipment, maternal education, and improper waste disposal practices were signi cant factors for the occurrence of diarrheal illness [8, 10,12,13,15,19,20]. In previous works of literature, the effect of improved WASH on childhood diarrhea morbidity was well established. However, in some instances, WASH interventions fail to appear long-term impact [21], due to reinfection and contamination of the living home as a result of dirt oors. As dirt oors facilitate and increased de lement of nger, uid, food, and materials that encounter these surfaces and possibly increase the risk of diarrhea among children [21][22][23][24]. A study from Zimbabwe showed that mothers of infants living in households with improved ooring were less likely to report diarrheal illness. Further, the association between ooring and diarrheal illness did not vary by the presence of improved/unimproved water or sanitation [23]. It was also found that dirt and mud oors are a known predictor of diarrhea and parasitic infestations [23,24]. Eliminating a dirt oor from the home results in dramatic reductions in childhood diarrhea and Soil-Transmitted Helminth (STH) [23,25]. For instance, replacing a dirt oor with a concrete oor reduces diarrhea by 49% [26].
In many developing countries, including Ethiopia replacing a dirt oor with concrete or other improved material is unaffordable to the poor. And the challenges remain due to cleaning of sand or soil oors are so di cult, the proximity of latrine pits, and unsanitary environmental surrounding. This may be further exacerbated by contamination of the oor by fecal matter brought in on shoes, and especially when animals live in close proximity to humans [22,27,28]. According to the recent 2016 EDHS report, the two most commonly used materials for ooring in Ethiopia are earth or sand (48%) and dung (33%) [18]. And children dwelling in households with mud oors are disproportionately affected by diarrheal diseases [29].
Numerous research articles from Ethiopia have identi ed the determinants of under-ve diarrhea [15,19,29,30]. However, speci c studies that focus on factors that in uence childhood diarrhea in households with sub-optimal ooring (i.e. earth, mud or sand, and dung) are limited and not su cient to show the underlying factors. As more than 80% of Ethiopians live and sleep on a dirt oor, the challenges remain enormous where diarrhea is spreads more easily in homes with dirt oors. Up to date, no study ever assessed the determinants of diarrhea solely among children residing in households with the suboptimal oor in Ethiopia. Therefore, this present study aimed to investigate factors associated with childhood diarrhea among children residing in households with sub-optimal ooring in Ethiopia.

Data Sources
The data source for this analysis was the 2016 Ethiopia Demographic and Health Survey (EDHS) [18]. It is a nationally representative household survey carried out based on a nationally representative sample of households that provide estimates at the national and regional levels.

Study design and sampling
The 2016 EDHS was cross-sectional by design. The sample in EDHS was designed to provide population and health indicators at the national and regional levels. The EDHS used a strati ed two-stage cluster sampling technique. In the rst stage, 645 enumeration areas (EAs) (202 urban and 443 rural areas) were selected with probability proportional to the EA size. Then, household listing was done for the selected EAs. In the second stage, a xed number of 28 households per cluster were selected using the newly created household list as a sampling frame [18]. In this study, we included all under-ve children living in households with suboptimal oor. A total of 9,097chidren from sub-optimal oor households and 645 community clusters were included in the current analysis ( Figure 1)..

Outcome variable
The outcome variable for this study was presence of diarrhea. Diarrhea was de ned as the passage of 3 or more loose or liquid stools per day, or more frequently than is normal for the individual. In the EDHS, mothers are asked if their children under ve had diarrhea in the past 2 weeks prior to the survey. The response was recorded as "yes" and "no".

Independent variables
The individual-and community-level variables included in the study are shown in Table 1, along with the coding and de nitions.

Data analysis
Data analysis was carried out using STATA version 14 (Stata Corp, College Station, Texas, United States) statistical software. We used the "svy" command to weight the survey data as per the recommendation of the EDHS. Sample weights were applied in order to compensate for the unequal probability of selection between the strata that were geographically de ned, as well as for non-responses. A detailed explanation of the weighting procedure can be found in the EDHS methodology report [10]. The EDHS data are hierarchical (children were nested in clusters). Children from the same cluster will be more similar to each other than children from different clusters. For this reason, we used a multilevel model which account the hierarchical nature of the EDHS data [31]. Accordingly, four models were tted to estimate both xed effects of the individual and community-level factors and random effect of between-cluster variation. Null Model ts were assessed using log-likelihood (LL), deviance, and Akaike Information Criterion (AIC). Loglikelihood (LL), AIC, and deviance were used to estimate the goodness of t of the adjusted nal model in comparison to the preceding models (individual and community level model adjustments). The LL, AIC, and deviance value for each subsequent model was compared and the model with the highest value of LL and lowest value of deviance and AIC was considered to be the best t model. Variables that had a relationship with childhood diarrhea (p<0.20) were considered for the nal model [32]. Adjusted ORs (AOR) with a 95 %CI (Con dence Interval) were used to declare statistical signi cance.
Ethics approval DHS Programme granted permission to download and use the data for this study after being registered and submitting a request with brie y stated objectives of the study. The Institution Review Board approved procedures for DHS public-use data sets that do not in any way allow respondents, households, or sample communities to be identi ed. There are no names of individuals or household addresses in the data les. The detail of the ethical issues has been published in the 2016 EDHS nal report, which can be accessed at: http://www.dhsprogram.com/publications.

Results
Characteristics of study participants and childhood diarrhea prevalence Table 2 shows the descriptive data of the respondents. Of the 9,097 children included in the analysis, half (n = 4,646, 51.1%) were male. The majority 78.4% of the children were aged 12 months or older and 70.7% currently breastfed. Over 70.4% of mothers of the children had no formal education, almost three-fourth of them were agricultural employees, and 52.9% were in the poor wealth quintile. The majority of study participants, (n = 8,681, 95.4%) were rural dwellers, 52.3% of the respondents use unimproved drinking water sources, and 94.5% use unimproved sanitation facilities. The overall prevalence of childhood diarrhea in households with suboptimal ooring in Ethiopia was 11.9% (95% CI:11.2-12.6).

Factors associated with childhood diarrhea in households suboptimal ooring in Ethiopia
On bivariable multilevel logistic regression analysis, child's age, number of under-ve children, currently breastfeeding, cough in the last two-week, child stool disposal, mother's age, mother's education level, respondent currently working, paternal education, household size, wealth quintiles, region, and ecological cluster were associated with childhood diarrhea. However, in the nal model at the individual level (child's sex, child's age, number of under-ve children, birth order, fever in the last two-week, cough in the last two- week, child stool disposal, mother's age, paternal education, and household size) and at community level region were signi cantly associated with diarrhea among children's in households suboptimal ooring in Ethiopia (p< 0.05) Table 3.

Measures of variation (random-effects) and model t statistics
Measure of variation (random intercept models) and model t statistics of diarrhea in households with suboptimal ooring in Ethiopia can see from Table 6. The ICC in the empty model was 10.36 %, indicating that 10.36% of the total variability for diarrhea was due to differences between clusters (Enumeration's areas), with the remaining unexplained 89.64% which is attributed to individual differences. Additionally, the ICC in the nal model (6.0%) suggested that residual community in uences were persistent even after adjusting for the individual-and community-level factors. This implies that there are other unmeasured community factors. In this study, the models were compared with deviance, and model III (a model with both individual and community level factor) was selected, had the lowest deviance (4,375.15) ( Table 6)..

Discussion
This study identi es factors associated with the prevalence of childhood diarrhea in households with sub-optimal oors in Ethiopia. Our study may be the rst to report factors associated with the prevalence of childhood diarrhea in households with sub-optimal ooring in the Ethiopian context. It was found that childhood diarrhea was associated with the age of the child, birth order, presence of cough and fever, maternal age, paternal education, household size, improper child stool disposal, number of under-ve children, and place of residence.
In this study, the odds of diarrhea among children older than 6 months were higher compared with those aged < 6 months, which was similar to the results of a study done in Rwanda [33], Pakistan [34], and Ethiopia [35]. The possible justi cations could be due to the fact that children older than 6 months usually crawling on the ground which increases the probability of getting and contracting lth materials, particularly those live-in households with the mud oor may expose to pathogenic microorganisms easily.
In addition, in this age, unhygienic and contaminated food as a result of sub-optimal ooring may increase the risk of diarrhea. On the other hand, younger infants-age less than 6 months would be protected against diarrheal diseases by different mechanisms such as maternal antibodies obtained through exclusive breastfeeding and they are less exposed to the contaminated oor because these ages are neither crawling nor walking and cannot easily pick dirt or other contaminated objects.
A lower risk for diarrhea was found for children in households with 2-3 under-ve children. This nding is in agreement with a cross-sectional study conducted in Ethiopia [8]. This might be due to when the number of children in the household increases it is expected that older children may guide younger sibling on hygiene behavior and attention from peer may increase. In contrast, other studies reported an increased risk of diarrheal frequency as the number of children increased [30,36].
In this study, lower birth order was associated with a lower risk of diarrheal morbidity. This nding supported cross-sectional studies conducted in Ethiopia [8,30]. The possible explanation for this may be as the childbirth order increases it is likely that children will be more exposed to diarrhea because the care and attention from parents decrease [30].
Children having cough and fever were found to have increased probability of childhood diarrhea by 6.09 and 1.62-fold, respectively, compared to children without cough and fever in the past 14 days prior to the survey. Having both conditions of cough and fever may increase the risk of diarrhea or the episodes of diarrhea among children may predispose children for cough and/or fever. Due to the nature of the study design, we are unable to determine which occurred rst. However, there the relationship could explain the association between pneumonia and diarrhea [37,38]. A study from Ethiopia also identi ed the positive association between diarrhea and cough, the study indicated the risk of diarrhea was four times higher in children having cough than their counterparts [8].
In this study, a signi cant association between diarrhea and child stool disposal was found. Somewhat surprisingly, our nding showed that the odds of diarrhea were lower among children whose stool was disposed of unsafe than children whose stools were disposed of safely. We were unable to locate other studies that have examined the relationship between presence of diarrhea and child stool disposal in households with sub-optimal ooring. However, studies reported a positive relationship between unsafe child stool disposal and increased risk of childhood diarrhea [30,39].
Living in the pastoralist area was also found to be protective for developing diarrhea, as compared to those dwelling in cities. This nding can satisfactorily explain by children of pastoralist communities tend to move from place to place, as a result, the risk of fecal-oral disease transmission from contaminated ooring may be minimal as compared to those children who live on earth oor in cities. In addition, among city dwellers, children spend much time on the contaminated ground and their exploratory behaviors including geophagia could expose them for diarrheal disease.
In this study, water and sanitation facility failed to be signi cantly associated with childhood diarrhea. This nding contradicts previous study which reported that unprotected source of drinking water and unimproved latrine were associated with the higher incidence of childhood diarrhea [40]. The possible reason for this discrepancy maybe due to water and sanitation interventions fail to show signi cant longterm impact in households with sub-optimal ooring, due to reinfection of children with diarrhea as dirt ooring can facilitate contamination. In many case, poor hygiene practices persistent in households with sub-optimal ooring and children were exposed to feco-oral infections regardless of the availability of water and sanitation facilities. Further, this nding has implication on WASH interventions that aimed to reduce childhood diarrhea; in households with sub-optimal ooring WASH intervention alone may not have an impact on childhood diarrhea.

Limitations
Though the study explored factors associated with childhood diarrhea in households with sub-optimal ooring it has some limitations. Firstly, the analyses were conducted using EDHS data collected in a cross-sectional survey, which prevents causal inferences. Secondly, because the information on childhood diarrhea was self-reported, there is the possibility of recall bias. Third, due to the secondary nature of the data, the present study was limited by unmeasured confounders. Despite these limitations, we used a multilevel model to account for the clustered nature of EDHS data, which enhances the accuracy of estimates.

Conclusion
This study provides the rst empirical evidence on factors associated with the prevalence of diarrhea in children residing in households with sub-optimal ooring in Ethiopia. Both individual-level (age of the child, birth order, presence of cough and fever, maternal age, paternal education, household size, improper child stool disposal, number of under-ve children) and community-level factor (place of residence) were revealed to be important factors for childhood diarrhea in households with suboptimal ooring. Our results suggest that household improved ooring interventions may yield more sustained reductions in the prevalence of childhood diarrhea. In addition, public health interventions and strategies designed to promote sanitation in households with sub-optimal ooring need to consider the identi ed factors through community assessments to identify community-speci c barriers. The data we used which is the 2016 Ethiopian Demographic and Health Survey was obtained from the DHS program (www.dhsprogram.com) but the 'Dataset Terms of Use' do not permit us to distribute this data as per data access instructions (http://dhsprogram.com/data/Access-Instructions.cfm). To get access to the dataset you must rst be a registered user of the website (www.dhsprogram.com) and download the 2016 Ethiopian Demographic and Health Survey.

Competing interests
The author declares that he has no competing interests.

Funding
No fund was received for the present review.
Authors' Contribution BS: Conceptualizes, design the study and data curation, performed the analysis, wrote and approved the nal manuscript. AK, DA, YT, DW, DZ and TA: Contribute to the analysis, critically reviewed the manuscript and approved the nal manuscript. All authors read and approved the nal manuscript before submission. Categorized into (1)  The variable was recoded as 'agrarian' (encompassing Tigray, Amhara, Oromia, Benishangul, SNNPR, and Gambela), 'pastoralist' (Afar and Somali regions) or 'city dweller' (Addis Ababa, Dire Dawa cites and Harari). An agrarian society is any community whose economy is based on producing and maintaining crops and farmland. A pastoralist society is any community whose economy is based on raising livestock. A city-dweller society is any city community

Ecological Cluster
Categorized into (1) <1500; (2) 1500-2500; and (3) >2500 a child who received one dose of measles vaccine at any time before the survey (according to a vaccination card, health facility, or the mother's report); b frequency of watching television was categorized as yes (less than once a week, at least once a week, and almost every day) and no (not at all).     Figure 1 Schematic diagram of study participant selection