A Multilevel Analysis Factors Associated With Diarrhea Among Under-ve Children in Ethiopia Using 2016 EDHS Data: Individual and Community Level Factors.

Background: Despite numerous attempts, diarrhea remains one of the leading causes of death among children under the age of ve around the world. Globally, diarrhea is the second prominent cause of death in children next to pneumonia. Every year, there are 1.7 billion children who develop diarrhea and at least 525,000 die from diarrhea. In low- and middle-income countries such as Ethiopia, diarrhea still remains one of the health problems due to its high morbidity and mortality. This study was conducted identify factors associated with diarrhea at community (cluster) level and individual level. Objectives: To assess individual and community level factors of childhood (0–59months) diarrhea in Ethiopia by using the 2016 Ethiopian Demography and Health surveys (EDHS) data, 2020. Methods: A cross-sectional secondary analysis of data pooled from 2016 Ethiopian Demographic and Health Surveys (EDHS) data was used. The analysis was done using Stata version 14.2. A multilevel logistic regression model was used to identify independent predictors of childhood diarrhea. Odds ratio with 95% CI was used in identifying the association between dependent and independent predictors Result: The prevalence of diarrhea in Ethiopia based on EDHS 2016 was 11.78%.The odds of diarrhea among children reside in rural area were 1.84 times more likely to develop diarrhea (AOR=1.82; 95% CI: 1.52-2.16) as compared to urban dwellers. Those children aged between 13 and 24 months were 2.2 times more likely to have diarrhea than (AOR=2.2, 2.15-2.98) their older counter parts (48-59 months). The measure of variation was also assessed by using ICC, MOR, and DIC with the result of 10.08, 1.56 and 316.18 respectively. Conclusion: Our ndings identied that childhood diarrhea was affected by not only individual level factors but also community-level factors. At the individual level (age of the women, number of under ve children in the households, age of the child, number of family members, maternal education, and the number of under-5 children) and the community-level, place of residence were signicant factors associated with childhood diarrhea in Ethiopia.


Introduction
Globally, diarrhea is the second prominent cause of death in children next to pneumonia. Every year, there are 1.7 billion children who develop diarrhea and at least 525,000 die from diarrhea. Child under 3 years old experience about 3 episodes of diarrhea per year (1). In low-and middle-income countries such as Indonesia, diarrhea still remains one of the health problems due to its high morbidity and mortality (2).
The passing of three or more loose or liquid stools a day is considered as diarrhea (or more frequent passage than is normal for the individual). Passing shaped stools on a regular basis is not diarrhea, nor is passing loose, "pasty" stools by breastfed infants. (1). Diarrhea is objectively de ned as passing a stool volume greater than 200 ml or weight 200g per 24 hours(3).
According to a 2016 study, diarrhea affected 23 % -25 % of those aged 6-23 months. The children of families who use unprotected wells have the highest incidence of diarrhea (18 %). In many developing countries, diarrhea-related deaths are still common. For example, according to a 2012 study, infant and under-ve mortality rates are still high in Ethiopia. (4,1).
In 2008, the global rotavirus-associated mortality among children aged less than 5 years was estimated to be 453,000 deaths, accounting for 37 percent of diarrhea-related deaths and 5% of all deaths in children aged less than 5 years (5).
The World Health Organization (WHO) recommends that all countries, particularly those with a high rate of diarrhea-related mortality among children under the age of ve, use rotavirus vaccines on a regular basis.
In Ethiopia, few rotavirus vaccines have been implemented into private or public health programs. Prior to implementing new vaccines, the WHO suggests that countries perform local surveillance studies.(6).
Many strategies were achieved by the Millennium development goals in the last 15 years and the strategy of Sustainable Development Goals (SDGs) is on the way, but still the prevalence of diarrhea is high in many parts of the world. Despite the emphasis given by the Ethiopian ministry of health and the respective regional health o ces to improve child health, still many children are dying due to easily preventable and treatable diarrheal disease in Ethiopia. It's important to understand the causes of diarrhea in children under the age of ve at community and individual level in order to learn the best preventive approach and formulate an effective strategy. As a result, this study was conducted to look into the individual and community level determinants of diarrhea among under ve children in Ethiopian based on EDHS 2016. It has to be done a lot to prevent the burden of diarrhea in this age group. Therefore, in an effort to combat the problem, this study contributed by exploring factors associated with diarrhea at community (cluster) level and individual level This in turn serve as an input for policy makers, others stakeholders or anybody it may concern to formulate an effective solution on these common factors that associated with diarrhea among this group age children speci cally where the prevalence is high in Ethiopia. Different Organizations can also use the nding of this study for assessing and evaluating interventions towards maternal and child health care which enables them redesign it accordingly. In addition, it can be used as a baseline for further research. In addition, it can be used as a baseline for further research.

Methods And Materials
Study Area and setting Ethiopia is situated in the northeastern part of Africa. It found between 30 and 150 north latitude and 330 and 480 east longitudes (40). It is bounded by six countries, namely: Eritrea, Djibouti, Kenya, Somalia, South Sudan and the Sudan, and the country covers an area of 1.1 million square kilometers ranging from 4,620m above sea level (41).The main climate type is tropical monsoon, with temperature climate on the plateau and hot in the lowlands. There are topographic-caused climatic variations broadly categorized into three: the "Kolla", or hot low lands which are found up to approximately 1,500 meters above sea level, the "Wayna Degas" which ranges between 1,500-2,400 meters above sea level and the "Dega" or cool temperate highlands 2,400 meters above sea level. Ethiopia has a total of 114,963,588 populations; of this 18,394,194 are under-ve children (41). The country has 20% of urban area, Government Expenditure on Health Care is 4 % of the total expenditure in the country. (42).
The global report on food crisis indicates that, Ethiopia's GDP Per Capita purchasing power parity (PPP) in the country is US $1916.1 and it has a poverty rate of 29.6%. The Human Development Index (HDI) rank is 174 among 187 countries and there are 9.7 Million people who are food insecured in the country. About 85% of the population is dependent on Agriculture in Ethiopia (41). There is a fertility rate of 4.6, infant mortality rate of 48 (per 1,000 live births) and child mortality rate of 67 per 1,000 live births in the country (43).

Study design and study period
A cross-sectional study design using secondary data from 2016 EDHS was conducted from February to May.

Source and Study Population
The source population is all Ethiopian children aged 0-59 months, who are living with the contemporary respondent, while the 2016 EDHS are being conducted. The study population was children who are incorporated in the 2016 Ethiopia Demographic and Health Surveys.
Sample Size determination and sampling procedures A total of 10,641children from EDHS 2016 were included from nine geographical regions and two administrative cities of Ethiopia. The sampling frame used for 2016 EDHS was the Ethiopian Population and Housing Census (EPHC), which was conducted in 2007 by the Ethiopia Central Statistical Agency (CSA) with Federal Ministry of Health (FMOH). The sampling frame contains information about the Enumeration areas (EA) location, type of residence (urban or rural), and estimated number of residential households. The samples for 2016 EDHS are designed to provide estimates of key indicators for the country as a whole, for urban and rural areas separately, and for each of the nine regions and the two administrative cities.
The 2016 EDHS sample was strati ed and selected in two stages. Each region was strati ed into urban and rural areas, which yielded 21 sampling strata. Samples of EAs were selected independently in each stratum in two stages. Clear strati cation and proportional allocation was achieved at each of the lower administrative levels by sorting the sampling frame within each sampling stratum before sample selection, according to administrative units in different levels, and by using a probability proportional to size selection at the rst stage of sampling.
From the rst stage, a total of 645 EAs (202 in urban areas and 443 in rural areas) were selected in 2016 EDHS and with probability proportional to EA size (based on the 2007 PHC) and with independent selection from each sampling stratum for this recent survey. A household listing operation was performed in all of the selected EAs. The resulting lists of households served as a sampling frame for the selection of households in the second stage. In the second stage of selection, a xed number of 28 households per cluster were selected with an equal probability systematic selection from the newly created household listing. Totally, 18,008 households were selected of which 17,067 are occupied. In all of the selected households, height and weight measurements will be collected from children aged 0-59 months, and women aged 15-49.The 2016 EDHS sample contained 18,008 households from 645 clusters, and 15,683 women aged 15-49 years were interviewed; the response rate was 95%.

Inclusion and exclusion criteria
Inclusion criteria: Children 0-59 months of age, complete data on diarrhea status and children with their mothers for measurements and completion of the questionnaires during survey.
Exclusion criteria: Children whose age is unknown, children who were not with their mothers/caregivers for and children diarrhea status for last two week before survey were not measured excluded from the study.

Study variables
Dependent variables was diarrheal status of the child (0-59months)

Independent variables
Individual level factors Socio-Demographic and Economic characteristics of the family (household) include age of child, sex of child and wealth status. Under maternal and child related factors incorporate education level, maternal Age, child nutritional status, maternal or caregiver's economical and educational status, hand washing practice during critical times, household wealth index, vaccination status of the child, child feeding practice, maternal employment status and maternal or caregiver's diarrhea history. Community level factors were source of drinking water, unprotected water sources, improper disposal of wastes, family member size, toilet facilities, no of Children in household, place of residence and region.

Operational de nition
Under Five Childhood Diarrhea: -It was de ned as the presence or absence of UFCD during the last two weeks prior to this study (44). Improved drinking water sources: -included piped water, public taps, standpipes, protected shallow wells, and springs.
Unimproved water sources: -were surface waters, unprotected shallow wells, and unprotected springs.
Improved latrines: -included any non-shared toilet, mostly ush toilets connected to piped sewer systems, septic tanks, and pit latrines, ventilated improved pit latrines, and pit latrines with slabs.
Unimproved latrines: -included the pour-ush type not connected to a sewer, pit latrines without a slab, open pits, and open defecation. Critical hand washing times were nominated as; before food preparation, eating, supplementary child feeding, and breastfeeding and after defecation and cleaning a child after defecation (43).
Vaccination status:-was considered fully vaccinated at 12 months, if the child had received the following vaccinations in the rst year of life and have con rmed immunization card (45).
Appropriate child feeding practice:-Early initiation of breastfeeding within rst hour of birth, exclusive breastfeeding for the rst six months followed by continued breastfeeding for up to two years and beyond with appropriate complementary foods after completion of 6 months(46).
Family member size: -number of family member living with fewer than ve children.
High birth order: -when the birth order duration is less than one year gab.

Data collection methods and Tools
Before extracting the data, permission was obtained from demographic health survey data center by persuading the purpose of the study. After getting permission, data were downloaded in Stata software data set form. Stata version 14 was used for cleaning and analysis of data.
For the purpose of this research, the under-ve children data separately was received from central statistical agency by registered in it. The link was received and downloaded.

Data Processing and Analysis
Data cleaning was done to check for consistency and missing value. Recoding, labeling and exploratory analysis was performed. Categorization was done for continuous variables using information from different literatures and re-categorization was done for categorical variables accordingly. Model tness was done by using Log likelihood, AIC, BIC and ICC. Sample weights were applied in order to compensate for the unequal probability of selection between the strata that are geographically de ned, as well as for non-responses.
Multilevel analysis was conducted after checking that the data is eligible for multilevel analysis that means ICC greater than 10%. ICC calculation formula is as follows: : Is the community (cluster) level variance.
: Is the standard logistic distribution, that is, the assumed individual variance?
Since DHS data are hierarchical, i.e., individuals (level 1) were nested with in communities (level 2), twolevel mixed-effects logistic regression model was tted to estimate both independent ( xed) effects of the explanatory variables and community-level random effects on childhood diarrhea.
Because the log of the probability of childhood diarrhea was modeled using two-level multilevel model as follows; Where, i and j are the level 1 (individual) and level 2 (community) units, respectively; X and Y refer to individual and community-level variables, respectively; is the probability of childhood diarrhea for the i th under-ve child in the j th community and is probability of under-ve child don't have diarrhea for the i th under-ve child in the jth community(cluster). The β's was the xed coe cients. Whereas, β0 is the intercept-the effect on the probability of childhood diarrhea in the absence of in uence of predictors; and showed the random effect (effect of the community on childhood diarrhea) for the jth community and showed random errors at the individual levels. By assuming each community had different intercept (β0) and xed coe cient (β), the clustered data nature and the within and between community variations were taken in to account.
Bi-variable and multilevel logistic regression analysis was done and model tness was checked. Those variables which had P-value less than 0.25 were candidates to build model 3 (model-3). After this analysis was performed, four models were constructed for this multilevel logistic regression analysis. The rst model was an empty model or null model without any explanatory variables, to evaluate the extent of the cluster variation on diarrhea among under ve children. The second model was adjusted for the individuallevel variables; the third model can be adjusted for community (cluster) level variables while the fourth model was adjusted for both the individual and community level variables simultaneously.
The measures of association ( xed-effects) estimates the associations between likelihood of under-ve children diarrhea and various explanatory variables were expressed as Adjusted Odds Ratio (AOR) with their 95 % Con dence level. A variable in which the con dence interval does not include the null value and p-value < 0.05 can be used to de ne statistical signi cance.
The measures of variation (random-effects) were reported by using intra-cluster correlation (ICC), Median Odds Ratio (MOR) and proportional change in variance (PCV). PCV was calculated to measure the variation between clusters. ICC was used to explain cluster variation while MOR is a measure of unexplained cluster heterogeneity (39).
The ICC shows the variation in under-ve diarrhea due to community characteristics. The higher the ICC, the more relevant was the community characteristics for understanding individual variation in childhood diarrhea. The ICC was calculated as follows:  procedures, a detailed review of questionnaire content, instruction on how to administer the paper and electronic questionnaires, mock interviews between participants in the classroom, and practice interviews with real respondents in areas outside the survey sample. Data quality management during the recent EDHS survey was published (43). An Initial exploratory analysis was conducted to check for outliers, missing and consistency of dataset.

Result writing and dissemination plans
The result of the study will be written and presented to Debre Berhan University, College of Health Sciences, and Department of Public Health. The nding of this study will be disseminated timely to all relevant stakeholders that may concern themselves. It will be sent for publication in scienti c journal, and online dissemination will be taken into account. Presentations on different occasions will be made on various seminars, workshops and scienti c conferences. were headed by male. Among the total households, 55% of them have more than 6 family members where as 45% of them have less than 5 members in the household. During multilevel binary logistic regression those variables like children who have taken measles vaccine, Rota 1 and Rota 2 vaccine, vitamin A supplementation, age of child between 12-23 months, household with a family member of 6 and above, unimproved toilet facility, mothers with no formal education and mothers age greater than 35 years were candidates for the nal model of multilevel analysis. Table 2 shows unadjusted or crude odds ratio (COR) results that were obtained when we are taken into account the effect of only one independent variable in the analysis. On binary multilevel logistic regression analysis, residence, educational status the mother, toilet facility, number of family members, sex of the household head, age of the child and Rota virus vaccination were associated with the outcome variable diarrhea(P-value < 0.25). (Table 2) The odds of diarrhea among children reside in rural area were 1.84 times more likely to develop diarrhea  -59 months). The odds of diarrhea among children whose mother had no formal education were 31% higher (AOR = 1.31(1.20-2.45) than children whose mother had attended formal education.
The odds diarrhea among children with more than ve family members in the household were 16% (AOR = 0.84(0.73-0.96) lower than households who have less than ve family members. Children live in households with more than 2 under ve children were 1.56 times higher (AOR = 1.56; 95% CI: 1.29-2.96) odds of experiencing diarrhea than families with greater than two under-5 children. (Table 3) Table 3 Predictors associated with childhood diarrhea recognized by multilevel multivariable logistic regression models.  In this study, the prevalence of diarrhea was 11.78% which was lower than other studies. This discrepancy might be due to the strengthening of the Health Extension Program (HEP), improving access to health care to meet the primary health services and the introduction of integrated community cause management program (17,21,22,23).
At the individual level, variables such as age of the child, sex of the child, maternal educational status, number of family members, sex of the household head, and numbers of under-5 children were signi cantly associated with childhood diarrhea. Similarly, at community-level residence and toilet facility shared with others were signi cantly associated with childhood diarrhea.
It is acceptable that the educational status of mother is more likely to in uence childhood diarrhea and educated mothers have a positive in uence on hygienic practices. In this study, the odds of diarrhea were higher among children whose mothers had no formal education than children whose mothers had formal education. The study ndings are consistent with other studies, which found higher odds of childhood diarrhea among children whose mothers had not attended formal education in Ethiopia, (27,39,42) Kenya (36).
In this study, the odds of diarrhea among children reside in rural area were 1.84 times more likely to develop diarrhea as compared to urban dwellers. This nding is supported by a study conducted in west Gojjam, Ethiopia (41). This could be explained by children and mothers who live in urban area may have good awareness about hand hygiene, sanitation and access to toilet facility. They could also have access to improved water sources.
In the present study, those children aged between 13 and 24 months were 2.2 times more likely to have diarrhea than their older counter parts (48-59 months). This nding is in line with a study done in Arba Minch, Benchi Maji, Sodo and EDHS 2011. (25,26,27,40). This can be explained by children during this age group start supplementary feeding. Besides, it is a time of crawling for children and they eat dirty particles whatever they get in the oor and mothers within this age category mayn't have experience on child care, good feeding and hygiene practices.
According to present study, the odds of diarrhea among children with more than ve family members in the household were lower than households who have less than ve family members. This is in line with a study conducted in Indonesia (37). This may be due to households having ve or more family members will get attention towards hygiene practice because they could get coach or see from the family members.
As a result, a child living in households with more than ve family member's under-5 children becomes less vulnerable to diarrhea. On the other hand, children in households having less than ve family member's under-5 children lack experience and necessary support from their older sibling toward toilet training and other sanitary practice, which possibly associate with childhood diarrhea.
The odds of experiencing diarrhea among households headed by female were 2 times higher (AOR = 2.06, 95%CI: 1.12-3.76) when compared with household headed by male. The possible explanation can be due to work overload on female when there is loss of partner or lack of support from family members'. As a result, they may lack time to give care and coach the children.
In our study, Children live in households with more than 2 under ve children were 56% higher odds of experiencing diarrhea than families with greater than two under-5 children. This is consistent with a study done in Soddo, Bahirdar Zuria, West Gojjam, Gondar and Tigray (40, 33, 41, 32, and 42). This can be explained by mothers mayn't get enough time to keep the hygiene and provide care and support to the children. As a result, it became a challenge to taking care of multiple young children.
The odds of diarrhea among children who have shared toilet with neighbors were 2.64 times higher when compared with children who haven't shared toilet facility with neighbors. This nding is consistent with a study conducted at Senegal, melbour. The possible explanation when toilet facilities shared by neighbor or de jure, the toilet may be contaminated by different infectious agent and can easily acquire by the children while the use the toilet facilities.

Strength and Limitation of the Study
We had used a multilevel model for analysis to take into account the clustered nature of the data and possible to know the individual level effect, community level factors effect and mixed effect and increase the accuracy of estimates. In addition to this, it can prevent ecological and atomistic fallacy. As a limitation of this study, we can't nd variables from the survey like health insurance, distance to health facility. Since it is a survey, there may be a possible recall bias and seasonal variation is not considered because the dataset is collected within a speci ed period of time.

Conclusion
Our ndings identi ed that childhood diarrhea was affected by not only individual level factors but also community-level factors. At the individual level (age of the women, number of under ve children in the households, age of the child, number of family members, maternal education, and the number of under-5 children) and the community-level, place of residence were signi cant factors associated with childhood diarrhea in Ethiopia. The ndings show that there is a need to consider some of the modi able factors in the existing interventions in order to improve child health outcomes at country level.

Recommendations
At Ministry of Health, designing innovative approaches to combat communicable diseases especially for under ve children and better to design peculiar policy for rural resident mothers.
The Regional Health Bureau ought to give priority on offering services for mothers who had not attended The authors declare that they have no competing interests among authors.

Ethics and consent
Ethical approval was obtained from the Ethical Review Committee of Debre Berhan University. Letter of authentication was obtained from DHS program before undertaking the study. Two-stage cluster sampling and sampling data frame EDHS 2016. Figure 2 shows the prevalence of Diarrhea based on EDHS 2016 data