Ethiopia is located in the Horn of Africa and shares a border with Eritrea, Djibouti, Somalia, Sudan, South Sudan, and Kenya. The country covers an area of 1.1 million km2 (square kilometer) with geographical diversity, ranging from 4,550 meters (m) above sea level down to the Afar depression 110m below sea level, which is comprised of over 80 ethnicities and speaking over 80 different languages (19). Administratively, Ethiopia is divided into nine regional states and two city administrations subdivided into 68 zones, 817 districts, and 16,253 kebeles (lowest local administrative units of the country) in the administrative structure of the country(20). Based on the 2018 world bank report Ethiopia had a total population of 109 million with a gross national income per capita of US$ 790 (21). Ethiopia's health system comprises three tiers: a primary health care unit, a general hospital, and a specialized hospital (22, 23).
Source of data
The data came from EMDHS 2019, specifically the under-five children's file (KR) (http://www.measuredhs.com). We were able to download the datasets after the measurement program allowed us to do so. The unweighted sample consisted of 5753 women who had live births in the five years preceding the survey. The 2019 EMDHS sample was stratified and selected in two stages, and interviews were conducted face-to-face with permanent residents and visitors who stayed in the residences the day before the survey. The 2019 EMDHS sampling frame is a composite of all census enumeration areas (EAs) created for the upcoming 2019 Ethiopia Population and Housing Census (PHC) conducted by the Central Statistical Agency (CSA). The census frame includes the complete list of 149,093 EAs created for the 2019 PHC. An EA is a geographical area with an average of 131 households. The sampling frame includes data on the EA's location, type of residence (urban or rural), and the estimated number of residential households (24)
The outcome variable for this study was institutional delivery, which was coded as “0” if the women gave birth at home and “1” if the women gave birth at a health facility. Institutions/facilities delivery was stated as the births at health institution/facility within five years afore the survey.
Individual level factors were women education level, household wealth index, birth interval, number of antenatal care visit, age of the women, media exposure and marital status.
Community level factors were residence, region, community educational status, community ANC coverage and community poverty. The EMDHS did not collect data that can directly describe the clusters’ characteristics except the place of residence and region. Therefore, other common community-level data were generated by aggregating the individual characteristics with our interest in a cluster. The aggregates were computed using the proportion of a given variables’ subcategory we were concerned on in a given cluster. Since the aggregate value for all generated variable was not normally distributed. It was categorized into groups based on the national median values.
Exposure to mass media: a frequency of listening to the radio and watching television were considered exposure to mass media in this study by excluding exposure to magazines and newspapers. So, women exposed to either television or radio at least once per week considered exposed, if not exposed at all, taken as not exposed (20).
Community women education: Was defined as the proportion of mother’s who attended primary/secondary/ higher education within the cluster. The aggregate of individual mother’s primary/secondary/higher educational attainment can show the overall educational status of women within the cluster. There were two categories for this variable with reference to the national median value: higher proportion of mother’s who attended primary/secondary/higher education and lower proportion of mother’s who attended primary/secondary/higher education within the cluster.
Antenatal care utilization: was defined as mothers who had at least four antenatal care visit(25).
Community antenatal coverage: The proportion of women in the clusters who had four and above antenatal care (ANC) from a skilled provider during the pregnancy of last delivery.
Community poverty status: It is defined as the proportion of poor or poorest mothers within the cluster. Within the cluster proportion of poor or poorest were aggregated and show over all poverty status within the cluster.
Data management and statistical analysis
The 2019 EMDHS data were pre-tested before the actual data collection. Data collectors had received training in interviewing techniques, field procedures, the content of the questionnaires, and how to administer both paper and electronic questionnaires; after all, questionnaires were finalized in English, then translated into Amarigna, Tigrigna, and Oromiffa(24). Since this was secondary data, the data were maintained by processing, editing, raw coding data, and re-coding, checking its completeness, and cleaning the missing values by running frequencies based on the research's interest. Sample weights were applied to compensate for the disproportional probability of sampling and non-response rate between the strata that have been geographically defined. A detailed explanation of the weighting procedure can be found in the EMDHS final report (24).Cross tabulations and summary statistics were used to describe the study population.
The aggregated home and health facility delivery count data were joined to the geographic coordinates based on each cluster unique identification code. Global spatial autocorrelations were assessed with ArcGIS version 10.5 using the Global Moran's I statistic (Moran's I) to evaluate whether the pattern expressed is clustered, dispersed, or random across the study areas. Moran's I values close to −1 indicated institutional delivery were dispersed, whereas I values close to +1 indicated institutional delivery were clustered, and distributed randomly if I value was zero. A statistically significant Moran's I (p < 0.05) led to the rejection of the null hypothesis, and indicated the presence of spatial autocorrelation as well as it detect the existence of at least one cluster, but not the specific location of the cluster(s) (26).
For positive global spatial autocorrelation, local spatial association indicators were used to assess clusters and outliers by comparing the values in each specific location with values in neighboring locations. It allows for decomposing the pattern of spatial association into four categories (quadrants) called Hot spot analysis (27). And this help us to identify the proportion of institutional delivery based on sampled enumeration area. Since geographic coordinates were collected at the cluster level, the unit of spatial analyses was 2019 EMDHS clusters. Finally, we employed Kulldorff's purely spatial scan statistic method using the Bernoulli probability model in SaTScan version 9.6 software to detect the local spatial clusters of areas with high home delivery. Its output presents the hotspot areas in circular windows, indicating the areas inside the windows are higher than expected distributions compared to the areas outside of the cluster windows (28). We used a maximum 50% of the population at risk for the spatial cluster size. A cluster was statistically significant if a p-value < 0.05. Interpolation- we run the empirical Kriging technique to predict values for areas where data points were not taken.
First a descriptive analysis was conducted for all individual- and community-level variables in order to examine the characteristics of the sample. Considering this hierarchical nature of the data and the assumption of independence among individuals within the same community and the assumption of equal variance across the community is violated in nested data. Therefore, flat models could underestimate the effect sizes' standard errors and lead to bias (loss of power or type I error), affecting the null hypothesis (29). Hence, in order to account the hierarchical nature of the EMDHS data and response variable multilevel logistic regression analysis was implemented to test the effect sizes of individual and community level factors on women’s decision to place of delivery. During analysis, the characteristics of women were taken as individual level (level-1) and characteristics of clusters were treated as community level (level-2).
Logit(p)= log(p/1-p)= β0j + β1jX1ij + β2jX2ij+……..βqjXqij
Since Β0j = γ00 + γ0sZsj+u0j
Βqj = γq0Xqij + γqsZsjXqij+uqjXqij
Or Logit(p) = γ00+γ0sZsj + γq0Xqij + γqsZsjXqij+u0j+uqjXqij
Where we have “q” explanatory variables at the lowest level and “s” explanatory variables at the highest level,
j = subscript indicates that this case belongs to the jth group
ij= subscript indicates that ith individual within jth group
γ00 = the overall intercept (fixed part)
Boj= is the random intercept varying at community level (group specific intercept)
Level one employs β s, while level two employs γ s regression coefficient
U0j = Error term of the intercept or deviation from the average intercept
Uqj = Error term of slope βqj or deviation from average slope βq due to level-2 explanatory variable Zsj
The intercept γ00 and slopes γ0s and γqs are fixed effects whereas uoj, uqj are random effects of level-2.
Model I (Empty model) was fitted without explanatory variables to test random variability in the intercept and to estimate the intra class correlation coefficient (ICC).
The intra-class correlation (rho)
Where σ2uo = variance due to group level error term (uoj) and π2/3 is level-1 variance.
Model II examined the effects of individual level characteristics, Model III examined the effect of community level variables and Model IV examined the effects of both individual and community level characteristics simultaneously. The p value <0.05 was considered as statistically significant. For measurements of variation (random effects), intra-class correlation coefficient (ICC), median odds ratio (MOR), and proportional change in variance (PCV) statistics were computed. Model comparison was made based on Akakie Information Criteria (AIC) and Deviance Information Criteria (DIC). The model with the lowest information criterion was considered to be the best fit model (29).
For this study, ethical clearance was taken from the ethical review committee of Salale University College Medicine and Health Sciences (CHS) with approval and a supporting letter. The EMDHS 2019 data were then obtained and used with the Central Statistical Agency of Ethiopia’s prior permission. We registered for dataset access and wrote the study’s title and significance on the website after completing a short registration form. Downloading of datasets was done using the accessed website at http://www.measuredhs.com on request with the help of ICF international. Downloading data were used only for this study. The dataset was not passed on to other researchers without the consent of EDHS. All EDHS data were treated as confidential, no need to identify any household or individual respondent interviewed in the survey.