Study design and period
Cross-sectional study design was conducted using the EDHS data of 2016. The study was conducted from January 18 to June 27, 2016.
Study area
The study was conducted in Ethiopia (3o-14o N and 33o - 48°E), situated at the eastern horn of Africa. The country covers 1.1 million square kilometers and has a great geographical diversity, which ranges 4550 meters above sea level down to the Afar depression to 110 meters below sea level. There are 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.
Source population
All reproductive women age 15-49 years old in Ethiopia who gave birth in the last 5 years preceding the survey.
Study population
All women aged 15 to 49 years in the selected enumeration areas who gave at least one birth in the last 5 years preceding the survey.
Inclusion and Exclusion Criteria
Inclusion Criteria
All women who gave birth in the past 5 years before the survey in selected enumeration areas were included.
Exclusion Criteria
Women with unknown places of delivery out of the geographical positioning system were excluded.
Sample Size Estimation
A total of 15683 women aged 15-49 were interviewed in 2016 Ethiopian demographic health survey data (EDHS) 8490 women were excluded because they didn't have a birth in the five years preceding the survey 7590 women's have at least one birth in five years before the interview was included.
Sampling procedures
Administratively, regions in Ethiopia are divided into zones, and zones into administrative units called woreda. Each woreda is further subdivided into the lowest administrative unit, called kebeles. During the 2007 census, each kebele was subdivided into census enumeration areas (EA), which were convenient for the implementation of the census.
A stratified two-stage cluster sampling procedure was employed where EA is the sampling unit for the first stage and households for the second stage. In 2016 EDHS, a total of 645 EAs (202 in urban areas and 443 in rural areas) were selected with probability proportional to EA size (based on the 2007 housing and population census) and with independent selection in each sampling stratum. Of these 18,060 households were included; 7590 women.
Study Variables
Dependent variable
Institutional delivery (Yes or No)
Independent Variables
Individual Level of Institutional Variables
Socio-demographic Factors
|
Obstetrics Factors
|
Age at first birth
|
Antenatal care service
|
Husband’s education
|
Parity
|
Religion
|
Birth order
|
Residence
|
|
Maternal education
|
|
Wealth index
|
|
Maternal age
|
|
Marital status
|
|
|
|
Community Level of Institutional Variables
Region (Pastoralist, Agrarian, and City)
Place of residence
Community-level of poverty
Community-level of education
Operational definition
Institutional delivery: The main response variable in this study was institutional delivery whether women had for the most recent live birth or not by assessed the EDHS place of delivery by asking women “Where did you give birth?
‘1’ coded as if the woman gave birth to a health institution (Governmental and Non-governmental health center) ‘0’ coded as if the women gave birth elsewhere
Antenatal care utilization: Having at least one visit for ANC.
Community-defined as based on the primary sample unit in the EDHS data. Community-level variables were driven by aggregating individual-level variables, the aggregated were computed using average values of the proportion of women in each category of a given variable based on national median values the aggregated values categorized into groups (42).
Community poverty: the proportion of women in the community those live households in the lower (poor) quintile of wealth index; categorized as high (proportion of women greater than median national value), and whereas low (proportion of women below the median national value).
Community education: the proportion of women in the community who primary and secondary, categorized as high (proportion of women greater than median national value) whereas low (proportion of women below-median national value).
Region: The 11 regions of Ethiopia, which are delineated for administrative purposes, were categorized into three contextual regions pastoralist(Afar, Gambella, Benishangulgumez and Somali) Agrarian(Amhara, Tigris, Oromia, and SNNP) and City (A.A, Diredewa, and Hariri) defined based on the living conditions of their population (43)
Data collection and quality assurance
Data Source
Data were obtained from the nationally representative 2016 Ethiopian Demographic and Health Survey (EDHS) which used a two-stage cluster sampling design with rural-urban and regions as strata. Approval letter for the use of this data was gained from the Measure DHS.
Methods of data analysis
First data Extracted and cleaning by Stata 14.1 version. A multi-level logistic regression analysis technique was employed in this study to account for the hierarchal structure of the DHS data and the binary response of the outcome variable Bivariate multilevel logistic regression analysis was performed to estimate the crude odds ratios at 95 % confidence interval and those variables which were statistically significant were considered in the multivariate analysis. Finally, multivariate multilevel logistic regression analysis was performed to estimate the adjusted odds ratios and to estimate the extent of random variations between communities. In the multilevel models, the fixed effects (measures of association) estimate the association between the likelihood of institutional delivery and the individual and community level factors and were expressed as odds ratio with their 95 % confidence intervals. The random effects are the measures of variation in institutional delivery across communities expressed as Intracluster Correlation Coefficient (ICC) and Proportional Change in Variance (PCV).
Model comparison was conducted by using the Log-Likelihood ratio test and the model that maximum LLR was selected as a better-fitted model. Geographical Information System (ArcGIS version 10.6) and spatial Sat scan was used to analyze spatial data.
Spatial Autocorrelation Analysis
The spatial autocorrelation (Global Moran's I) statistic measures were used to evaluate whether the institutional delivery patterns are dispersed, clustered, or randomly distributed in the study area. Calculated Moran’s I values close to −1 indicate institutional delivery Is randomly distributed whereas Moran’s I close to +1 indicate institutional delivery clustered distributed. A statistically significant Moran’s I (p < 0.05) leads to rejection of the null hypothesis and indicates the presence of spatial autocorrelation. Local Moran’s I will be used to investigate the local level cluster locations of institutional delivery. Local Moran's I measure whether there were positively correlated (high-high and low-low) clusters or negatively correlated (high-low and low-high) clusters of high values (High-High), and clusters of low values (Low-Low). It also measures outlier in which high value is surrounded primarily by low values, and an outlier in which a low value is surrounded primarily by high values. Value for ‘I’ indicated that a case is surrounded by cases with dissimilar values; this case is an outlier.
Hot spot analysis (Getis-OrdGi* statistic)
Gettis-OrdGi* statistics were computed to measure how spatial autocorrelation varies over the study location by calculating Gi* statistic for each area. Z-score is computed to determine the statistical significance of clustering, and the p-value computed for the significance. The p-value associated with a 95% confidence level is 0.05. If the z-score is between −1.96 and +1.96, the p-value would be larger than 0.05, and could not reject the null hypothesis; the pattern exhibited could very likely be the result of random spatial processes. If the z-score falls outside the range, the observed spatial pattern is probably too unusual to be the result of random chance, and the p-value would be small to reflect this. In this case, it is possible to reject the null hypothesis and proceed with figuring out what might be causing the statistically significant spatial pattern in the data. Statistical output with high Gi* indicates “hotspot” whereas low Gi* means a cold spot.
Incremental analysis
Incremental spatial autocorrelation measures spatial autocorrelation for a series of distances and optionally creates a line graph of those distances and their corresponding z-scores. Statistically significant peak z-scores indicate distances where spatial processes promoting clustering are most pronounced. These peak distances are often appropriate values to use for tools with a Distance Band or Distance Radius parameter.
Incremental spatial autocorrelation
The incremental spatial autocorrelation of institutional delivery utilization showed that the maximum peak, where the spatial clustering is highly significant at a distance of 151366.65 meters, with a corresponding z- score of 25.53 (p-value <,0.01).
Ethical consideration
Ethical clearance was obtained from the Institutional Review Board (IRB) of the Institute of Public Health, College of Medicine and Health Sciences, University of Gondar. The survey data was received from the Measure DHS International Program which authorized the data-sets. All the data used in this study are publicly available, aggregated secondary data with not having any personal identifying information that can be linked to particular individuals, communities, or study participants. Confidentiality of data was maintained anonymously (See Additional file 3).