Skilled birth attendant utilization trends, determinant and inequality gaps in Ethiopia | BMC Women’s Health

Study setting

Ethiopia has the second largest population in Africa with an estimated population of 114.9 million in 2020 [16]. The country has a four-tier health care system with a broad base primary health care unit (PHCU) serving the majority of the population. Skilled birth attendant services are given free of charge in all public health facilities with a reasonably functioning referral chain. Ethiopia has expanded free delivery services greatly along with increased demand creation campaigns including what is referred to in the country as the ‘pregnancy conference’ [17, 18].

Study design and population

This study was based on multiple EDHS surveys (cross-sectional surveys) conducted in the period 2000–2019. The surveys were conducted at approximately 5 years intervals. The study population constitutes women in the reproductive age group, 15–49 years of age, who gave at least one live birth within five years preceding the surveys. A demographic health survey (DHSs) is a standardized and nationally representative cross-sectional survey, it has a standard protocol for selecting study populations [19].

Sampling method

The DHS typically follows a two-stage stratified cluster sampling procedure to select a nationally representative sample proportionate to the population size of each regional state in the country. The sampling frame was prepared for each round based on the most recent population and housing census. The lowest sampling unit is an Enumeration Area (cluster), which is a geographic area consisting of several dwelling units which served as a counting unit for the census. Enumeration Areas (EAs) were selected using probability proportional to population size. A complete household listing was carried out in all of the selected EAs. In each EA, a fixed number of households per cluster were selected with an equal probability of selection from each newly prepared household list. All women aged 15–49 in the selected households were included in the survey [19,20,21,22].

Data collection

Data were collected using a structured questionnaire consisting of several modules including household and women modules that are relevant to this study [16]. The data were collected through face-to-face interviews with eligible women aged 15–49 who gave birth within five years preceding the survey at home by trained data collectors. If the woman had more than one child in the five years preceding the survey, information on the use of delivery assistance was collected for the last birth [16, 23]. The data for this analysis were accessed by requesting the DHS program website ( after explaining the purpose of this study. The dependent variable of this study is skilled birth attendant coverage and the main independent variables were household wealth status, women’s education, and residency area.

Data analysis

The national coverages of skilled birth attendants were calculated by dividing the number of women who reported having skilled delivery by the total number of pregnant women in each survey. The projected coverage for 2025 was derived by calculating a smoothed yearly coverage and compared with the target set for 2025 in HSTP II. The annual increment between surveys was calculated by subtracting the coverage of the last survey from the current survey and then dividing it by the number of years between the two surveys [24].

The equality analysis was done for household wealth quintiles (lowest or poorest, second, middle, fourth, highest or richest), maternal education (No education, Primary school, and Secondary school and above), and place of residence (rural/urban) using the WHO HEAT model [25].

The Toolkit is a software application that facilitates the assessment of health inequalities by using disaggregated data and summary measures through visualized interactive graphs, maps, and tables. The relative concentration index was performed for household wealth status and maternal education and a graph of the urban–rural ratio was done for a place of residency.

The inequality analysis was done for the year 2016 since the data from 2019 was not publicly available.

Study variables were re-coded to meet the desired classification. To overcome the unbalanced distribution of regional samples for national estimates, sampling weights were used during analysis where the inverse probabilities of selection for each observation allow us to reconfigure the sample as if it was a simple random draw of the total population and hence yield accurate population estimates for the main parameters of interest. It also provides a measure of how many individuals a given sampled observation represents in the population.

After searching for literature, the following nine independent variables were included in the analysis namely: Maternal education, wealth status, residence, maternal age, partner’s level of education, number of children, distance to a health facility, maternal occupation, and husband occupation. These independent variables were selected due to their positive association with skilled birth attendance [19, 26, 27].

All analyses were done using Stata version 14. Inferential analysis was used to examine the relationship between the independent variables and the outcome variable. Specifically, binary logistic regression was conducted. All results of the binary logistic analyses were presented as odds ratios (ORs), with 95% confidence intervals (CIs). Variables that show a statistically significant association (p < 0.05) at the bivariate level were further analyzed at the multivariate level by logistic regression. All the variables were included in the multivariate model once they were significantly associated at the bivariate level. This is because these variables showed an influence on the outcome variable and there is a need to identify whether each has been confounded by another variable or not. The adjusted odds ratio (AOR) was used to determine the presence of an association between the dependent and independent variables for which 95% CI was determined. The final regression model was tested for fitness using the Pearson goodness of fit test.


Leave a Reply

Your email address will not be published. Required fields are marked *