Maternal health is the health of women during pregnancy, childbirth, and the postpartum period [1]. Globally, approximately 287,000 women die each year (about 800 women every day) of pregnancy and childbirth-related causes [2, 3]. This is despite the Maternal Mortality Ratio (MMR) having dropped by 34% (339–223 maternal deaths per 100,000 live births) worldwide between 2000 and 2020 [3]. Achieving the Sustainable Development Goal (SDG) of MMR to less than 70 maternal deaths per 100,000 live births by 2030 is a critical global target [4]. However, the global average annual maternal mortality reduction rate was about one-third of the 6.4% needed to achieve the SDG by 2030 [5]. Maternal mortality rates also vary widely by country and region, with the highest rates of maternal deaths occurring in Low and Lower-Middle Income Countries (LLMICs) [3], with maternal deaths in Sub-Saharan Africa (SSA) accounting for more than two-thirds (70%) of the global burden [3].

In Ethiopia, there has been a 70% reduction in MMR from 1400 per 100,000 live births in 1990–420 per 100,000 live births in 2013, a decline of 5% per annum [6]. Despite this remarkable success, the World Health Organization (WHO) reported Ethiopia still had a very high MMR of 267 deaths per 100,000 live births equalling 3.6% of global maternal deaths in 2020 [3]. In contrast, Ethiopia’s population accounted for just 1.5% of the global population in 2022 [7]. In addition, about 30% of child-bearing-age women have unintended pregnancies, and 38% of pregnant women experience one or more complications, such as severe bleeding and prolonged labour in Ethiopia [8, 9]. However, the majority of maternal deaths are preventable through the provision of high-quality care during pregnancy, childbirth, and the postnatal period [10, 11].

The Ethiopian government is strengthening Reproductive, Maternal, Neonatal, and Child Health (RMNCH) interventions to end preventable maternal and child deaths and aims to achieve Universal Health Coverage (UHC) by 2030 [12]. The UHC service coverage index of 80% of essential health service utilization is a critical target to ensure access to quality health services by 2030 [13, 14]. Progress toward UHC requires ongoing efforts to improve health systems and expand access to quality health services for all [14]. Continuum of maternal health service (CMHS) is among the vital strategies to reduce maternal morbidity and mortality. Continuum of maternal healthcare service is the integrated and seamless provision of care and services to women throughout their pregnancy, childbirth, and postpartum period [15]. However, only 25.5% of Ethiopian women had received complete CMHS in 2022 [16]. The Ethiopian government provides maternal health services free of charge at public health facilities to ensure access to pregnancy and childbirth-related healthcare services [17].

Health service utilization is an approach that considers the perspectives of healthcare professionals, the production of services, and the preferences and circumstances of patients [18]. Applying health service theoretical models is critically important to explain various determinants of health service utilization [19]. Andersen's behavioural model for health service utilization is among these theoretical models that can be employed to elucidate the multifaceted factors influencing women’s utilization of healthcare services. This model provides a structured lens to analyse and interpret the complexities of healthcare utilization behaviours in terms of predisposing, enabling, and illness-specific factors that shape people's decisions on healthcare seeking [20, 21]. The first category, predisposing factors encompasses the socio-demographic and socio-cultural characteristics of respondents that precede their health conditions and offer insights on shaping individuals' healthcare behaviour. The second category is enabling factors, facilitators that empower individuals to access healthcare services effectively and explore the practical resources and support systems. The third category, need factors, focuses on the immediate determinants prompting individuals to seek health services. This dimension probes into the perceived health status of respondents, shedding light on the intrinsic motivations and urgencies driving healthcare utilization [20, 21]. Our conceptual framework is illustrated using these three components of Andersen’s health service utilization model (Fig. 1).

Conceptual framework for a continuum of maternal healthcare services

Early initiation of ANC visits, closer distance to health facilities, arranging birth preparedness and complication readiness plans, attending higher education, planned pregnancy, and autonomy of women from previous studies were the facilitators for CMHS utilization [16, 22]. However, no studies conducted at a national level have prompted the need for rigorous examination and analysis of the trends and contributing factors influencing the change in the proportion of complete CMHS utilization in Ethiopia. This study aims to uncover the factors for the change in complete CMHS utilization in Ethiopia using decomposition analysis. The findings will contribute to a nuanced understanding of the dynamics and offer actionable insights for policymakers seeking to enhance maternal healthcare programs in Ethiopia. This study promises to be a pivotal resource in informing evidence-driven decisions, ultimately fostering nationwide improvements in complete CMHS utilization.

Secondary data analysis was undertaken to measure the trends of complete CMHS utilization using Ethiopian Demographic and Health Survey (EDHS) data from 2011 to 2019. We used data from Tigray, Afar, Amhara, Benishangul-Gumuz, Gambela, Harari, Oromia, Somali, and South Nation, Nationality, and People (SNNP) regions of Ethiopia, and two city administrations (Addis Ababa and Dire-Dawa). The public healthcare sector in Ethiopia is organized into a three-tier system that aims to provide comprehensive healthcare services to the population of Ethiopia. The details of Ethiopia’s healthcare system are described in the 2015 Health Sector Transformation Plan (HSTP) of Ethiopia [12].

All women who had at least one ANC visit for their recent childbirth during the survey period from EDHS 2011 to 2019 were eligible for the study. The EDHS is a nationally representative survey conducted in Ethiopia to gather data on various demographic and health indicators. These surveys were conducted under the Ethiopian Public Health Institute (EPHI) Division of the Ethiopian Ministry of Health (MoH). The data is publicly available at https://www.dhsprogram.com/data/dataset_admin/login_main.cfm. The EDHS is based on the Population and Housing Census conducted by the Ethiopian Central Statistical Agency (CSA), and its sampling frame employed Enumeration Areas (EAs) [23,24,25]. These EAs are geographical units smaller than administrative units used for conducting surveys. In each EDHS from 2011 to 2019, each region of Ethiopia was stratified into urban and rural areas, resulting in a total of 21 sampling strata. Samples of EAs were selected independently in each stratum in two stages. In the first stage, a total of 624 EAs (437 rural and 187 urban) in 2011, 645 EAs (443 rural and 202 urban) in 2016, and 305 EAs (212 rural and 93 urban) in 2019 were selected with proportional allocation to EA size. In the second stage, household listing operations were performed in all selected EAs. On average, 27–32 households per cluster were selected proportional to the cluster size using systematic sampling. Relevant variables were extracted from Individual Record (IR) file datasets. In the data cleaning procedure, we applied the following phases to identify the eligible women in each dataset. In the first phase, we excluded women who did not give birth before the survey period. In the second phase, we excluded those women who did not have at least one ANC visit. In the third phase, we excluded missing data, including “do not know” responses and non-dejure residents. Finally, total weighted sample of 10,768 women (3333 in 2011, 4590 in 2016, and 2845 in 2019) were included in the study (Fig. 2). The details of sampling procedures are described in the EDHS reports [23,24,25].

Final sample sizes included in the analyses

Complete CMHS was the outcome variable. Socio-economic characteristics, such as the age of women, religion, maternal education, residence (urban vs. rural), current marital status, geographic region, age at first birth, household’s wealth index, obstetric history (e.g., the timing of the first ANC visit, urine and blood samples taken during pregnancy, measuring blood pressure during pregnancy, caesarean delivery for women’s most recent childbirth), radio ownership, and television ownership, and access to electricity, were covariates. Continuum of maternity care is an integrated and seamless provision of care and services to women throughout their pregnancy, childbirth, and postpartum period. This involves the use of four or more ANC visits, facility birth attendance, and postnatal care to ensure the health and well-being of the woman [15]. We categorized the responses for each broad category of maternal healthcare services as “no” and “yes” recoded as 0 and 1, respectively. If women reported “no” for one or more of these maternal healthcare services, we considered them as not receiving complete CMHS. On the contrary, women who reported “yes” to all three broad maternal healthcare services were considered to have complete CMHS [15]. In this study, we used the old WHO guideline [26], which recommends a minimum of four ANC visits as a standard of ANC care. This guideline was used across all surveys, including in 2019, as Ethiopia started implementing the updated 2016 WHO recommendations which require at least eight ANC contacts for a positive pregnancy experience from February 2022 onward. In addition, a woman was considered to have had a postnatal care visit if a woman had a health check before discharge, after discharge, or after delivery at home. Additionally, we operationalised some of the independent variables that needed operational definitions for further clarity. As such, we considered a woman had been aware of or being informed about pregnancy-related complications if she was told about at least one of the following signs of pregnancy complications: vaginal bleeding, vaginal gush of fluid, severe headache, blurred vision, fever, abdominal pain, and convulsion. A woman was considered to have an ANC visit in her first trimester or early initiation of an ANC visit if she had at least one ANC visit within her first three months of pregnancy.

Stata software was used to analyse the data (Stata Corp. 2023. Stata Statistical Software: Release 18. College Station, TX: Stata Corp LLC). After extracting relevant variables and data cleaning, we appended 2011, 2016, and 2019 EDHS data to show the trends and applied a multivariate decomposition analysis. Before any statistical analysis, the data were weighted using survey weight to adjust differences in the probability of selection to maintain the survey's representativeness and to get reliable estimates. Descriptive statistics were reported in the form of texts, figures, and tables.

We performed decomposition analysis to identify the main factors influencing the observed changes over time in the utilisation of CMHS amongst Ethiopian women. Firstly, we divided the survey rounds into three phases: Phase One 2011–2016, Phase Two 2016–2019, and Phase Three 2011–2019. Descriptive analysis was undertaken to show the variations in complete CMHS utilization stratified by various covariates over time. The multivariate decomposition analysis model decomposes the probability of the change in the outcome variable into two components, compositional and coefficient effects [27, 28]. Compositional (endowment) effects are the changes or differences in outcomes that arise due to the varying composition of a population or group, rather than changes in individual behaviors or characteristics. On the other hand, coefficient effects are the changes in outcomes that arise due to differences in the influence of specific factors (coefficients) or behavioral effects on the outcome of interest. We used mvdcmp STATA command to assess the contribution of covariates [27]. In this analysis, we assessed the probability of the change in the proportion of complete CMHS utilization using the reference group (EDHS 2011) and the comparison group (EDHS 2019). We defined A and B as the comparison year 2019 and reference year 2011 surveys, respectively. The logistic regression model for the logit function of complete CMHS utilization can be represented as follows:

$$} - B = F\left( \beta_ } \right)F\left( \beta_ } \right)\, = \underbrace \beta_ } \right) - F\left( \beta_ } \right)} \right]}}_ + \underbrace \beta_ } \right) - F\left( \beta_ } \right)} \right]}}_$$

(1)

In Eq. (1), we have chosen group A as the comparison group and group B as the reference group. The E component in Eq. (1) is attributed to compositional differences weighted by coefficients of the comparison group or endowments, usually called the explained component. The E component also reflects a counterfactual comparison of the difference in covariates from group A’s perspective (that is, the expected difference if group A were given group B’s distribution of covariates). The C component is attributed to changes in the coefficients weighted by the characteristics of the reference group, which is usually labelled as the unexplained component. The C component also reflects a counterfactual comparison of outcomes from group B’s perspective in which the expected difference if group B experienced group A’s behavioural responses to X [27]. In our study, E represents a counterfactual comparison of the difference in the characteristics of complete CMHS utilization for 2019 (that is, the expected difference in complete CMHS utilization if the 2019 survey was given the 2011 distribution of covariates). On the other hand, C depicts a counterfactual comparison of complete CMHS utilization from the 2011 perspective. This was the anticipated difference in complete CMHS utilization when the participants in the 2011 survey experienced similar behavioural responses with that of the 2019. Beta (β) coefficient and 95% confidence interval were used to determine the factors contributing to the change in proportion of complete CMHS utilization between 2011 and 2019.