Study design and setting

A community-based cross-sectional study was conducted from November 5 to December 5, 2024 to examine determinants of the completion of continuum of maternal healthcare services among rural women in East Shoa Zone, Eastern Ethiopia. The zone is located approximately 100 km southeast of Addis Ababa covering an area of 10,232 km² within the Great Rift Valley. It has an estimated population of 1,386,490 of whom 306,830 are women of reproductive age. Administratively, it comprises 11 woredas, 2 town administrations, and 292 kebeles [14]. The study setting is characterized by dispersed rural settlements and travel-time barriers to health facilities which may affect women’s continuity of maternal healthcare utilization [15]. Health services are delivered within Ethiopia’s three-tier healthcare delivery system with potential variability in referral linkages and facility readiness across service levels [16]. The public health service infrastructure comprises 3 primary hospitals, 51 health centers, 5 comprehensive health posts, and 227 basic health posts [14]. Despite this service-delivery structure, maternal health service coverage remains substantially lower than the corresponding regional averages. According to the DHIS2 routine health service report, antenatal care, skilled birth attendance, and early postnatal care coverage were 71%, 52%, and 66%, respectively, compared with the regional averages of 97%, 70%, and 94% [17]..

Study population and eligibility criteria

The source population consisted of women residing in rural areas of East Shoa Zone, Oromia Region who had given birth during the last 12 months preceding the survey period. The study population included all eligible women in the selected kebeles who had given birth during the 12 months before the interview date. Women who were critically ill or had lived in the kebele for less than 12 months at the time of data collection were excluded.

Sample size calculation and sampling procedure

The required sample size was calculated using a single-population proportion formula, assuming an estimated prevalence of 21.5% for completion of the maternal continuum of care [18], and a 95% confidence level. A 4% margin of error (d = 0.04) was used instead of 5% to increase the precision of the estimate, acknowledging the associated increase in sample size.

To account for clustering, a design effect was incorporated into the sample-size calculation. Although the design effect is commonly estimated as DEFF = 1 + (m − 1) ρ, no reliable ICC estimate was available from a closely comparable study. Assigning an arbitrary value to ρ could have resulted in an inaccurate adjustment. Therefore, a conservative design effect of 2.0 was adopted to accommodate the anticipated variation in kebele sizes and the context-specific nature of the ICC. Finally, we included a 10% allowance for non-response. This yielded a final calculated sample size of 901 participants. Following establishment of the sampling frame and proportional allocation across kebeles, eligible women selected through systematic random sampling were traced to their households for participation.

During field implementation, 1,038 eligible women were identified and recruited from the sampling frame, of whom 997 completed the interviews. During the house-to-house survey, public health facilities providing at least basic maternal health services to the selected communities were also identified. Health centers that had been operational for at least one year prior to the data collection period were eligible for inclusion in the facility-level assessment. A service utilization–based linkage approach was used to integrate household and health facility data. Women were asked to identify the primary public health facility where they received maternal health services during their most recent pregnancy. Based on these reports, 12 public health facilities specifically health centers were included in the facility survey. Each woman was linked to the readiness score of the reported facility. This reflects the predominantly rural PHCU-based service delivery context where a catchment health center serves as the principal point of contact and referral hub for maternal health services.

A multi-stage sampling technique was employed to select study participants. In the first stage, four rural districts were randomly selected from the total of eleven rural districts in East Shoa Zone. In the second stage, 34 kebeles (hereafter referred to as clusters) were selected using simple random sampling. Kebele is the lowest administrative unit in Ethiopia. It represents a defined geographic community served by a local health post. In this study, kebeles were used as clusters in the multi-stage sampling procedure. It also reflects the local service and social context in which women reside and receive maternal health services.

In the third stage, within each selected kebele, a complete listing of eligible women who gave birth in the 12 months preceding the survey date was conducted in each selected kebele. The listing was prepared using family folders and registration logbooks maintained at health posts. In the Ethiopian Health Extension Program, family folders are household-level records maintained by Health Extension Workers at health posts. These records contain basic information on household members, including demographic characteristics, reproductive history, maternal and child health status, and service utilization. Together with health post registration logbooks, they were used to identify eligible women and construct the sampling frame. This process identified a total of 2,076 eligible women across 34 kebeles, which constituted the sampling frame. The number of eligible women varied across kebeles, ranging from 32 to 124, with an average of approximately 60 women per kebele. This variation reflects differences in population size across clusters and was accounted for through proportional allocation during sampling.

The required sample size (n = 901) was proportionally allocated to each kebele based on the number of eligible women identified. Within each kebele, a systematic random sampling technique was applied. The sampling interval was determined by dividing the total number of eligible women by the allocated sample size, yielding an interval of approximately k ≈ 2.3. For practical implementation, a fixed interval of k = 2 was used, and every second eligible woman was selected following a randomly chosen starting point. This procedure resulted in the selection of 1,038 women across all kebeles. Of these, 997 women completed the interviews, yielding a response rate of 96.1%.

Data collection and procedures

At the household level, data were collected through face-to-face interviews using structured questionnaires administered during home visits. Health facility data were obtained through direct observation using standardized checklists. The health facility assessment checklist was adapted from the World Health Organization’s Service Availability and Readiness Assessment (SARA) tool. It covered six key domains: general amenities, basic equipment, standard precautions for infection prevention, diagnostic capacity, essential medicines and commodities, and staff and guidelines.

Data collection was conducted electronically using the KoboToolbox platform to enhance accuracy and efficiency. The questionnaires and checklists were developed based on a comprehensive review of relevant literature, national maternal and child health (MCH) guidelines, World Health Organization (WHO) recommendations, and the Ethiopian Demographic and Health Survey (EDHS) [19,20]. These tools were translated into the local languages, Afan Oromo and Amharic, and pre-tested in a similar setting to ensure clarity and contextual relevance. Necessary adaptations were made based on the pre-test findings.

Twenty trained data collectors with prior experience in survey implementation and four supervisors with a Master’s degree in Public Health (MPH) were recruited. All field personnel received a three-day intensive training on the study objectives, ethical considerations, and data collection procedures.

Measurement

The maternity continuum of care (CoC) was defined as complete when a woman received all essential components of maternal health services, including a minimum of four antenatal care (ANC) visits, skilled birth attendance by a trained health professional such as a physician, health officer, nurse, or midwife, and postnatal care (PNC) provided within the first 48 hours following delivery [21]. The antenatal care component of the continuum of care was operationalized using a threshold of four or more ANC visits (ANC4+). Although the World Health Organization’s 2016 guideline recommends a minimum of eight ANC contacts, ANC4 + was retained in this study to ensure consistency with Ethiopia’s routine monitoring system, including DHIS2 and national RMNCAH indicators. It also facilitated comparison with previous studies conducted in similar settings.However,the ANC4 + category includes all visits beyond four, and therefore also captures women who attended eight or more ANC contacts. For analytical purposes, completion of the continuum of care was treated as a binary variable and coded as “1” if all three components were fulfilled and “0” otherwise.

Knowledge of pregnancy danger signs was assessed by asking participants to name known danger signs. Women who identified at least two of the four key danger signs (vaginal bleeding, severe headache, blurred vision, and swelling of the face or legs) were categorized as knowledgeable. Those who identified fewer than two were considered not knowledgeable [22].

Adequacy of ANC contents was assessed based on women’s self-reported receipt of ten essential components recommended by the World Health Organization (WHO) [23], as collected through a household survey. The components included measurement of body weight and blood pressure, urine and blood testing, administration of tetanus toxoid vaccination and deworming medication, provision of iron and folic acid supplements, and counseling on birth preparedness and recognition of key obstetric danger signs. The contents was classified as adequate if a woman reported receiving at least eight out of the ten specified ANC service components at least once during her most recent pregnancy. Women who received fewer than eight components were classified as having inadequate ANC contents [24,25].

Explanatory variables were selected based on established conceptual frameworks and prior empirical evidence on determinants of maternal continuum of care utilization. The variables were classified as individual, household, and contextual-level factors. Individual-level variables included maternal age, educational status, parity, pregnancy intention, timing of antenatal care initiation, adequacy of antenatal care content, knowledge of pregnancy danger signs, and travel time to the nearest health facility. Household wealth status was measured at the household level but assigned to individual women and analyzed as a Level 1 explanatory variable. The contextual-level variable was health facility readiness which was linked to each woman’s reported primary source of maternal health care and represented the service-delivery environment. Variables were structured across two analytical components within a multilevel framework. Level-1 variables captured both individual and household characteristics, while Level-2 components captured clustering at the kebele level and broader contextual influences.

The multilevel structure of the analysis was defined by clustering of women within kebeles, with random intercepts specified at the kebele level to account for between-kebele variation and unobserved contextual heterogeneity. In addition, health facility readiness was included as a contextual service-delivery factor linked to each woman’s reported source of maternal health care, reflecting differences in the service environment rather than a kebele-level administrative characteristic.

Facility readiness was measured using WHO SARA tools. Domain scores were calculated as the mean proportion of tracer items available within each domain. The domains included General amenities, Basic equipment, Standard precautions for infection prevention, Diagnostic capacity, medicines and commodities, and Staff and guidelines. The overall readiness index was then computed as the average of these domain scores, in line with WHO recommendations. Then readiness scores were linked to women based on the primary facility where they reported receiving most maternal services, reflecting the PHCU-based provision of maternal care in this predominantly rural setting.

Household wealth index was constructed using Principal Component Analysis (PCA) based on ownership of selected household assets and housing characteristics. Variables included household utilities, productive and non-productive assets, and housing conditions (e.g., source of drinking water, sanitation facilities, wall and flooring materials). The first principal component, which explained the largest proportion of variance among the asset variables, was retained to generate a composite household wealth score. The resulting scores were ranked and categorized into three groups: lower, medium, and higher wealth status. This tertile classification was used to enhance interpretability and to ensure adequate sample size within each category for multilevel analysis.

Outcome variable

The outcome was completion of the continuum of maternal health care (CoC), defined as receipt of all three services: (i) ≥4 antenatal care visits (ANC4+), (ii) skilled birth attendance (SBA), and (iii) postnatal care (PNC) within 48 hours after delivery. To facilitate analysis, we also constructed stage-specific indicators and applied binary coding as follows:

Continued care at the pregnancy level. This measured whether a woman attained ANC4 + . It was coded 1 if she received at least four antenatal visits and 0 otherwise (n = 426). Continued care at the delivery level. This captured whether a woman attained both ANC4+ and SBA. It was coded 1 if she had ≥ 4 ANC visits and delivered with a skilled attendant, and 0 if she had ≥ 4 ANC visits but did not receive SBA (n = 315). Complete CoC at the postpartum level. This represented full completion of the continuum—ANC4 + , SBA, and PNC within 48 hours postpartum. It was coded 1 if all three components were met and 0 otherwise (n = 217).

Data processing and analysis

Data were processed and analyzed using STATA version 17. Descriptive analysis was performed to summarize the utilization patterns of key maternal health services across the continuum of care. Retention and drop-out rates between successive service components were calculated to describe women’s progression along the continuum of care (ANC, skilled birth attendance, and postnatal care) and to identify stage-specific gaps in service utilization. Sampling weights were calculated as the inverse of the probability of selection at each stage: woreda, kebele, and household. Kebele selection probabilities were determined using probability proportional to size. Weights were normalized to preserve the effective sample size and applied to descriptive analyses using Stata’s survey (svy) commands which account for clustering at the kebele level.

Although the maternal continuum of care can be conceptualized as a sequential pathway across pregnancy, childbirth, and the postpartum period, the primary analytical outcome in this study was full completion of the continuum of care. Therefore, binary multilevel logistic regression was used to identify explanatory variables associated with complete CoC. Stage-specific descriptive indicators and regression models were also examined as secondary analyses to support interpretation of continuation from ANC4+ to skilled birth attendance and early postnatal care.

To account for the hierarchical structure of the data where women nested within kebeles, a two-level random-intercept logistic regression model was fitted. Women were considered Level 1 units, while kebeles were considered Level 2 units. Individual-level variables included maternal age, education, parity, pregnancy intention, timing of ANC initiation, adequate ANC contents, knowledge of danger signs, and travel time to the nearest health facility. Household-level characteristics, including household wealth status were measured at the household level but assigned to individual women and analyzed as Level 1 explanatory variables.

This approach was used because the woman was the primary unit of analysis, and households were not modeled as a separate random-effect level. The random intercept at the kebele level was included to account for intra-kebele correlation and unobserved contextual heterogeneity. This method adjusts for intra-cluster correlation and enables partitioning of variance into within- and between-cluster components. [26]. Three sequential two-level random-intercept logistic regression models were fitted to examine factors associated with completion of the continuum of care.

Model 0 (Null model) included no predictors and was used to quantify between-community variance:

Model 1 (Individual level) added individual-level explanatory variables I_ij.

Model 2 (Individual + contextual factor) further included contextual service-delivery factor, i.e., health facility readiness linked to each woman’s reported source of care. C_j.

Where, P_ij is the probability of the outcome of interest for woman i in the community j; β₀is the fixed intercept (overall mean on the log-odds scale); β₁ and β₂ are fixed-effect coefficients for the individual and community-level predictors respectively; I and C refer to individual and community-level independent variables respectively, and u_0j is the community-specific random intercept capturing unobserved between-community heterogeneity. Fixed effects were expressed as odds ratios (ORs) with 95% confidence intervals (CIs) and p-values. Random effects were assessed using the intra-class correlation coefficient (ICC) [27]. The ICC was calculated as:

In this expression, σ1² denotes the between-community (level-2) variance, π represents the mathematical constant pi ≈ 3.14159 and π²/3 denotes the assumed individual-level (level-1) residual variance. In logistic multilevel models with binary outcomes, the level-1 residual variance is not estimated directly; instead, it is assumed to follow a logistic distribution with fixed variance π²/3 ≈ 3.29. [26].

Ethical considerations

Ethical approval for this study was obtained from the Institutional Review Board of the College of Health Sciences, Addis Ababa University (Protocol No. 059/24/SPH, dated 01 November 2024). Written informed consent was secured from all participants prior to their participation in the study. To ensure privacy and confidentiality, personal identifiers were intentionally omitted from the data collection form.

Socio-demographic characteristics of participants

A total of 1,038 women were invited to participate of whom 997 completed the interviews, resulting in a response rate of 96.1%.The majority of the respondents 651 (65.3%) were between 25 and 34 years old, with a mean age of 26.58 years (SD ± 4.319). Most participants were married 984 (98.7%) and predominantly identified as Oromo in ethnicity 863 (86.6%).

Regarding educational attainment, nearly half of the women 450 (45.14%) had no formal education. In terms of employment, a substantial proportion of their husbands 808 (81.04%) were engaged in farming. With respect to household wealth distribution, 420 women (42.13%) belonged to the lower wealth quintile. Table 1

Download:

• PNG
  larger image
• TIFF
  original image

Table 1. Socio-demographic characteristics of women who gave births within the past 12 months in East Shoa Zone, Eastern Ethiopia, 2024 (n = 997).

https://doi.org/10.1371/journal.pone.0356392.t001

Obstetrics and reproductive health characteristics

Approximately three-fifths of the women demonstrated adequate knowledge of obstetric and reproductive health services. Approximately half of the participants, 507 (50.9%), were multiparous.. A substantial proportion of the pregnancies, 728 (73.02%), were reported as both planned and intended. Furthermore, 518 women (51.96%) had a birth-to-pregnancy interval of two years or less. Obstetric complications were reported by 115 women (11.53%), while 138 (13.84%) had experienced adverse fetal outcomes in previous pregnancies. Table 2

Download:

• PNG
  larger image
• TIFF
  original image

Table 2. Obstetric and reproductive health characteristics of women who delivered within the past 12 months in East Shoa Zone, Eastern Ethiopia, 2024.

https://doi.org/10.1371/journal.pone.0356392.t002

Descriptive presentation of maternal health care services utilization

Among all participants, 736 women (73.82%) received at least one antenatal care (ANC) visit during their most recent pregnancy, while 426 (45.3%) attained minimum of four or more ANC contacts. Of those who initiated ANC, only 165 (22.4%) began care within the first trimester. Furthermore, among women who received at least one ANC visit, 172 (23.4%) did not receive adequate ANC service components recommended by the World Health Organization.

With regard to delivery services, only half of the most recent 515(51.65%) births were attended by skilled birth attendants. The predominant reasons for home delivery included the sudden onset of Labor 208(43.70%), lack of transportation 159(33.4%), and poor quality of service 68(14.3%). Additionally, 115 women (11.53%) experienced complications during or immediately after childbirth. The most commonly reported complications were severe vaginal bleeding 40 (34.8%), abdominal pain 39 (33.9%), severe headache 11 (9.6%), fever 11 (9.6%) and blurred vision 5 (4.4%). Regarding Postnatal care only half (50.65%) of the women reported receiving at least one postnatal care (PNC) service. Nevertheless, only 3.13% of women attended postnatal care two or more times. Table 3

Download:

• PNG
  larger image
• TIFF
  original image

Table 3. Distribution of study participants by their maternal health care utilization in East Shoa Zone, Eastern Ethiopia, 2024.

https://doi.org/10.1371/journal.pone.0356392.t003

Completion of the maternal health care continuum

After applying sampling weights to account for the multistage design and clustering at the kebele level, 45.3% (95% CI: 42.1, 48.5) of women received ≥4 antenatal care visits, 33.0% (95% CI: 30.1, 36.0) received both ANC ≥ 4 and skilled birth attendance, and 23.2% (95% CI: 20.6, 25.9) completed the full continuum of maternal health care (ANC ≥ 4, SBA, and PNC within 48 hours). In absolute terms, 426 women received four or more antenatal care (ANC) visits. Among these, 315 subsequently delivered in a health facility with skilled birth attendance. However, only 217 women completed the full continuum of maternal health care, defined as receiving four or more ANC visits, skilled birth attendance, and postnatal care within 48 hours after childbirth.

The completion of the maternity continuum of care (CoC) varied significantly across age and parity categories. A chi-square test showed statistically significant differences in CoC completion across age groups (χ² = 14.06, p = 0.001). Women aged 25–34 years had the highest completion rate (25.2%), followed by those aged 15–24 years (16.5%), while the lowest rate was observed among women aged 35–49 years (10.8%). Similarly, CoC completion differed significantly across parity categories (χ² = 6.41, p = 0.04). Women with primiparity had a completion rate of 23.3%, compared with 22.7% among multiparous women, and 12.5% among grand multiparous women. Fig 1.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. Proportion of women who completed the maternal continuum of care among those who gave birth in the 12 months preceding the survey in East Shoa Zone, Eastern Ethiopia, 2024.

https://doi.org/10.1371/journal.pone.0356392.g001

ANC = Antenatal Care (minimum four visits); SBA = Skilled Birth Attendance; PNC = Postnatal Care within 48 hours of delivery.

Multilevel analysis

In the multilevel model analysis, we initially assessed whether the data warranted the inclusion of random effects at the kebele (cluster) level. The null model demonstrated substantial variation in the odds of completing the continuum of care across communities, supporting the application of a multilevel modeling approach. The observed mean cluster size was mˉ = 29 and the ICC was ρ = 0.435, implying DEFF = 1+ (mˉ − 1) ρ = 1+ (29 − 1)×0.435 = 13.18. The intraclass correlation coefficient (ICC) revealed that 43.5% of the total variance in the completion of the continuum of care was attributable to unmeasured differences between clusters. After adjusting for all study explanatory variables, the between-cluster variance decreased to 27.9% in Model 1 and further to 25.8% in Model 2. Nonetheless, the residual variance at the cluster level remained substantial. This suggests that the continued presence of unmeasured kebele-level influences on the completion of continuum of care services despite accounting for level-1 and level-2 explanatory variables. Table 4

Download:

• PNG
  larger image
• TIFF
  original image

Table 4. Random effects parameters for the three multilevel analysis models.

https://doi.org/10.1371/journal.pone.0356392.t004

Determinants of completion of the continuum of maternal health care

The adjusted multilevel model identified several significant determinants of completing the continuum of maternal health care. At the individual and household level (level-one), maternal education, timing of the first antenatal care (ANC) visit, history of poor fetal outcomes, and household wealth status were significantly associated with completion of the continuum of care. Women who attained at least secondary education were over 3 times more likely to complete the continuum of care compared to those with no formal education (AOR = 3.27; 95% CI: 1.37, 7.79).

Early initiation of ANC within the first 12 weeks of gestation was also significantly associated with completion of the continuum of care; women who booked ANC early had over five times higher odds of completing the continuum of care than those who initiated ANC later (AOR = 5.16; 95% CI: 3.01,8.83). Similarly, women with a history of poor fetal outcomes were 2.46 times more likely to complete the continuum of care (AOR = 2.46; 95% CI: 1.33, 4.51). Furthermore, those women in the higher wealth quintile had 2.57 times greater odds of utilizing all recommended maternal health services compared to women in the lower wealth quintile (AOR = 2.57; 95% CI: 1.41,4.67).

Proximity to a health facility was significantly associated with completion of the maternal continuum of care. Women living within one hour’s travel time to a health facility had higher odds of completing the continuum of care than those residing farther away (AOR = 1.87; 95% CI: 1.01, 3.17). In addition, women who received care from health facilities with higher readiness scores were substantially more likely to complete the continuum of care compared with those served by facilities with lower readiness levels (AOR = 5.19; 95% CI: 1.48,18.16).

Stage-specific regression models were also fitted to examine continuation across the maternal care pathway. Early ANC initiation was associated with ANC4 + , ANC4+ with skilled birth attendance, and complete continuum-of-care utilization, with the strongest association observed for complete CoC (AOR = 5.16; 95% CI: 3.01,8.83). Similar cross-stage associations were observed for history of poor fetal outcome, shorter distance to a health facility, and high facility readiness. Women’s autonomy and planned pregnancy were more strongly associated with later-stage outcomes. Women with autonomy had higher odds of ANC4+ with skilled birth attendance (AOR = 1.72; 95% CI: 1.15, 2.56) and complete CoC (AOR = 3.12; 95% CI: 1.89, 5.14). Women with planned pregnancies also had higher odds of ANC4+ with skilled birth attendance (AOR = 2.16; 95% CI: 1.40, 3.32) and complete CoC (AOR = 4.18; 95% CI: 2.17, 8.01). Table 5

Download:

• PNG
  larger image
• TIFF
  original image

Table 5. Primary and stage-specific multilevel logistic regression models for maternal continuum-of-care utilization among women who gave birth in the 12 months preceding the survey in East Shewa Zone, Ethiopia, 2024.

https://doi.org/10.1371/journal.pone.0356392.t005

Study strengths and limitations

This study has several strengths. It used community-level data from rural women who gave birth within the 12 months preceding the survey. This reduced the potential for long recall periods compared with surveys using five-year birth histories. The study also linked household-level data with contextual factors, particularly health facility readiness, to comprehensively examine determinants of completion of the maternal continuum of care. In addition, the use of multilevel analysis accounted for clustering of women within kebeles and allowed assessment of between-community variation.

This study also has limitations. The use of self-reported maternal health service utilization may have introduced recall bias and social desirability bias. To reduce recall bias, the study included women who gave birth within the 12 months preceding the survey. The adequacy of ANC contents was assessed using selected service components, with all components weighted equally for analytical simplicity although their clinical importance may differ. Therefore, this measure should be interpreted as an indicator of ANC content adequacy rather than the full multidimensional quality of ANC care. The cross-sectional design also limits causal and temporal interpretation of the associations observed.

The empirical design effect of 13.18 exceeded the planning assumption of 2.0, indicating stronger within-community correlation than expected and a smaller effective sample size. Although this may reduce statistical power, the multilevel models accounted for the observed clustering. Future studies in similar settings should consider higher ICC assumptions or sensitivity analyses during sample-size planning. Linking women to a single facility based on their self-reported primary source of care may have introduced misclassification, particularly if services across pregnancy, delivery, and postnatal care were obtained from different facilities. However, the study was conducted in a rural PHCU-based context where maternal care is commonly coordinated through a dominant catchment facility. The retrospective design also did not allow verification of facility use at each contact, which may have led to non-differential misclassification of facility readiness.

The study used four or more antenatal care visits as the ANC component of continuum-of-care completion. Although ANC4 + was retained as an operational threshold because of its use in Ethiopia’s routine monitoring system, comparability with previous studies, and the potential difficulty of accurately recalling the exact number of contacts beyond four, it does not represent the current standard for optimal antenatal care. The WHO recommends a minimum of eight ANC contacts. The use of ANC4 + may therefore have overestimated continuum-of-care completion and underestimated the service gap that would have been observed under an ANC8 + definition. Accordingly, the reported prevalence should be interpreted as completion according to the study’s operational and programmatic definition rather than as full adherence to the WHO-recommended ANC8 + schedule. Finally, the binary classification of complete versus incomplete care may obscure differences between women who partially completed the continuum and those who dropped out at earlier stages. To partly address this limitation, stage-wise descriptive indicators and regression models were included to support interpretation of continuation across ANC4 + , skilled birth attendance, and early postnatal care. Future studies may consider ordered, multinomial, or other models to examine dropout patterns more explicitly.