Effectiveness and cost-effectiveness of health belief model-based companion education on maternal childbirth satisfaction and outcomes: cluster randomized controlled trial Authors: Gedamu Abera Zegeye, Muluemebet Abera Wordofa, Afework Mulugeta Published: 2026 ABSTRACT AbstractBackgroundExpanding free maternal health services in Ethiopia advances gender justice. However, structural inequities, low satisfaction, and limited use of labour companionship remain. Lower self-efficacy in companions and views of labour wards as female-only spaces are major barriers. This study measured the effectiveness and cost-effectiveness of a structured Health Belief Model (HBM)-based education session for chosen labour companions in public facilities.MethodsThis parallel-group cluster-randomized trial was conducted in 12 public health facilities in Addis Ababa from January 15 to June 5, 2025 implementation phase. Facilities were randomized 1:1, assigning six clusters (222 primigravidae dyads) to each arm. Primigravidae were screened at 12–14 weeks of gestation. The intervention group completed five HBM-based sessions for labour companions from 26 weeks to admission for labour and delivery. The control group received only standard antenatal care. The primary outcome was maternal satisfaction, measured by the 25-item Pregnancy and Childbirth Questionnaire (PCQ). Secondary outcomes included delivery mode, labour duration, and newborn complications. We used Generalized Estimating Equations (GEE) for estimation and validated the results using a 9,999-replication Wild Cluster Bootstrap (WCB). Economic assessment took a health system view, using micro-costing to calculate the Incremental Cost-Effectiveness Ratio (ICER) per newborn complication avoided.ResultsThe intervention had a highly significant positive impact on enhancing maternal satisfaction (β = 21.14, WCB 95% CI: 14.14 to 28.50, p < 0.001) compared to the control group. A notable reduction in prolonged labour > 8 h (aOR = 0.82 =, 95% CIWCB: [0.74, 0.91], PWCB < 0.001) was accompanied by a significant drop in newborn complications (4.5% vs. 11.26%; aOR = 0.93, 95% CIWCB: [0.90, 0.97], PWCB = 0.0041). Descriptively, higher spontaneous vaginal delivery rates were observed in the intervention group (77.0% vs. 66.2%), alongside reduced Cesarean pathways (14.9% vs. 23.0%), though these modal variations did not achieve formal statistical significance (p > 0.05). Economic modeling revealed that the intervention was highly cost-effective, demonstrating a > 90% probability of cost-effectiveness at Ethiopia’s willingness-to-pay benchmark, driven by the avoidance of newborn complications and neonatal intensive care costs.ConclusionsPreparing labour companions through a socio-culturally adaptive, theory-based approach substantially improves subjective maternal satisfaction and optimizes key physiological parameters, including labour duration and immediate neonatal transition outcomes. While it does not significantly alter the mechanical mode of birth or maternal complications, it provides an exceptionally cost-effective strategy that optimizes low-resource health systems, offering high clinical value without escalating institutional overhead costs.Trial registrationIt was registered with Pan African Clinical Trials Registry (PACTR) with Trial ID: PACTR202503666794479 on March 17, 2025, retrospectively registered due to a lack of awareness. URL: https://pactr.samrc.ac.za/TrialDisplay.aspx?TrialID=33509Supplementary InformationThe online version contains supplementary material available at 10.1186/s12889-026-28117-2.Keywords: Companion support, Maternal Satisfaction, Respectful Maternity Care, Birth Outcomes, Cost-effectiveness, Ethiopia Introduction Even though Ethiopia promotes free maternal health, the country still has one of the highest maternal mortality rates in the world. This is mostly due primarily to persistent structural inequalities and underutilization of evidence-based practices like labour companionship [1, 2]. National studies show that only a small number of women in Ethiopia have a labour companion during childbirth in health facilities [3]. Even though national guidelines recommend labour companionship, it is not used consistently. This is due to different factors such as barriers at the health facility, cultural beliefs, and organizational factors [4]. Many individuals still see labour wards as spaces only for women, and companions often lack confidence in their role. Maternal satisfaction with maternity care is low, with only 55% to 67% of women reporting satisfaction, which is below the national goal of 85% or higher [5]. Studies in Ethiopia revealed that labour companionship positively influences maternal satisfaction during intrapartum care [6]. Continuous support from a chosen labour companion reduces childbirth fear, alleviate labour pain, and substantially increases the likelihood of a positive and satisfactory childbirth experience compared to giving birth alone [7]. In Ethiopia, only approximately 11% to 20% of women are able to choose a labour companion, resulting in most women giving birth without consistent support [8]. Health facilities rarely promote labour companionship, healthcare providers and families often lack awareness of its benefits [9]. The Health Belief Model (HBM) provides a valuable framework for studying labour companionship in Ethiopia, as it explains the psychological and cultural factors that affect the presence of companions in labour wards [10, 11]. Findings shown that application of the HBM can enhance the effectiveness of labour companions and improve respectful maternal care [12]. The present study applies the HBM to investigate the low prevalence of labour companionship in public health facilities in Addis Ababa. The HBM is particularly relevant because it addresses modifiable psychological factors, such as perceived barriers and self-confidence, which are closely associated with Ethiopian cultural norms [13]. In Ethiopia it is routine practice to keep labour companions outside of labour ward due to concerns about privacy or interference. Nevertheless HBM-based approaches can help change these views and highlight the benefits of labour companion support during childbirth [4]. Labour companions’ optimal practice in Ethiopia can lower the risk of poor care, reduce fear during childbirth, and offer better emotional support. HBM helps to design specific solutions for the challenges of labour companions face, turning them from passive observers into confident, active supporters. Numerous labour companions may not realize their support can help to prevent complications like prolonged labour, reduce labour pain, or newborn complications. The HBM framework helps companions see how serious birth outcomes can be and encourages them to take an active role in maternal health care. In busy public health facility labour wards in Addis Ababa, labour companions often feel like they are excluded and unequipped to help during labour and childbirth [14]. HBM-based brochures and orientation sessions delineate clear roles and responsibility, showing labour companions when and how to help. This bridges the gap between wanting to help, knowing what to do and women’s desire to have labour companion during childbirth. Numerous labour companions in Ethiopia lack confidence in supporting women in labour and childbirth during intrapartum period. By offering education for both the mother and companion, this study aims to build the companion’s self-confidence and ability to help during labour. When these beliefs are strengthened, companions are more likely to provide active and continuous support. This can lead to higher maternal satisfaction, shorter labour, and less demand on healthcare workers for non-clinical tasks. This study looks at how HBM-based education for chosen companions affects delivery outcomes, maternal satisfaction, and the cost of the strategy. Companions, who understand their role and the risks and benefits of childbirth may be more likely to give the right support, reduce barriers, ease anxiety, and build self-confidence. Methods Settings and study period A parallel-group, cluster-randomized trial was employed in 12 public health facilities in Addis Ababa, Ethiopia. Cluster randomization was selected at the facility level to reduce the risk of ‘spillover effects’ or ‘informational contamination,’ which often occurs when individuals within the same clinical environment are assigned to different intervention arms. The reporting of this trial followed the CONSORT recommendations specific to cluster-randomized controlled designs. This study adhered to the CONSORT 2025 checklist, which identifies the specific page numbers where each reporting item is addressed. Items such as randomization, blinding, and statistical methods were addressed in the manuscript, provided in supplement 1(See supplement File1). Furthermore, this supplement complements the trial protocol and statistical analysis plan located in supplement2 (see supplement File2). The research took place in Addis Ababa, Ethiopia, with implementation phase from January 15 to June 5,2025. Every healthcare facility that was chosen for this study was assigned as a cluster. Trial registration and methodological disclosure This trial was registered with the Pan African Clinical Trials Registry (PACTR) under the unique identifier PACTR202503666794479 (URL: https://pactr.samrc.ac.za/TrialDisplay.aspx? TrialID=33509). While the registration was completed retrospectively on March 17, 2025, due to initial lack of awareness regarding international prospective registration timelines. To eliminate any risk of post-hoc selection bias, data mining, or selective outcome reporting, the comprehensive trial protocol, primary hypothecated outcomes, and the full statistical analysis plan (SAP) were finalized, date-stamped, and administratively locked prior to any data extraction, code execution, or unblinding by the evaluation team. Inferential estimates for primary maternal satisfaction metrics and secondary clinical birth outcomes were processed via an Intention-to-Treat (ITT) framework using Generalized Estimating Equations (GEE) and validated through a 9,999-replication Wild Cluster Bootstrap to guarantee that statistical parameters were not altered by retrospective administrative tracking. Study timeline statement The study was conducted over a 21-month period, spanning from December 2023 to August 2025, and was executed sequentially across four distinct operational phases: Phase I: Formative Research (December 23, 2023 – October 2024): This initial phase comprised mixed-methods formative research: qualitative and quantitative studies, the adaptation and validation of research instruments, comprehensive facility assessments, and the development of Information, Education, and Communication (IEC) materials. Phase I: Formative Research (December 23, 2023 – October 2024): This initial phase comprised mixed-methods formative research: qualitative and quantitative studies, the adaptation and validation of research instruments, comprehensive facility assessments, and the development of Information, Education, and Communication (IEC) materials. Baseline Assessment (November 20 – December 27, 2024): Following the formative phase, a baseline assessment was conducted over approximately five weeks to establish pre-intervention benchmarks. Baseline Assessment (November 20 – December 27, 2024): Following the formative phase, a baseline assessment was conducted over approximately five weeks to establish pre-intervention benchmarks. Phase III: Implementation (January 15 – June 5, 2025): The active intervention phase took place over nearly five months. This involved the execution of a 5-session Health Belief Model (HBM)-based educational intervention targeted at both pregnant women and their chosen companions.Phase IV: Endline Assessment (July – August 2025): The study concluded with a final endline evaluation conducted over a two-month period to assess post-intervention outcomes (Fig. 1). Phase III: Implementation (January 15 – June 5, 2025): The active intervention phase took place over nearly five months. This involved the execution of a 5-session Health Belief Model (HBM)-based educational intervention targeted at both pregnant women and their chosen companions.Phase IV: Endline Assessment (July – August 2025): The study concluded with a final endline evaluation conducted over a two-month period to assess post-intervention outcomes (Fig. 1). Sequential schematic workflow and timeline of the study phases, spanning from formative research (December 2023) through implementation and endline assessment (August 2025) Data quality assurance Sample size calculation For this parallel-arm cluster-randomized controlled trial, the required sample size was calculated using the methodology of Hooper and Bourke [15] to account for a two-stage cluster design. The Intracluster Correlation Coefficient (ICC) was utilized to measure the degree of baseline correlation among individual outcomes within the same health facility cluster. To detect a conservative 20% absolute difference in maternal childbirth satisfaction between the intervention and control arms assuming a baseline satisfaction prevalence of 40% [16] under routine care the sample size calculation incorporated an estimated ICC (ρ) of 0.05, a cluster autocorrelation coefficient of 0.8014, a two-sided alpha level (α )of 0.05, a statistical power (1-β) of 80%, a 1:1 allocation ratio, and an average cluster size (m) of 37 dyads. This yielded a pre-attrition target enrollment of 444 primigravidae-companion dyads across 12 clusters (222 dyads per arm) to absorb an anticipated 10% loss-to-follow-up rate. As documented in the trial flow diagram, a total of 434 dyads completed the full intrapartum protocol (n = 216 in the intervention group and n = 218 in the control group), resulting in a true attrition rate of 2.25% (10 dropouts out of 444). In compliance with the Intention-to-Treat (ITT) principle, all 444 initially randomized participants were retained in the primary inferential models, with missing endpoints elegantly handled using maximum likelihood estimations within the GEE architecture under the Missing-at-Random (MAR) assumption. Population and eligibility Eligible primigravidae were initially screened and provisionally enrolled during early antenatal care (ANC) contacts at 12–14 weeks of gestation to establish baseline tracking metrics. The baseline study was conducted after verifying that eligible pregnant women and their potential companions of choice fulfilled the following criteria: primigravidae status, low-risk single pregnancy, an expressed desire to receive companion support, and a planned spontaneous vaginal delivery. Following baseline allocation, active dyadic educational sessions utilizing the Health Belief Model framework formally commenced at 26 weeks of gestation, continuing sequentially across scheduled ANC visits at weeks 30, 34, and 36, as well as at admission for labour and delivery. Thereafter, enrolled dyads were followed continuously through the intrapartum phase into the postpartum period. All postpartum women who completed the follow-up period were included in the endline assessment (Fig.2). The CONSORT diagram outlines the sampling strategy, distribution of study units, evaluation of eligibility for participants, and the effectiveness of role orientation sessions in Addis Ababa from November 2024 to August, 2025 Randomization, stratification, and blinding In this study, non-adjacent healthcare facilities in Addis Ababa were selected to prevent information contamination regarding the intervention. Participants were allocated into intervention and control groups through computer-generated cluster randomization among facilities with operating theaters and a cesarean section rate of at least 5%. Clusters were stratified on their mean number of primigravida women assisted monthly and geographic location. A randomization sequence ensured unbiased assignment of these clusters in a 1:1 ratio. Specifically, six clusters were allocated to either the intervention group, while the remaining assigned to control group. Health facilities not included in the trial functioned as buffer zones between the two groups to further prevent information contamination, and data collectors were concealed during the allocation process, though the arms could still be identified by participants. Based on the eligibility criteria for health facilities, health facilities allocation to intervention or control group was done using https://www.randomizer.orgonline. Due to the nature of this educational and behavioral intervention, complete double-blinding was unfeasible; participating primigravidae, their chosen companions, and the delivering ward midwives were aware of group allocations. To control for observer bias, outcome assessments for maternal satisfaction and record extractions for secondary clinical metrics were conducted by independent, trained data collectors who were masked to the cluster allocation assignments and did not participate in delivering the HBM-based education session (Fig. 2). Sampling procedures As a result, the sampling structure was the list of primigravida women who met the requirements. Using a randomization number between one and five, the sample interval was established. Based on the inclusion criteria, the chosen pregnant women were assessed; if the next woman on the list was selected to be ineligible next woman on the list was chosen. Intervention and control arm descriptions Intervention group (HBM-Based Dyadic Education) Enrolled dyads (the pregnant woman and her chosen companion) received standard antenatal care combined with a structured orientation program grounded in the Health Belief Model (HBM) framework. The program consisted of 5 distinct 20-to-30-minute sessions led by trained maternal health professionals during scheduled ANC visits (at weeks 26, 30, 34, and 36) and reinforced upon labour room admission. The session targeted cognitive shifts by addressing Perceived Susceptibility (risks of unsupported labour), Perceived Severity (prolonged labour and neonatal trauma), and Perceived Benefits (labour shortening, pain management). Sessions used demonstrations, brochures, and task-structuring to trigger Cues to Action and enhance the companion’s Self-Efficacy in providing physical support (massage, mobility assistance) and emotional support (continuous verbal reassurance) (Fig. 3 and SuppFile2). Conceptual Framework of the HBM-based Companion Education Intervention, explaining the pathway from educational input to clinical and economic outcomes Control group (Usual Care) Participants assigned to the control clusters received the existing standard public ANC and intrapartum services alone. In this arm, companions did not receive structured pre-birth role orientations. Companionship during labour was provided inconsistently according to varying facility-specific shift protocols, and support was primarily limited to passive physical presence or tangible administrative tasks. Conceptual framework of the HBM-based companion education intervention Monitoring and evaluation The acceptability and comprehensibility of the draft materials were evaluated. Supervisory visits to the study healthcare facilities and consultative meetings with the research team and investigators were utilized to supervise the implementation. Following monitoring, any necessary adjustments were made, and the contextual obstacles discovered during the intervention were addressed as needed. Outcome measures The primary outcome was maternal satisfaction. Spontaneous vaginal delivery rates, duration of labour, Apgar scores, and cesarean section rates served as secondary outcomes. Data collection procedures Data collectors assessed women for eligibility during their initial antenatal care (ANC) visit. Those meeting the inclusion criteria and expressing interest in the study provided preliminary consent to be contacted for formal enrollment. A health facility participant list was created by the data manager and arranged according to the Medical Registration Number (MRN). These lists were then sent to maternal health practitioners, who were responsible for locating and enrolling mothers at their respective health institutions. Baseline data were acquired by data collectors during enrollment. Eligible health facilities were randomized by the data manager one week after the enrollment period ended. To minimize social desirability and response bias, the satisfaction surveys were administered by independent data collectors separate from the intervention team. Surveys were conducted in private settings to ensure participant anonymity and reduce the pressure to provide favorable responses. Outcome measures and instrumentation Primary outcome Maternal childbirth satisfaction was assessed post-delivery using the 25-item Pregnancy and Childbirth Questionnaire (PCQ) [17]. The PCQ evaluates satisfaction across sub-groups of personal behavior and educational information using a 5-point Likert scale, yielding a total raw score ranging from 25 to 125. To facilitate direct clinical comparisons, raw satisfaction scores were linearly transformed to a standardized 0–100% scale using the following pre-specified mathematical formula: Secondary outcomes Secondary clinical and neonatal parameters extracted from institutional delivery records included the mechanical mode of delivery (categorized multinomial outcomes: spontaneous vaginal delivery [SVD], assisted vaginal delivery [AVD], or Cesarean section [CS]) ; total duration of labour (dichotomized as prolonged labour > 8 h versus normal progression ≤ 8 h) ; cervical dilatation status at the time of admission ; and immediate neonatal transition indexes, including 1-minute and 5-minute APGAR scores alongside direct incidences of immediate newborn complications. Economic evaluation framework A formal economic evaluation was conducted from a health system perspective across a lifetime horizon for newborn outcomes, in accordance with the Consolidated Health Economic Evaluation Reporting Standards (CHEERS 2022) guidelines [18]. Costing was derived using a micro-costing, ingredient-based approach, aggregating all resources required for implementation, including healthcare provider training, staff time allocations, administrative overhead, and the production of physical IEC leaflets and booklets. To eliminate macro-economic inflation biases, all cost values collected during the 2025 implementation phase were deflated back to a 2022 baseline using the national Gross Domestic Product (GDP) deflator, matching the Ethiopian National Health Accounts and the Essential Health Services Package (EEHSP) data framework [19]. To evaluate incremental cost-effectiveness with appropriate contextual nuance, two distinct Willingness-to-Pay (WTP) thresholds were applied to the economic evaluations. First, a macro-demand threshold was established at 700 USD, representing approximately 1x Ethiopia’s nominal GDP per capita, in alignment with traditional WHO-CHOICE ‘Best Buy’ criteria. Second, a highly rigorous empirical supply-side threshold was established at 230 USD. This low threshold is derived from the Ethiopian Essential Health Services Package (EEHSP) resource allocation ledger, reflecting the estimated marginal opportunity cost of health care production within the public infrastructure essentially capturing the threshold at which existing services are crowded out by new health interventions. The probability of the HBM-based companion education being cost-effective was simulated across both reference points to ensure policy relevance. Costs were converted from Ethiopian Birr (ETB) to United States Dollars (USD) based on the official 2022 fiscal exchange rate (1 USD = 51.75 ETB). The key primary economic outcome was the Incremental Cost-Effectiveness Ratio (ICER), defined mathematically as the incremental cost per immediate newborn complication avoided. To address structural and parameter uncertainty within the economic evaluation, a Probabilistic Sensitivity Analysis (PSA) was executed using 1,000 Monte Carlo simulations. Gamma distributions were assigned to resource costs, and Beta distributions were assigned to binary probabilities (e.g., Newborn complication rates). A deterministic one-way sensitivity analysis was concurrently conducted and visualized using a Tornado diagram to identify key drivers of the Incremental Cost-Effectiveness Ratio (ICER). Data quality assurance The operational timeline of this intervention was solely confined to the timeline-based milestones to completely rule out any possibility of post-hoc parameter inflations or biases. Baseline measurements were done between November 20 and December 27, 2024, and the intervention phase was between January 15 and June 5, 2025. The final longitudinal measurements and endline assessments were done during July - August, 2025. After the fieldwork was done, there was reconciling double data entry and electronic validation of the outliers. Main database of the study was locked and got audited officially on 25th September, 2025. There was no post-hoc adjustments made after that; all subsequent statistical evaluations and health economic sensitivity assessments were strictly in line with the pre-specified trial protocol. Data processing and statistical analysis After completing the data validation and consistency check process, statistical analysis was performed based on descriptive statistics that summarize the characteristics of the study population. The chi-squared (χ2) test was used to examine baseline categorical characteristics in the two study groups. The tests of homogeneity of variances and normality of the continuous metrics were conducted using Levene’s test and normal quantile plot, respectively, before the main inferential tests were performed. For measuring changes in maternal childbirth satisfaction and secondary clinical outcomes, Generalized Estimating Equations (GEE) approach was used, taking into account the existence of a clustering effect within the 12 health facilities [20]. Regarding the primary metric maternal childbirth satisfaction (the continuous variable measured on the five-point Likert scale of the PCQ tool) GEE analysis was conducted with a Gaussian family and an identity link function, where the raw regression coefficients (β) are shown. For the secondary outcome, which includes both continuous variables (e.g., birthweight measured in grams) and binary variables (e.g., newborn complications [Yes/No] and APGAR score [Intermediate/Normal]), GEE analysis was used with a Binomial family and a logistic regression model for binary and linear regression for continuous metric, respectively. For categorical and binary birth outcomes (specifically prolonged labour, admission phase status, labour induction, labour augmentation, maternal complications, and newborn complications), the Generalized Estimating Equations (GEE) models were executed utilizing a Binomial distribution family paired with a logit link function. To ensure intuitive clinical interpretation and align directly with the reported values, raw regression coefficients(β) were systematically exponentiated to yield adjusted Odds Ratios (aOR = Exp (β) accompanied by their corresponding 95% Wild Cluster Bootstrapped confidence intervals (CIWCB). Conversely, continuous metrics (such as standardized maternal satisfaction percentages and absolute birth weights in grams) were modeled via a Gaussian family paired with an identity link function, with effects directly reported as raw regression coefficients (β) or mean differences (MD). All models included an exchangeable structure for working correlation. Due to the small number of clusters (K = 12), Wild Cluster Bootstrap (WCB) analysis with 9,999 replications and Webb 6-point weights was run to obtain the robust p-value. To preserve the family-wise error rate across multiple secondary birth outcomes, Benjamini-Hochberg false discovery rate (FDR) adjustments were applied to all secondary hypothesis tests, with statistical significance sustained at a sharpened q-value threshold of < 0.05. In order to minimize attrition bias, the intention-to-treat (ITT) approach was adopted. Thus, the 10 participants lost to follow-up (n = 6 in the intervention group and n = 4 in the control group) were included in the primary GEE model based on all available clinical tracking data up until attrition point under assumption of missing at random(MAR). Two-tailed statistical significance was set at an alpha level of 0.05 (p < 0.05). Data processing and statistical computations were conducted using SPSS software and R software version 4.5.1. Ethical statement We would like to confirm that all methods and experiments involving human participants were done based on ethical and legal guidelines such as The Declaration of Helsinki and GCP Guidelines [21]. Following the institution and regional ethics guidelines, “The Effect of Continuous Labour Support by Companion of Choice on the Outcome of Birth, Level of Satisfaction and Cost Effectiveness Among Pregnant Women in Addis Ababa, Ethiopia: A Cluster Randomized Control Trial” study protocol has been performed. The study obtained an approval from Jimma University Institute of Health Institutional Review Board on August 11, 2022 (Ref No: JUIH/IRB/108/22). This has been followed by obtaining ethics clearance from Addis Ababa City Government Health Bureau on December 23, 2023 (Ref No: A/H/2077/227). Moreover, administrative ethical clearance for participant tracking and data collection was securely maintained under the Addis Ababa City Government Health Bureau extension (Ref No: A.A./25018), which provided valid continuous coverage through December 2, 2025, safely encompassing the study’s conclusion, final database lock, and official audit on 25th September, 2025. Informed consent and participant rights A thorough and informed consent from all the pregnant women involved and the chosen labour companion for the research was acquired before conducting the research. The involvement in this project was entirely done on a voluntary basis where all the individuals irrespective of their literacy level was clearly told and made aware about their absolute freedom not to participate and to withdraw at any time during the course of this experiment without being penalized. In this connection, all the subjects for the experiment were thoroughly informed about the purpose of the research, methods involved risks factors, and data security concerns. Where participants with poor literacy ability were concerned, ethical considerations were taken to make sure that there was no coercion at all and that all the participants were well informed about everything regarding the study. Information sheets and consent forms given to all the participants with poor literacy abilities were read out in full detail in a private environment in the language of the participant (Amharic). This was done in the presence of an unbiased literate person who did not belong to the health care facility or the researchers in any way. Collectors of data employed a systematic “teach back method” in order to make sure that all participants fully understood about the voluntariness of the study and its aim. Thumbprints of the right thumb of participants were signed by this independent individual in order to ensure the validity of the procedure and the freedom of participants’ choice. All the participants’ data collection was done under a private and isolated environment using valid tools in order to protect their rights. No identifier was recorded in the whole process of the research. Results Participants ‘characteristics A total of 444 primigravida women were enrolled (222 in the intervention arm and 222 in the control arm). Although 434 participants completed the endline assessment (216 intervention, 218 control), all 444 original participants were included in the final intention-to-treat analysis by handling missing data via GEE modeling. Baseline characteristics, including maternal age (mean 29 ± 5 years) and companion demographics, were well-balanced across both groups (p > 0.05) (Table 1). Table 1Baseline Demographic and Socioeconomic Characteristics of the Participants, Partners, and Companions between the Intervention and Control Groups (N = 444)VariablesIntervention(n = 222) n%Control(n = 222) n%p-valueAge of the WomenMean age (SD)29 ± 5.229 ± 4.60.74 20–244219.14319.37 25–297232.437232.43 30–347332.726529.30 35 and above3515.94218.94Woman’s education0.36 Absence of reading and writing167.2146.30 Reading and writing33154519.80 Elementary school6730.35826.6 Secondary school5022.525123.0 College and beyond5625.05424.30Woman’s occupation0.76 Housewife5826.365625.45 Employed10346.811050.1 Private work5926.85424.54Marital Status0.94 Unmarried2712.162511.28 Married19587.8419788.72Partner’s Education Absence of reading and writing115.094.00.61 Reading and writing4218.904018.02 Elementary school5022.54922.07 Secondary school6931.007332.88 College and beyond5022.54922.7Partner’s Occupation0.23 Employed12958.1112255.00 Private work8136.489140.99 Others(Daily laborer,125.494Companion’s age0.19 20–243314.543616.36 25–296027.726227.73 30–346629.545525.0 35–39187.73208.6 40–44167.272310.45 45 and above2913.182611.82Companions’ Education0.34 No read and Write135.4125.0 Read and write6128.25324.1 Elementary school5324.15726.1 Secondary School6328.637433.63 College and above3214.12611.36Companion’s Occupation0.82 Housewife6328.636127.72 Employed11050.110447.3 Private work3717.274520.45 Others125.0124.54Companions’ prior experience of support provision0.365 No18082.1918383.71 Yes4217.783916.38Missing data for participants lost to follow-up (n=6 intervention, n=4 control) were handled via GEE modeling under the Intention-to-Treat framework. Data are presented as frequencies and percentages (n, %) for categorical variables, and as means ± standard deviations (Mean ± SD) for continuous variables (Age). P-values are derived from Independent Samples t-tests for continuous data and Pearson’s Chi-Square (X2 ) tests (or Fisher's Exact tests, where appropriate) for categorical variables. Statistical significance is set at p < 0.05 Baseline Demographic and Socioeconomic Characteristics of the Participants, Partners, and Companions between the Intervention and Control Groups (N = 444) Age of the Women Mean age (SD) Missing data for participants lost to follow-up (n=6 intervention, n=4 control) were handled via GEE modeling under the Intention-to-Treat framework. Data are presented as frequencies and percentages (n, %) for categorical variables, and as means ± standard deviations (Mean ± SD) for continuous variables (Age). P-values are derived from Independent Samples t-tests for continuous data and Pearson’s Chi-Square (X2 ) tests (or Fisher's Exact tests, where appropriate) for categorical variables. Statistical significance is set at p < 0.05 Our process evaluation successfully establishes the behavioral mechanism of action. HBM-based education for labour companions demonstrated significantly higher average engagement levels across all tracked subscales compared to the control group, yielding an overall composite support score of 6.75 pm 0.55 out of a maximum possible 9 points, compared to just 2.30 pm 0.60 in the control arm (p < 0.001) (Supplementary Table 3). Effect of Health Belief Model (HBM)-based education for labour companion on birth outcomes The secondary outcomes evaluated via Generalized Estimating Equations (GEE) with Wild Cluster Bootstrapping (WCB) error adjustments are outlined in Table 2. Mothers in the intervention arm had a significantly lower risk of prolonged labour. In the control group, the proportion of individuals with labor lasting more than 8 h was 75.9% (169/222), which decreased to 66.4% (147/222) in the intervention group with a protective effect (aOR = 0.82, 95% CIWCB: [0.74, 0.91], PWCB < 0.001). Similarly, the initial clinical facility assessment showed significant difference with fewer mothers in the intervention group being in the latent phase (< 5 cm cervical dilatation) at admission than those in the control group (46.8% vs. 56.8%; aOR = 0.80, 95% CIWCB: [0.70, 0.91], PWCB = 0.0038).There were no statistically significant differences in mechanical delivery paths and intrapartum interventions between study arms. The rate of spontaneous vaginal delivery was 77.0% in the intervention group compared to 66.2% in the control group. Neither assisted vaginal delivery (aOR = 0.99, 95% CIWCB: [0.94, 1.04], PWCB = 0.4820) nor emergency Cesarean section (aOR = 0.97, 95% CIWCB: [0.90, 1.04], PWCB = 0.3552) were statistically significant. Similarly, rates of labor induction (aOR = 0.96, PWCB = 0.0641), labor augmentation (aOR = 0.99, PWCB = 0.7857) and absolute maternal complications (6.8% vs. 9.5%; aOR = 0.95, 95%CIWCB: [0.89,1.01], PWCB = 0.0797) safely differed within uniform parameters. Neonatal outcomes were significantly improved physiologically in the intervention arm. Intermediate 1-minute APGAR scores (6–7) were significantly less common in the intervention arm (30.2%) versus control arm (42.8%) (aOR = 0.63, 95% CIWCB: [0.60, 0.67], PWCB < 0.001) although these differences were equalized by the 5-minute APGAR score (aOR}= 1.04, PWCB = 0.4524). Overall newborn complications were significantly lower in the intervention arm compared to controls (4.50% vs. 11.26%; aOR = 0.93, 95% CIWCB: [0.90, 0.97], PWCB = 0.0041). Finally, modeling of continuous birth weights with an identity link function confirmed a highly significant positive mean adjustment in favor of the intervention cohort (MD = 169.05 g, 95% CIWCB: [97.81, 239.32], PWCB < 0.001) (Table 2). Table 2GEE Model Evaluations of Structured HBM-Based Companion Education on Childbirth and Neonatal Outcomes among participants in Addis Ababa, Ethiopia (N = 444)Outcome VariablesReporting CategoryTotal (N = 444),n (%)Control (n = 222), n (%)Intervention (n = 222), n(%)Effect Estimate(95% CIWCB)aP-valueWCBbMode of DeliverySpontaneous Vaginal (SVD)318 (71.62)147 (66.21)171 (77.03)1.00 (Reference)—Assisted Vaginal (AVD)42 (9.45)24 (10.81)18 (8.10)aOR0.99 (0.94, 1.04)0.4820Cesarean Section (CS)84 (18.92)51 (22.97)33 (14.86)aOR: 0.97 (0.90, 1.04)0.3552Duration of Labour≤ 8 h128 (28.83)53 (23.87)75 (33.78)1.00 (Reference)—> 8 h316 (71.17)169 (76.12)147 (66.21)aOR:0.82 (0.74, 0.91)p < 0.001Cervical Dilatation at Entry≥ 5 cm (Active Phase)214 (48.19)96 (43.24)118 (53.15)1.00 (Reference)—< 5 cm (Latent Phase)230 (51.80)126 (56.76)104 (46.85)aOR: 0.80 (0.70, 0.91)0.0038Labour InductionNo406 (91.44)196 (88.28)210 (94.59)1.00 (Reference)—Yes38 (8.56)26 (11.71)12 (5.40)aOR: 0.96 (0.92, 1.00)0.0641Labour AugmentationNo406 (91.44)200 (90.09)206 (92.79)1.00 (Reference)—Yes38 (8.56)22 (9.90)16 (7.20)aOR: 0.99 (0.92, 1.07)0.7857Maternal ComplicationsNo408 (91.89)201 (90.54)207 (93.24)1.00 (Reference)—Yes36 (8.10)21 (9.45)15 (6.76)aOR: 0.95 (0.89, 1.01)0.0797APGAR Score (1 min)8–10 (Normal)282 (63.51)127 (57.20)155 (69.82)1.00 (Reference)—6–7 (Intermediate)162 (36.48)95 (42.79)67 (30.18)aOR:0.63 (0.60, 0.67)p < 0.001APGAR Score (5 min)8–10 (Normal)416 (93.69)205 (92.34)211 (95.04)1.00 (Reference)—6–7 (Intermediate)28 (6.30)17 (7.66)11 (4.95)aOR:1.04 (0.92, 1.17)0.4524Newborn ComplicationsNo409 (92.11)197 (88.74)212 (95.49)1.00 (Reference)—Yes35 (7.88)25 (11.26)10 (4.50)aOR: 0.93 (0.90, 0.97)0.0041Birth Weight (grams)Mean (SD)2806 (300)2690 (126)2859 (369)MD: 169.05(97.81, 239.32)p < 0.001Abbreviations AVD Assisted Vaginal Delivery, CS Cesarean Section, GEE Generalized Estimating Equations, SVD Spontaneous Vaginal DeliveryaaOR = Adjusted Odds Ratio derived via exponentiated Generalized Estimating Equations (GEE) model parameters [Exp(β)] utilizing a logit link function for categorical binary and multinomial risk frameworks; MD = Mean Difference derived directly via linear GEE modeling using an identity link function for continuous measurements. Both mathematical model designs control strictly for intra-cluster correlation across the 12 public health facilities in Addis AbababWCB = Wild Cluster Bootstrap. To fully adjust for the bounded size of institutional facility clusters (K=12) and rule out Type I error inflation (false positives), confidence intervals (CIWCB) and respective P-valueWCB are extracted using using a 9,999-replication Wild Cluster Bootstrap (WCB) with Webb weights (K=12 clusters) to account for cluster-level variations. Missing data were handled via Generalized Estimating Equations (GEE) under an Intention-to-Treat (ITT) framework. Continuous Birth Weight was modeled utilizing a Gaussian family (identity link). Categorical and binary outcomes were evaluated via binomial GEE (Binomial family, logit link, exchangeable correlation structure) GEE Model Evaluations of Structured HBM-Based Companion Education on Childbirth and Neonatal Outcomes among participants in Addis Ababa, Ethiopia (N = 444) MD: 169.05 (97.81, 239.32) Abbreviations AVD Assisted Vaginal Delivery, CS Cesarean Section, GEE Generalized Estimating Equations, SVD Spontaneous Vaginal Delivery aaOR = Adjusted Odds Ratio derived via exponentiated Generalized Estimating Equations (GEE) model parameters [Exp(β)] utilizing a logit link function for categorical binary and multinomial risk frameworks; MD = Mean Difference derived directly via linear GEE modeling using an identity link function for continuous measurements. Both mathematical model designs control strictly for intra-cluster correlation across the 12 public health facilities in Addis Ababa bWCB = Wild Cluster Bootstrap. To fully adjust for the bounded size of institutional facility clusters (K=12) and rule out Type I error inflation (false positives), confidence intervals (CIWCB) and respective P-valueWCB are extracted using using a 9,999-replication Wild Cluster Bootstrap (WCB) with Webb weights (K=12 clusters) to account for cluster-level variations. Missing data were handled via Generalized Estimating Equations (GEE) under an Intention-to-Treat (ITT) framework. Continuous Birth Weight was modeled utilizing a Gaussian family (identity link). Categorical and binary outcomes were evaluated via binomial GEE (Binomial family, logit link, exchangeable correlation structure) Maternal childbirth satisfaction analysis Data were analyzed using Generalized Estimating Equations (GEE) to account for the correlated nature of longitudinal observations within subjects. To ensure the reliability of the findings against potential distributional assumptions or small-cluster biases, a Wild Cluster Bootstrap (WCB) with 9,999 replications was performed. To evaluate the impact of the intervention while accounting for the clustered nature of the data and repeated measures, a Generalized Estimating Equation (GEE) model was utilized, supplemented by a Wild Cluster Bootstrap (WCB) estimation to ensure robust standard errors. The primary outcome of interest was the interaction between study group and time (Treatment Effect). The baseline analysis indicated no statistically significant difference in the outcome between the intervention and control arms (Estimates (β) = -0.52, SE = 1.64, 95% Wald CI: [-3.75, 2.70], p = 0.750), confirming comparable baseline equivalence. Similarly, no significant secular trend over time was observed within the reference control group from baseline to endline (Estimates (β) = -1.19, SE = 0.69, 95% Wald CI: [-2.55, 0.17], p = 0.085; WCB 95% CI: [-2.41, 0.28], p = 0.114). However, a highly significant treatment effect was observed over time. The group-by-time interaction coefficient indicates that the intervention group achieved a multi-unit increase of 21.14 relative to the control group from baseline to endline (Estimates (β) = 21.14, SE = 3.13, 95% Wald CI: [15.01, 27.27], p < 0.001). To account for potential cluster-related biases or small-cluster constraints, the Wild Cluster Bootstrap was applied. The bootstrapped estimates strongly validated the primary model, demonstrating a highly robust and stable intervention effect (Estimates (β) = 21.14, WCB 95% CI: [14.14, 28.50], p = 0.0002)(Table 3). Table 3GEE-WCB Model Estimates for Maternal Childbirth Satisfaction among participants in Addis Ababa, EthiopiaPredictor ParameterRegression Coefficient (β)Standard Error (SE)95% Wald Confidence IntervalGEEp-value95% Wild Cluster Bootstrap CIWCBp-valueIntercept50.293.21[44.01, 56.57]< 0.001——Group (Control vs. Intervention)-0.521.64[-3.75, 2.70]0.750[-3.12, 2.15]0.722Time (Baseline vs. Endline)-1.190.69[-2.55, 0.17]0.085[-2.41, 0.28]0.114Treatment Effect (Interaction)21.143.13[15.01, 27.27]< 0.001[14.14, 28.50] < 0.001 Standard errors and Wald tests account for clustering within the 12 facilities via an exchangeable correlation matrix. Bold text highlights statistical significance at alpha = 0.05 GEE-WCB Model Estimates for Maternal Childbirth Satisfaction among participants in Addis Ababa, Ethiopia Standard errors and Wald tests account for clustering within the 12 facilities via an exchangeable correlation matrix. Bold text highlights statistical significance at alpha = 0.05 The impact of the intervention on the average level of satisfaction scores for mothers is presented in Fig. 4 depicting changes over time from the first to the last moment of the study. Initially, the scores of participants were rather similar because their 95% confidence intervals overlapped. But throughout the experiment period, the control group did not experience any change concerning average satisfaction scores, while an increase in the average satisfaction score was observed in the intervention group. The average score of the participants in the intervention group changed by 21 units more than that of the control group. This change in dynamics of two curves is clearly depicted in Fig. 4, showing the effects of the intervention (Estimates (β) = 21.14, p < 0.001). Additionally, the fact that 95% confidence intervals do not overlap shows the success of the positive intervention impact (Fig. 4). Comparison of Mean Maternal Childbirth Satisfaction Scores between Control and Intervention groups at Baseline and Endline. Error bars represent 95% confidence intervals. The significant divergence at endline represents a substantial treatment effect Analysis of predictors the effect of companion education on maternal satisfaction Baseline maternal factors associated with maternal satisfaction were found to be occupation and marital status. Specifically, housewives (MR = 0.777; 95% CI: [0.672, 0.898]; p < 0.001) and formally employed women (MR = 0.806; 95% CI: [0.684, 0.951]; p = 0.011) had significantly lower arithmetic mean satisfaction scores than mothers involved in private work. Furthermore, unmarried mothers had significantly lower mean satisfaction scores than their married counterparts (MR = 0.857; 95% CI: [0.745, 0.986]; p = 0.031). However, there was no significant difference in the mean score of mothers based on age and education level (p > 0.05). On the other hand, baseline companion factors revealed that mothers with labour companions who were not formally educated had significantly lower mean satisfaction scores by 21.3% than mothers with companions having college degrees or higher (MR = 0.787; 95% CI: [0.651, 0.95 [0.651, 0.952]; p = 0.014). Conversely, having a companion who worked in private business significantly increased the mean satisfaction score by 29.5% compared to having a housewife companion (MR = 1.295; 95% CI: [1.019, 1.644]; p = 0.034). The companion’s age and specific relationship type to the mother showed no significant independent main effects on the satisfaction scores. Several clinical and labour-related variables demonstrated strong independent associations with maternal satisfaction. Women who did not undergo labour augmentation experienced a significant increase in mean satisfaction (MR = 1.256; 95% CI: [1.090, 1.446]; p = 0.002) compared to those who did. Institutional shift times also played a significant role: deliveries occurring during the evening shift carried an 18% higher mean satisfaction score compared to night-shift deliveries (MR = 1.180; 95% CI: [1.018, 1.366]; p = 0.028). Notably, a lack of previous labour companion support was strongly associated with a reduction in mean satisfaction (MR = 0.789; 95% CI: [0.692, 0.900]; p = 0.001). Other clinical factors including mode of delivery, duration of labour, induction status, cervical dilatation at admission, and pregnancy or childbirth complications did not significantly influence maternal satisfaction within the multivariable model. Primary Interaction Effect (Group*Time = 0): A highly significant interaction effect was observed at baseline/control conditions (MR = 0.534; 95% CI: [0.372, 0.766]; p = 0.001), indicating that the baseline group mean satisfaction score was roughly 46.6% lower than the reference condition (Group X Time = 1). Significant variations in satisfaction over time were observed when factoring in maternal occupation. The arithmetic mean. satisfaction scores over time decreased significantly for housewives (MR = 0.812; 95% CI: [0.692, 0.953]; p = 0.011) and employed individuals (MR = 0.726; 95% CI: [0.619, 0.851]; p = 0.001) relative to those engaged in private work. The interaction effect for unmarried status over time was also statistically significant, showing a lower mean satisfaction score (MR = 0.809; 95% CI: [0.658, 0.995]; p = 0.045) over time compared to married individuals. There were no significant interaction effects on mean maternal satisfaction scores over time regarding companion relationship type, labour augmentation status, or companion education level (p > 0.05; Table 4). Table 4Generalized Estimating Equations predicting the effect of companion education on maternal satisfaction in Addis Ababa, EthiopiaParameterExp(B)95% Wald CI for Exp(B)P-valueIntercept34.729[22.355, 53.952] 0.001 Age Category 20–240.909[0.757, 1.092]0.309 25–290.949[0.806, 1.116]0.524 30–341.023[0.863, 1.213]0.796 35 and above1.000[Reference]Educational Status No formal education1.082[0.823, 1.422]0.574 Reading and writing0.989[0.826, 1.185]0.907 Elementary school1.075[0.925, 1.249]0.348 Secondary school0.908[0.785, 1.051]0.196 College and beyond1.000[Reference]—Occupation Housewife0.777[0.672, 0.898] 0.001 Employed0.806[0.684, 0.951] 0.011 Private work1.000[Reference]—Marital Status Unmarried0.857[0.745, 0.986] 0.031 Married1.000[Reference]—Relationship Type Mother and mother-in-law0.991[0.837, 1.174]0.918 Partner/spouse0.925[0.783, 1.093]0.359 Sisters0.850[0.721, 1.003]0.054 Friends1.000[Reference]—Age of Companions 20–241.031[0.868, 1.223]0.729 25–290.999[0.855, 1.167]0.990 30–340.944[0.821, 1.087]0.424 35–390.951[0.801, 1.128]0.561 40–441.027[0.863, 1.222]0.762 45 and above1.000[Reference]—Companion Education No formal education0.787[0.651, 0.952] 0.014 Read and write1.012[0.851, 1.203]0.897 Elementary school1.000[0.866, 1.154]0.995 Secondary School1.048[0.922, 1.190]0.477 College and above1.000[Reference]—Companion Occupation Housewife1.000[Reference]— Employed1.161[0.930, 1.450]0.188 Private work (Business)1.295[1.019, 1.644] 0.034 Mode of Delivery Spontaneous Vaginal Delivery (SVD)1.107[0.954, 1.284]0.182 Caesarean Section (CS)1.029[0.861, 1.228]0.755 Assisted SVD1.000[Reference]—Duration of Labour ≤ 8 h1.071[0.954, 1.203]0.244 >8 h1.000[Reference]—Augmentation No1.256[1.090, 1.446] 0.002 Yes1.000[Reference]—Induction of Labour No0.924[0.773, 1.106]0.389 Yes1.000[Reference]—Complications Childbirth No1.030[0.867, 1.225]0.734 Yes1.000[Reference]—Time of Birth (Shift) Day time1.043[0.933, 1.165]0.462 Evening1.180[1.018, 1.366] 0.028 Night1.000[Reference]—Newborn Complications No1.033[0.815, 1.309]0.791 Yes1.000[Reference]—Pregnancy Complications No0.907[0.725, 1.136]0.397 Yes1.000[Reference]—Cervical Dilatation at Admission < 5 cm0.964[0.876, 1.062]0.458 ≥ 5 cm1.000[Reference]—Previous Labour Companion Support No0.789[0.692, 0.900] 0.001 Yes1.000[Reference]—INTERACTION TERMSGroup*Time Group*Time = 00.534[0.372, 0.766] 0.001 Group*Time = 11.000[Reference]—Group*Time*Occupation Group*Time*Housewife0.812[0.692, 0.953] 0.011 Group*Time*Employed0.726[0.619, 0.851] 0.001 Group*Time*Private work1.000[Reference]—Group*Time*Marital Status Group*Time*Unmarried0.809[0.658, 0.995] 0.045 Group*Time*Married1.000[Reference]—Group*Time*Relationship Type Group*Time*Mother & Mother-in-law0.866[0.702, 1.069]0.181 Group*Time*Partner1.023[0.845, 1.239]0.814 Group*Time*Sister0.857[0.688, 1.068]0.170 Group*Time*Friends1.000[Reference]—Group*Time*Augmentation Group*Time*No1.141[0.888, 1.465]0.302 Group*Time*Yes1.000[Reference]—Group*Time*Companion Education Group*Time*No formal education0.903[0.653, 1.247]0.535 Group*Time*Read and write1.114[0.888, 1.399]0.351 Group*Time*Elementary school*0.629[0.270, 1.466]0.240 Group*Time*Secondary School*1.225[0.330, 1.452]0.200 Group*Time*College and above1.000[Reference]—Exp (B) = Mean Ratio (MR), Reference= Reference category Generalized Estimating Equations predicting the effect of companion education on maternal satisfaction in Addis Ababa, Ethiopia Exp (B) = Mean Ratio (MR), Reference= Reference category Cost-Effectiveness Analysis (CEA) The total cost of the HBM-based education for labour companion training is estimated at 15.11 USD (781.49 ETB) per individual. This includes: Training materials and educational aids, Personnel time for health education delivery and Administrative overhead for the intervention. The primary clinical outcome, Newborn Complications, demonstrated a statistically significant absolute reduction of 6.76% (p = 0.0041). Based on this effect size, the ICER is calculated as follows: ICER = Cost of HBM-Based Education/ Absolute Reduction in Newborn Complication = 15.11/ 0.0676 = 223.52USD. For every 223.52 USD (11,567 ETB) invested across a cohort of women, the health system successfully avoids one major neonatal complication. In the context of the Ethiopian public health framework, this investment yields significant saving by mitigating the need for high-cost neonatal intensive care (NICU) resources, specialized respiratory support, and extended facility stays. Given that these clinical gains are achieved at a fraction of the cost of treating such complications, the intervention is considered a highly cost-effective “best buy” for improving neonatal safety. Incremental cost-effectiveness and distribution The base-case analysis yielded an Incremental Cost-Effectiveness Ratio (ICER) of 223.52 USD per neonatal complication avoided. The cost-effectiveness plane reveals that the majority of incremental cost-effectiveness pairs for significant clinical outcomes are situated in the “South-East” quadrant. This indicates that the HBM-based labour companion education represents a cost-effective strategy, providing superior health effectiveness (safety) while resulting in lower or neutral incremental costs compared to standard care due to the avoidance of expensive clinical complications. In the absence of a standardized national threshold for patient-reported outcomes in Ethiopia, we utilized the Ethiopian Essential Health Service Package (EEHSP) prioritization criteria and the WHO-CHOICE ‘Best Buy’ benchmarks as a frame of reference. Given Ethiopia’s commitment to the Compassionate, Respectful, and Caring (CRC) initiative and Universal Health Coverage (UHC), an intervention achieving high-certainty safety gains at an ICER of 223.52 USD which is well below the traditional threshold of 1x GDP per capita (estimated at 700 USD) is considered highly cost-effective and fiscally sustainable within the Ethiopian public health system. The certainty of the economic value was achieved through probabilistic analysis conducted using PSA, which ran for 1,000 iterations. According to the analysis, the willingness to pay criteria determined according to the country’s health policies indicate that the strategy under discussion is cost-effective in most of the simulations considering the critical indicators among neonates. Therefore, the “certainty of evidence” is an indication that the economic value of the suggested HBM-based program is consistent despite the variability in clinical effectiveness and resource unit prices, mostly because of less demand for neonatal intensive care facilities and specific respiratory assistance. The results on the Cost-Effectiveness Acceptability Curve (CEAC) illustrate that the probability of achieving cost-effectiveness remains robust across both the empirical supply-side and macro-demand thresholds. To ensure strict methodological alignment with our clinical endpoints, the horizontal axis (x-axis) is standardized to denote Willingness-to-Pay (WTP) in USD per Newborn Complication Avoided, moving from 0 to 2,000 USD (Fig. 5). Cost-Effectiveness Acceptability Curve (CEAC) modeling the probability that HBM-based companion education is cost-effective compared to routine care, as a function of the health system’s Willingness-to-Pay (WTP) threshold per Newborn Complication Avoided Extended clinical value analysis Beyond the primary satisfaction outcome, the 15.11 USD investment yields a “bundle” of clinical improvements that suggest potential cost-offsets (money saved by avoiding expensive complications) (Table 5). Table 5Economic Implications of Clinical Improvements Attributed to HBM-Based InterventionClinical OutcomeMD/RR of Effect (Intervention vs. Control)Economic Implication (Cost-Avoidance Potential)Prolonged Labour19.6% Relative Risk ReductionDecreased facility resource intensity, reduced bed occupancy, and optimized staff allocation.Cervical Dilatation Status at Admission10.0% Absolute Reduction in Late Latent-Phase Admissions (< 5 cm)Optimizes labour ward triage efficiency, reduces unnecessary prolonged institutional bed occupancy during early labour, and enhances throughput.Neonatal Complications6.76% Absolute ReductionManagement of resources used for NICU which are expensive.Birth Weight+ 169 g Mean DifferenceReduction of both immediate and future expenses related to Low Birth Weight patients.All clinical outcomes listed demonstrated statistical significance at p < 0.05MD Mean Difference, RR Relative Risk Economic Implications of Clinical Improvements Attributed to HBM-Based Intervention All clinical outcomes listed demonstrated statistical significance at p < 0.05 MD Mean Difference, RR Relative Risk Probabilistic Sensitivity Analysis (PSA) and CEAC In PSA on the proposed HBM model-based study, probabilistic distributions are applied to estimate all clinical and cost parameters in order to account for the uncertainty of such variables. The newborn complication variable can be defined as an Important Variable (Level I) serving as a “driver” for this simulation model. Its p value (0.0041) suggests that the probabilistic distribution will be relatively narrow and favorable to the proposed intervention during 1,000 Monte Carlo simulations. In the CEAC plot, the probability of cost-effectiveness of the proposed intervention can be observed at various “willingness to pay” rates for preventing the development of complications among newborns. Given the effectiveness of the intervention in terms of absolute reduction of newborn complications by 6.76% (the control group 11.26% vs. the intervention 4.50%), the graph should illustrate high probabilities of cost-effectiveness at low WTP rates. However, incorporating non-significant variables, such as Mode of Delivery (p = 0.355), causes some disturbance to the model. Thus, there might be fluctuations in costs related to delivery influencing the economic result of the project regardless of its highly certain clinical benefits. (Table 6). Table 6Assumptions for Statistically Significant, Non-Significant and Borderline Variables in PSA/CEAC for HBM-based companion education intervention in Addis Ababa, EthiopiaVariableNatureDistribution (GEE Family)Rationale / Measurement Approach(WCB)P-valueMaternal satisfactionContinuousGaussianPrimary outcome: Standardized continuous scores(0–100%) from the 25-item Pregnancy and Childbirth Questionnaire (PCQ); modeled with an identity link function.p < 0.001APGAR 1-minuteBinaryBinomialNeonatal transition measure: Categorized into binary clinical risk levels (“Intermediate” vs. “Normal”) and modeled with a logit link function.p < 0.001Birth weight (g)ContinuousGaussianKey physical neonatal health metric: Continuous physical birth measurement in grams; checked for normality and homogeneity of variance.p < 0.001Duration of labourBinaryBinomialMeasure of clinical efficiency: Evaluated using a dichotomous threshold based on clinical criteria (“Prolonged labour > 8 hours” vs. Normal progression).p < 0.001Cervical dilatationBinaryBinomialPredictor of delivery outcomes: Analyzed as a dichotomous state at admission to track clinical status (“Latent phase < 5 cm” vs. Active phase).p = 0.0038Newborn complicationBinaryBinomialMeasure of safety and neonatal health: Simple incidence occurrence (“Yes” vs. “No”); serves as the primary driver of the cost-effectiveness model.p = 0.0041Induction of labourBinaryBinomialClinically relevant baseline intervention status; statistically non-significant.p = 0.0641Maternal complicationBinaryBinomialObserved risk trend; did not cross the threshold for formal statistical significance.p = 0.0797Mode of deliveryCategoricalMultinomialExploratory clinical pathway mapping three distinct modes (Spontaneous Vaginal, Assisted Vaginal, and Cesarean Section).p = 0.3552APGAR 5-minuteBinaryBinomialSecondary neonatal outcome: Binned identically to the 1-minute score (“Intermediate” vs. “Normal”) to track resuscitation recovery kinetics.p = 0.4524AugmentationBinaryBinomialBaseline obstetric intervention event occurrence (“Yes vs.No”)p = 0.7857WCB Wild Cluster Bootstrap Assumptions for Statistically Significant, Non-Significant and Borderline Variables in PSA/CEAC for HBM-based companion education intervention in Addis Ababa, Ethiopia Primary outcome: Standardized continuous scores (0–100%) from the 25-item Pregnancy and Childbirth Questionnaire (PCQ); modeled with an identity link function. WCB Wild Cluster Bootstrap Cost-Effectiveness Acceptability Curve (CEAC) and Tornado diagram for sensitivity analysis Cost-effectiveness Acceptability Curve (CEAC) and Tornado Diagram for Sensitivity Analysis. The Cost Effectiveness Acceptability Curve (CEAC) shows an increase in probability of cost effectiveness for the intervention, when the threshold Willingness-to-Pay (WTP) increases rapidly. At the lower boundary of 230 USD, which is the empirical supply side threshold, the intervention attains a moderate probability of cost effectiveness, affirming that the intervention works well within the stringent resource constraint. The higher the WTP value, the curve becomes almost certain, indicating strong economic efficiency as the intervention works better than the GDP per capita of 700 USD. The two threshold settings are both feasible and economically comparable, in conformity with EEHSP prioritization model. NICU costs, C section rate, and training costs significantly determine the ICER in the sensitivity analysis, while utility of complication avoided and treatment efficacy moderately affect it. The vertical line in green color at 700 USD and horizontal one in red at 230 USD set the thresholds of the model. These indicate that changes in parameters do not make significant impacts on ICER as shown in the graph; thus, the cost effectiveness of the intervention is robust (Fig. 6). Cost-Effectiveness Acceptability Curve (CEAC) and Tornado Diagram for Sensitivity Analysis of HBM-based Labour Companion Education in Addis Ababa, Ethiopia Clustered CEAC Cost Effectiveness Acceptability Curve (CEAC): CEAC is plotted using the probability of the intervention maintaining its cost-effectiveness over different Willingness–To-Pay (WTP) levels. Layered Cost-Effectiveness Acceptability Curve (CEAC): In probabilistic sensitivity analysis, from Fig. 7 below, when the threshold willingness-to-pay (WTP) equals to the ICER of the intervention (223.52 USD per neonatal complication avoided), the probability of the Health belief Model (HBM)-based structured education program being the most cost-effective strategy is above 85%. It shows that the intervention economically is robust in the face of parameter uncertainty. Significant variables are indicated by the green layer while non-significant variables are indicated by the yellow layer, with both shaded by 95% confidence intervals. The significance gap between significant and non-significant variables indicates disparity between statistically robust and explorative variables. At the minimum WTP level of 230 USD (which equals to Ethiopia’s GDP per capita) representing supply side resource availability, the probability is moderately high for significant variables. With an incremental cost-effectiveness ratio (ICER) of 223.52 USD per neonatal complication avoided, it indicates that the program is economically feasible within the specified resource limitation, hence aligning with EEHSP priorities. As the willingness to pay increases to high level (700 USD – 1× GDP per capita), the probability becomes practically certain (Fig. 7). Clustered Cost-Effectiveness Acceptability Curve for HBM-Based Education for Labour Companions in Addis Ababa, Ethiopia Parallel incremental cost-effectiveness scatterplot In the left figure, non-significant and marginal results tend to be located close to the origin, meaning there are not many benefits beyond WTP. In the right figure, there is a clustering of significant results in the upper right quadrant, suggesting that cost-effectiveness is strongly demonstrated through positive incrementalwithin reasonable costs. Each dot corresponds to a clinical variable depending on its impact (blue = positive, red = negative) (Fig. 8). Parallel Incremental Cost-Effectiveness Scatterplots from Probabilistic Sensitivity Analysis (PSA) for HBM-Based Labour Companion Education in Addis Ababa, Ethiopia Combined two way sensitive analysis The graph showing the ICER contour plots reveals how the incidence of newborn complications on the x-axis and the mode of delivery on the y-axis affect the economic efficiency of the program. It shows an increase in efficiency along the spectrum of red (> 2000 USD), orange (1500 USD), yellow (1000 USD), light green (500 USD), and dark green (< 0 USD). The graph shows an enormous drop in ICERs as the rate of newborn complications decreases and spontaneous vaginal delivery increases, proving the program’s cost-effectiveness. The high clustering of variables newborn complications, APGAR1 minute, and birth weight in the green zone proves its economic significance under Ethiopian health system willingness to pay thresholds. The clustering of significant variables into ICER zones is clear proof of the robustness of the HBM-based educational model. The graph shows that the combination of low newborn complications and spontaneous vaginal delivery pushes ICER values into very cost-effective zones (< 500 USD), while the combination of high newborn complications and cesarean deliveries pushes the ICER value into poor cost-effectiveness zones (> 2000 USD). The plot demonstrates that reductions in neonatal complications combined with spontaneous vaginal delivery shift ICER values into highly cost-effective zones. In contrast, higher complication rates and cesarean deliveries cluster in poor cost-effectiveness regions. The standardized axes highlight the synergistic impact of neonatal safety and delivery efficiency, confirming the robustness of the HBM-based education model under Ethiopian willingness-to-pay thresholds (Fig. 9). Incremental Cost-Effectiveness Ratio (ICER) Contour Plot per Newborn Complication Avoided Discussion This cluster-randomized controlled trial shows the effectiveness of a theory-based, Health Belief Model (HBM) education intervention on improving the humanistic and clinical indicators of intrapartum care in public healthcare clusters in Addis Ababa, Ethiopia. This study methodically restructures the ecosystem of antenatal and intrapartum care, questioning the historical paradigm of isolated, passive, or actively excluded labour companionship in low- and middle-income countries. The HBM’s theoretical framework was operationalized specifically to remove the multi-dimensional cognitive hurdles that would otherwise prevent labour companions from providing high-utility intrapartum care [10]. In most intrapartum care in sub-Saharan Africa, where labour companions are permitted at all, they act primarily as passive observers or structural navigators rather than active clinical and emotional buffers [22]. By providing education focused on the core concepts of the Health Belief Model (HBM), chosen labour companions were put on notice to become aware of about benefits continuous support, roles, responsible of labour companions and be trained in the techniques of continuous support required to help both themselves and the woman in childbirth achieve psychological and physical stability [23, 24]. In the current study the HBM-based education for labour companions intervention helped mothers to have positive experienced after giving birth compared to mothers who were not supported by their chosen labour companions through the entire childbirth experience, as well as specific improvements on birth outcomes. These birth outcome effects should be viewed with caution and should not be misconstrued as providing inadequate optimization during the entire continuum of clinical practice for women giving birth. The HBM-based education for labour companion intervention helped reduce the likelihood of having a prolonged labour, reduce the incidence of newborn complication, reduce lower Apgar score and reduce latent phase admission, but there were no statistically significant differences between the two groups in terms of mode of childbirth by spontaneous vaginal delivery, by caesarean section, or experiencing the incidence of acute maternal complications. While labour companionship can directly optimize the behavioral and psychophysiological aspects of childbirth, they cannot replace broader categorical and system-level structures and processes of both institutional and surgical processes related to public health. However, this intervention positively improved the maternal experience of childbirth; therefore mothers who received this intervention reported a statistically significant satisfaction than mothers who received the usual care, with an increase of 21.14 points on the Pregnancy and Childbirth Questionnaire (PCQ) scale than those receiving routine care. This substantial increase outpaces the 15% increase noted by role-orientation birth companion model in East Africa, suggesting that a structured, dyadic framework under the role-orientation birth companion model surpasses generic approaches [25]. Within the HBM matrix, brochures and antenatal sessions functioned as explicit “Cues to Action” that bridged the gap between a companion’s innate desire to assist and their “Self-Efficacy” in performing active supportive tasks, such as massage and mobility guidance, thereby transforming the birthing environment into a woman-centered care [13, 26]. The Generalized Estimating Equations (GEE) model provides significant socio-demographic insights into the dynamics of maternal satisfaction confirmed by the Wild Cluster Bootstrap framework. The model shows that housewives and women with formal employment have significantly lower satisfaction levels than women with self-employment in private business. This variation reflects levels of economic and psychological autonomy. Women working in private business in urban Ethiopia often have higher levels of baseline financial resilience, self-efficacy, and experience of navigating complex institutional bureaucracies [27]. On the contrary, housewives are more vulnerable to institutional stress [28] as they often suffer a more acute loss of personal control in very rigid, medicalized public hospital wards. Furthermore, the strong negative association we found for unmarried mothers suggests the impact of socio-cultural stigma on maternal safety. In the socio-cultural context of Ethiopia, out-of-wedlock childbearing is stigmatized and can undermine family support networks. These women’s lower satisfaction scores highlight the need for behavioural support, as they do not have the built-in social capital that married status automatically confers in traditional societies [28]. These findings offer a rationale for the inclusion of well-structured companion support as a protective structural tool designed specifically to buffer vulnerable sub-populations from isolated psychological birth trauma. GEE analysis of labour companion demographics provides insight into the socio-technical dynamics of intrapartum care. Labour companions with no formal education had mean satisfaction score of 21.3% lower than those with college education. A companion with no formal education in a complex public health hospital faces significant communication barriers [29]. They are less equipped to understand clinical jargon, more anxious to address medical equipment, and do not have the confidence to engage with maternal health care providers. Thus, their capacity to serve as an effective advocate for the labouring woman is structurally compromised [30]. Instead, companions who have a college education perfectly navigate the institutional space and can translate medical instructions for the mother and advocate for respectful care [29]. Interestingly, having a companion employed in private business increased maternal satisfaction by 29.5%. The positive correlation probably is related to good interpersonal communication, negotiation and problem solving skills which are inherently acquired in the private entrepreneurial sector in Addis Ababa. These companions are very skilled in negotiating interpersonal friction in stressful situations, constructively dealing with overburdened public health workers, and obtaining physical comfort measures for the woman in labour without provoking institutional conflict. This finding highlights that the educational and operational background of the companion is a critical factor in their ability to convert antenatal HBM training into active intrapartum advocacy. The GEE model also shows strong clinical and structural factors of maternal satisfaction besides socio-demographic variables. Mothers who are subjected to oxytocin acceleration experience a sudden and overwhelming escalation of physical pain and visceral distress that heavily drives down their subjective childbirth experience [29]. When labour progresses naturally and efficiently without exogenous augmentation partially facilitated by the companion’s non-pharmacological comfort measures the mother avoids this severe physical trauma, leading to a major elevation in satisfaction [26]. Besides, highly significant impact of institutional shift timing where childbirth occurring during the evening shift carried an 18% higher satisfaction score compared to the night shift reveals a critical health system delivery reality. During overnight hours, maternal healthcare providers are heavily reduced, senior consultants are typically off-site, and individual midwives face severe sleep deprivation and physical fatigue while managing high admission volumes [28]. Finally, the substantial negative impact of a complete lack of previous labour companion support confirms that the psychological shock of childbirth is severely magnified when a primigravida woman faces the experience without any concurrent peer safety framework [31]. This finding provides empirical support for the structural integration of HBM- based education for labour companions to break the cycle of isolated obstetric trauma [32]. Beyond subjective experiences, participants in the intervention group revealed a significant reduction in prolonged labour lasting 8 h and above establishing a major acceleration in parturition distribution (66.4% in the intervention group vs. 75.9% in the control group). These clinical outcomes were closely linked to the cervical status of the labouring women upon admission. Women in the intervention arm were significantly less likely to present in the late latent phase of labour (cervical dilatation <5 cm), with only 46.8% presenting in the latent phase compared to 56.8% in the control arm. The observed reduction in prolonged labour duration (> 8 h) and lower rates of late latent-phase admissions within the intervention group point toward improved labour progression efficiency. Within the framework of the HBM, preparing companions antenatally may help lower maternal anxiety upon facility admission. This supportive presence potentially limits the behavioral and psychological stress responses typically triggered by unfamiliar clinical environments, thereby facilitating optimal natural labour progression. However, as endogenous hormone dynamics were not measured, these associations remain hypothetical interpretations of the behavioral intervention’s success. Consequently, uterine contractions achieve maximum mechanical efficiency, optimizing fundal dominance and accelerating cervical transition [33]. Furthermore, addressing Perceived Barriers regarding mobility empowered companions to encourage vertical positions, utilizing gravitational force to intensify endogenous oxytocin expression via Ferguson’s reflex [32, 34]. The observed reduction in intermediate 1-minute APGAR scores and newborn complications within the intervention arm suggests that targeted companion support may have a positive downstream effect on early neonatal transition. While our study did not directly monitor intrapartum biophysical profiles or placental perfusion pathways, existing behavioral literature suggests that continuous emotional and physical reassurance may help reduce maternal distress during labour. This reduction in stress, alongside shorter labour durations, represents a plausible pathway through which birth outcomes are optimized, though further clinical studies utilizing direct physiological metrics are warranted to confirm these underlying mechanisms. Furthermore, prolonged labour directly increases the physical trauma of the fetal head against the maternal pelvic floor, escalating the risk of intracranial molding and deep decelerations in fetal heart rate [35]. By utilizing HBM-based education for labour companions to optimize labour speed and structurally reduce maternal distress, the intervention preserves continuous, uncompromised blood flow through the spiral arteries, effectively insulating the fetus from intrapartum hypoxic insults. Neonates in the intervention arm demonstrated a highly significant reduction in intermediate 1-minute APGAR scores establishing a lower probability of immediate newborn depression and a superior extrauterine transition. This is accompanied by a major drop in newborn complications. The lack of statistical divergence at the 5-minute APGAR mark reflects the high clinical efficacy of immediate, standardized neonatal resuscitation protocols uniformly deployed by delivery room personnel across both study arms [36].Additionally, newborns in the intervention group revealed a significantly higher mean birth weight compared to controls. This statistical variation is highly unlikely to stem from a direct physiological effect of behavioral companion education; rather, it may reflect residual confounding, understated baseline variations in maternal nutritional status, or minor differences in gestational age at delivery across clusters that were not fully captured by the GEE model. From an economic standpoint, this analysis confirms critical system-level implications of a cost-effectiveness neonatal condition prevention approach. Unlike macro-economic efficiency assessments the current economic analysis is entirely predicated on the Newborn Complication rate as the sole economic driving factor [37]. The HBM-based education for labour companion intervention yields a structural clinical optimization strategy in the form of a substantial absolute 6. 76% reduction in Newborn Complications. Intervention implementation including training, administration, and booklets costs 15.11USD (781.49ETB) per dyad, leading to an ICER of precisely 223.52USD (11,567ETB) per newborn complication avoided. On the cost-effectiveness plane, this intervention largely resides in the North-East-dominant quadrant, signifying additional cost for high impact safety benefits. Still, with explicit structural cost offsets for avoidance of extremely high expense NICU and ventilation and hospitalizations, the approach favors the South-East-dominant quadrant, making it economically sustainable for low-resource urban environments [38]. From a cost perspective, the ICER of 223. 52USD per neonatal complication avoided makes this approach extremely cost-effective for the Ethiopian 1× GDP per capita income measure. And, a 6. 76% absolute risk reduction in newborn complications implies a significant capacity for a health system to forego resource-dense secondary newborn care needs. Empirical evidence from this study should be leveraged to advocate for the structural integration of a theory-based education for labour companions intervention into the existing maternal health structure. Practical implication Healthcare facilities, especially in areas with low resources, should integrate the option of a chosen labour companion as a standard component of care, a practice endorsed by the WHO. The application of structured, theory-based (HBM) education for both the pregnant woman and her chosen companion is crucial. This approach enhanced perceived benefits by emphasizing how a companion improves pain management, reduces anxiety, and addressing common concerns such as lack of privacy or interference with medical care through clear guidelines and communication protocols and building the companion’s confidence in their ability to provide effective support (e.g., massage, encouraging words, alerting staff to changes).The study’s findings support the use of HBM-based education as an effective intervention reduce intermediate APGAR scores, shorten the duration of labour, and show favorable descriptive trends toward increased vaginal delivery rates, though the latter did not reach formal statistical significance. Healthcare providers can use the HBM framework to foster a more positive and respectful childbirth experience, as accompanied women report significantly higher satisfaction rates and reduced fear and anxiety during childbirth. Policy implications These findings carry profound implications for the national health policy architecture of East Africa. Rather than relying on unvalidated reductions in surgical intervention rates, which remained statistically non-significant after adjusting for the False Discovery Rate, the economic mandate for integrating HBM-based companion education into national policy rests firmly on its capacity to reduced neonatal complications and enhance maternal humanistic outcomes within a highly favorable willingness-to-pay profile (< 1× GDP per capita). Reflecting this, the Federal Ministry of Health of Ethiopia has placed patient-centered, high-safety delivery metrics at the center of its Health Sector Transformation Plan II (HSTP-II), primarily driven by the nationwide Compassionate, Respectful, and Caring (CRC) initiative. This formal mandate is designed to humanize institutional care, eliminate obstetric violence, and accelerate progress toward Sustainable Development Goals (SDG) 3.1 (reducing global maternal mortality to < 70 per 100,000 live births) and 3.2 (ending preventable neonatal and under-5 deaths). Historically, the nationwide scale-up of patient-centered care has been hindered by a deep-seated administrative perception that humanizing care requires unsustainable capital investments, such as restructuring physical delivery suites or expanding the formal midwifery workforce. This study fundamentally dismantles that assumption. Study strengths and limitations The main methodological strength of this study is the study design. By employing a cluster-randomized controlled trial in 12 separate health facility clusters, it has virtually eliminated contamination between arms which would have been possible if randomization had been done at individual level within shared, open-ward public Health facilities. Companions in the intervention cluster would have observed, learned, and then demonstrated support behaviors to control cluster companions. In an individual level randomization in such facilities, companionship to pregnant women in different arms will necessarily result in information contamination which will severely dilute observed treatment effect. Secondly, employing sophisticated statistics represents a major advancement over descriptive obstetric studies. GEE modeling comprehensively corrected for the intra-cluster correlation and time-dependent effects related to repeated measures of pregnant women. Crucially, GEE estimations were confirmed through 9,999 replicate Wild Cluster Bootstrap using Webb weights thus rendering impossible any bias arising from small clusters. Because there were only 12 facility clusters for this study, typical asymptotic statistics assumption could result in over-estimation and subsequent type 1 errors while under-estimation could result in under-powered studies. Wild Cluster Bootstrap gives statistically reliable p-values and confidence interval with the smallest sample size. Lastly, performing a thorough, probabilistic economic analysis concurrently with a clinical trial is a rarity in maternal child health. Study limitations Despite these strengths, several limitations must be acknowledged. First, due to the highly visible nature of behavioral interventions, it was methodologically impossible to blind the participating women, their labour companions, or the attending facility midwives to the group assignments. Second, maternal satisfaction was assessed via subjective, patient-reported recall at the endline interview. Although these interviews were conducted systematically, patient-reported outcome measures can be influenced by social desirability bias, where postpartum mothers may over-report satisfaction out of gratitude for a safe delivery or a healthy infant. Third, while early enrollment at 12–14 weeks gestation allowed the study to evaluate long-term behavioral changes on birth weight, it introduced a vulnerability to tracking maternal compliance across the second and third trimesters, which could introduce unmeasured variations in home-based diet or rest behaviors. Also, this trial was conducted entirely within the public health infrastructure of Addis Ababa, a highly dense, urban metropolitan environment. The findings may therefore lack direct generalizability to deep rural settings in Ethiopia, where obstetric care is delivered across primary health center networks characterized by vastly different structural constraints, longer geographical transit delays, and different socio-cultural companion dynamics. Fourth, while our study rigorously adjusted for multiple secondary endpoints using sharpened q-values, only a subset of secondary clinical markers retained their significance. The trial was fundamentally powered around the primary outcome of maternal satisfaction; hence, the interpretation of secondary outcomes is bounded by this constraint, and broad clinical or surgical improvements should not be inferred beyond the specific validated parameters of labour duration and immediate neonatal transition. Finally, the short-term clinical follow-up period used in our health economic evaluation introduces specific parameter and structural uncertainties regarding long-term resource trajectories. Because our economic model relies strictly on immediate birth outcomes specifically the acute newborn complication rate as the sole driver of the Incremental Cost-Effectiveness Ratio (ICER), it does not capture subsequent healthcare utilization or potential chronic morbidity beyond the immediate postpartum period. While this narrow intrapartum-to-postpartum horizon adheres directly to empirical supply-side and public health resource allocation ledgers in low-resource maternal health evaluations, our findings must still be viewed within these parameters. Consequently, our cost-effectiveness conclusions remain exploratory concerning long-term impacts. Future longitudinal cohorts should track these mother-infant dyads over extended postpartum periods to confirm whether the structural avoidance of acute neonatal complications and early intensive care admissions translates into sustained, multi-year economic and developmental benefits for the health system. Conclusion In conclusion, this cluster-randomized controlled trial provides robust evidence that a theory-based, prenatal educational intervention for labour companions, structured around the Health Belief Model (HBM), significantly improves maternal satisfaction and individual immediate neonatal transition markers in public health facilities. Activating labour companions over a multi-month prenatal period effectively reduced sociocultural barriers and enhanced their self-efficacy, resulting in a clinically stable increase in subjective childbirth satisfaction. While long-term modeling suggests the framework is highly cost-effective at 223.52 USD per neonatal complication avoided, these economic projections remain exploratory and warrant extended longitudinal tracking to confirm their lifetime structural impacts. Patient and public involvement statement Level of involvement Patients and members of the public first become involved in the design and proposal stage of the research study. Patients have been chosen to be on the CAG if they have already been involved with their local healthcare provision and from various backgrounds to ensure a good mix of views. Open recruitment through notice boards around the local area, and through various community initiatives is encouraged in an attempt to include the views of people from different ages, genders, and health conditions. Research questions and outcomes Both of our patient representatives have been involved with defining the primary research questions so that they address the priorities of patients. Patient representatives attended various meetings to achieve this and information gathered through surveys gave a good broad idea of what was felt by the majority; discrepancies in views were resolved through mediated discussion. Our patient representatives were involved in the design of our outcome measures so that they would be meaningful to patients. Study design and conduct Patients have been asked for feedback on our protocol and have considered how appropriate our proposed procedures are and time requirements, to ensure that patient burden is reasonable and that our information leaflet and consent form (patient-friendly version) are acceptable. Dissemination Patients will contribute to writing our project website, to include a summary of findings in ‘lay’ English, and will contribute to spreading the word of our findings via social media sites and local support groups. Supplementary Information Supplementary Material 1. Supplementary Material 2. Supplementary Material 3. Acknowledgements We thank you the study participants, data collectors, facility Heads and supervisors for their vital role in generating this research evidence. We acknowledge Jimma University, Faculty of Public health for their facilitation. We also thank Addis Ababa health Bureau for their cooperation and support letter. Lastly, we would like to thank you our friends for their unwavering support. Abbreviations Antenatal Care Cluster Randomized Controlled Trial Cesarean Section Generalized Estimating Equations Gross Domestic Product Health Belief Model Information, Education, and Communication Role Orientation Session Spontaneous Vaginal Delivery Wild Cluster Bootstrap World Health Organization Authors’ contributions G.A.Z. and M.A.W. drafted the research plan. A.M. helped to edit and review the manuscript. G.A.Z., M.A.W., and A.M. helped in the statistical analyses, and interpretation of findings, and proofreading. The final version of the paper has been reviewed and endorsed by all authors. Funding This study was funded by Jimma University. The funder had no role in the study design, data collection, analysis, interpretation of data, or writing of the manuscript. Data availability Can be obtained from the Corresponding author, upon request. Declarations Ethics approval and consent to participate Ethical approval for this study was granted by the Jimma University Institute of Health Institutional Review Board, Jimma University (Ref. No: JUIH/IRB/108/22; August 11, 2022). Further regional clearance was obtained from the Ethics Review Committee of the City Government of Addis Ababa Health Bureau (Ref. No: A/H/2077/227; December 23, 2023). A formal extension for continuous oversight was granted by the Addis Ababa City Government Health Bureau (Ref. No: A.A./25018/) on November 30, 2024, ensuring valid ethical coverage through December 2, 2025. We confirm that all methods and experiments involving human participants were performed in accordance with the relevant guidelines and regulations, including the Declaration of Helsinki and Good Clinical Practice (GCP) standards. Written informed consent was obtained from all participants, including the pregnant women and their chosen companions, prior to enrollment. In cases of limited literacy, the information sheet was read aloud in the presence of an independent witness, and consent was documented using the participant’s thumbprint and the witness’s signature. Participation was entirely voluntary, and all individuals were informed of their right to withdraw from the study at any time without penalty or loss of access to standard healthcare services (consent to participation see Annex1). Consent for publication Not applicable. Competing interests The authors declare no competing interests. Footnotes Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. References Associated Data Supplementary Materials Supplementary Material 1. Supplementary Material 2. Supplementary Material 3. Data Availability Statement Can be obtained from the Corresponding author, upon request.