Effect of community engagement interventions on child health care utilization: a quasi-experimental study in Afar and Benishangul Gumuz Regions, Ethiopia Authors: Abebe Gebremariam Gobezayehu, John N Cranmer, Gurmesa Tura Debelew, Tamiru Kassa Desalew, Aynalem Hailemichael Frew, Dagnew Hagezom Kahissay, Mariamawit Asfaw, Meles Solomon, Daniel Ngemera, Tiyese Chimuna, Bizuhan Gelaw Birhanu Published: 2026 ABSTRACT AbstractBackgroundChild health in Ethiopia is influenced by socioeconomic disparities, limited healthcare access, weak community engagement, and low service utilization, with greater challenges in resource-limited regions such as Afar and Benishangul Gumuz. To address these issues, Emory University Ethiopia, in partnership with UNICEF, Ministry of Health, and regional health bureaus, implemented community engagement interventions from February 2021 to April 2024 to improve health-seeking behavior and service use. However, their effects on care-seeking for sick children and infant immunization remain unassessed. This study addressed this gap.MethodsA pre–post quasi-experimental study was conducted in 14 Woredas (7 per region). Over three years, interventions included training quality improvement teams to identify mothers and children needing care and link them to facilities, alongside community festivals, dialogues, women’s conferences, and social and behavior change activities. Baseline and endline surveys were conducted in January 2021 and April 2024, respectively, with baseline immunization data drawn from the 2019 Demographic and Health Survey. Mixed-effects multilevel logistic regression assessed intervention effects by controlling for potential confounders (p < 0.05).ResultsAt the endline, 21.9% of children under 18 months of age (26.6% in Afar and 17.6% in Benishangul Gumuz) presented symptoms of common illnesses, a decrease from 43.3% at baseline (48.8% in Afar and 37.8% in Benishangul Gumuz). Among them, 90.4% (89.0% in Afar and 92.3% in Benishangul Gumuz) sought treatment at health facilities, an increase from 83.4% at baseline (75.1% in Afar and 94.1% in Benishangul Gumuz). The percentage of full immunization coverage increased from 40.0% at baseline to 66.4% at the endline. Mothers who participated in community interventions were nearly four times more likely to have their children fully immunized (AOR = 3.96; 95% CI: 1.70, 9.20). Other predictors included residence, parity, maternal education, knowledge, attitudes and healthcare utilization.ConclusionsCommunity engagement interventions significantly improved treatment-seeking behavior and the full immunization rate in the study area. Future efforts should focus on context-specific strategies that engage existing community platforms and address factors such as parity, education, and geographic location to enhance child health care.Supplementary InformationThe online version contains supplementary material available at 10.1186/s12889-026-27407-z.Keywords: Community engagement, Child health, Treatment seeking, Full immunization, Ethiopia Background The Sustainable Development Goal (SDG) targets 3.2.1 and 3.2.2 aim to reduce the under-five mortality rate (U5MR) to 25 or fewer and the neonatal mortality rate (NMR) to 12 or fewer per 1,000 live births by 2030. However, a recent estimate indicates that 4.9 million under-five children and 2.3 million neonates die globally each year, corresponding to a U5MR of 37 and an NMR of 17 per 1,000 live births, both of which are higher than the SDG targets. The burden is highest in sub-Saharan Africa, where the U5MR and NMR remain at 71 and 27 per 1,000 live births, respectively [1]. Ethiopia has made notable progress in reducing under-five mortality, which declined from 166 deaths per 1,000 live births in 2000 to 55 in 2019. Despite this achievement, both under-five mortality and neonatal mortality remain higher than the global average, and reductions in neonatal mortality have critically lagged. The neonatal mortality rate decreased from 49 to 33 per 1,000 live births between 2000 and 2019, a 32.7% decline over 20 years. As a result, neonatal deaths account for more than 52% of all under-five deaths, reflecting persistent gaps in newborn care, particularly in underserved settings [2]. These challenges are more pronounced in resource-limited, conflict-affected and pastoral regions. Benishangul Gumuz had the highest under-five and neonatal mortality rates at 90 and 58 per 1,000 live births, respectively and Afar had U5MR of 55 and an NMR of 38, both exceeding national figures and indicating a need for targeted interventions [2]. Achieving the SDG targets by 2030 in the country requires strengthened neonatal and child health services and equitable access to quality care across socioeconomic, geographic, and cultural contexts. However, persistent disparities driven by poverty, limited health system capacity, and unequal resource distribution contribute to preventable conditions such as malnutrition, low immunization coverage, inadequate treatment, and excess child mortality, particularly in resource-limited settings [3, 4]. Addressing these gaps necessitates multifaceted, people-centered approaches that integrate health system strengthening with culturally appropriate community engagement to improve health-seeking behavior, service utilization, and overall child health outcomes [5, 6]. Previous studies have shown that context-specific community engagement interventions substantially improve child health outcomes. Community-based maternal, neonatal, and child health programs in resource-limited settings have increased immunization coverage, improved nutritional status, and reduced child mortality, highlighting the importance of interventions tailored to local contexts and community realities [7–9]. In Ethiopia, community engagement in health programs has evolved from minimal involvement in the past to more structured and participatory approaches. Recent initiatives, including the Health Extension Program and participatory planning, have strengthened community involvement in health care delivery. National frameworks, including HSTP-I (2016–2020), HSTP-II (2021–2025), and reproductive health strategies (2021-2025), prioritized community participation in planning, implementation, and monitoring, alongside stronger community health systems, local evidence use, and community-based health education to promote ownership [10–14]. Despite these efforts, the effectiveness of community engagement in improving child health care utilization remains under studied in resource-limited settings, particularly in areas with dispersed settlements, mobile pastoralist populations, poor transportation, and other structural barriers. This gap is most pronounced for care-seeking for sick infants and full immunization, especially in Ethiopia’s Afar and Benishangul Gumuz regions. This study addressed this gap by implementing and rigorously assessing the effects of community engagement interventions in these contexts, providing actionable insights to enhance interventions, improve child health care, and accelerate progress toward the child survival targets of the sustainable development goals. Furthermore, a formative assessment conducted in January 2021 by UNICEF and Emory University Ethiopia in Afar and Benishangul Gumuz regions identified key barriers to maternal, neonatal, and child health (MNCH) services, including delays in seeking and receiving care, poor service quality, and weak community engagement. These findings informed the project “Optimizing Care for Newborns and Children in Afar and Benishangul Gumuz Regions,” which aimed to improve maternal health knowledge, attitudes, and care-seeking behaviors through context-specific community engagement interventions [14]. After more than three years of implementation (February 2021–April 2024), Emory University Ethiopia and its partners evaluated the interventions’ effectiveness. This manuscript reports the impact of these community engagement interventions on child health care, focusing on treatment-seeking for sick children and full immunization coverage. Methods study setting The study was conducted in selected intervention Woredas within Afar and Benishangul Gumuz regions of Ethiopia. The community-engagement interventions were implemented in 11 Woredas/Woredas with 172 kebeles (the smallest administrative unit next to the Woreda) in Benishangul Gumuz and 10 woredas with 45 kebeles in Afar region. Both regions experience persistent socioeconomic disadvantages and limited access to quality maternal, newborn, and child health (MNCH) services. They were strategically selected because of their high neonatal and under-five mortality rates and underperforming health systems, making them national priorities for targeted community-engagement initiatives [2]. Benishangul Gumuz, in northwestern Ethiopia along the Sudanese border, is characterized by sparsely distributed populations, limited health infrastructure, shortages of skilled health workers, and the region with the country’s highest infant, neonatal, and under-five mortality rates. Afar, in northeastern Ethiopia, is predominantly pastoralist, with harsh desert conditions, extreme temperatures, and highly mobile populations that challenge primary health care delivery. In both regions, the low utilization of skilled maternal and newborn services and persistently high neonatal mortality underscore the urgent need for context-specific community engagement strategies to enhance care-seeking behaviors and MNCH service uptake. Interventions The interventions aimed to overcome barriers in seeking, accessing, and receiving neonatal and child health care. Various community engagement strategies were implemented to increase demand for MNCH services. The details of the intervention are described below. Community quality improvement (QI) interventions This intervention focuses on strengthening community-level problem solving and accountability through regular monthly QI meetings. The participants include a multidisciplinary QI team of 12–14 members drawn from each village/kebele, comprising Health Extension Workers (HEWs), village and clan leaders, religious leaders, Traditional Birth Attendants (TBAs), and Women Development Group (WDG) representatives. These members play a critical role in identifying and referring pregnant women, sick infants, and postnatal mothers to appropriate health services. They also analyze local health post data to guide decisions, identify key barriers to care-seeking, and design context-specific solutions to improve maternal, newborn, and child health (MNCH) outcomes within their communities. Community dialogue Community dialogue sessions were conducted quarterly to promote awareness and collective understanding of MNCH issues. Participants include HEWs, pregnant and lactating women, husbands/partners, family members, TBAs, WDG representatives, and influential community figures such as elders, clan leaders, kebele administrators, and religious leaders. The primary role of these participants was to engage in open discussions that promote health awareness, encourage positive behavior change, and address misconceptions related to maternal and child health. These forums also enhance community accountability by creating a shared responsibility for improving health practices and supporting timely utilization of health services. Community festival The community festival was a broader engagement platform organized quarterly to integrate service promotion with cultural and social mobilization. It involves participants from the woreda health office, health workers, HEWs, pregnant and lactating women, families, TBAs, WDGs, and community leaders, including religious and clan leaders and kebele administrators. Assigned community mobilizers engage and inform communities about the festival one day prior to the event. During these events, participants disseminate MNCH messages, promote available health services, and in some cases provide services directly through health corners equipped with tools such as portable ultrasound. The festival also addressed cultural barriers that hinder service uptake and included recognition and rewarding of community champions and model families to reinforce positive health behaviors. Pregnant women conference This intervention consisted of monthly gatherings of approximately 30 pregnant women, identified and mobilized by TBAs and WDGs, and facilitated by HEWs. The conferences provided a supportive environment for peer learning and experience sharing among pregnant women. Participants discuss their experiences, challenges, and best practices related to pregnancy and childbirth. The HEWs play a facilitative role by providing health education and promoting key services such as antenatal care (ANC), skilled delivery, and postnatal care (PNC). This approach helped to build confidence, improve knowledge, and encourage timely utilization of maternal health services. Women empowerment & social and behavior change communication (SBCC) This intervention aimed to enhance women’s decision-making power and improve household-level health practices through structured SBCC activities conducted quarterly. Participants included pregnant and lactating women, husbands/partners, mothers-in-laws, influential family members, women leaders, volunteers, HEWs, and facility-based health staff. Their roles involve engaging in health education sessions, utilizing multimedia communication tools to reinforce key MNCH messages, and participating in household-level discussions to influence behaviors. This intervention emphasized inclusive engagement of key decision-makers within the family to address social norms and barriers, ultimately promoting sustained behavior change and improved health outcomes. To ensure intervention fidelity and adequate dose, standardized implementation manuals were developed by Emory University officers. Monthly activity reports and quarterly aggregated performance data were systematically tracked throughout the intervention period (February 2021–April 2024). Monitoring and evaluation were conducted through quarterly joint supportive supervision involving Emory university local project officers, Regional Health Bureau child health experts, and woreda health offices and facilities. In addition, regular quarterly performance review meetings and learning sessions were conducted to assess progress, address implementation gaps, and promote continuous improvement of interventions (Table 1). Table 1Community engagement intervention strategies in afar and Benishangul Gumuz Region, Ethiopia, Feb 2021-April 2024Intervention typeParticipantsRoles in the interventionCommunity Quality improvement (QI)(monthly meeting)QI team members composed of 12–14 individuals one per village/kebele: members include HEWs, Village leaders, clan leaders, Religious leaders, Traditional Birth Attendant (TBAs)and Women Development Group (WDG) representatives• Identify & refer cases (pregnant women, sick infants and postnatal mothers)• Local problem-solving (identify community problems and design community solutions focusing on care seeking• They use health post data for their actionsCommunity Dialogue (quarterly event)HEWs, pregnant women and lactating mothers, husbands/partners and family members, TBAs, WDGs, Community leaders and elders, clan leaders, kebele administrators, and religious leaders• Awareness creation• Behavior changes messaging• Misconceptions clarifying• Community accountabilityCommunity Festival (quarterly event)Woreda health office, health workers, HEWs, pregnant women and lactating mothers, husbands and family members, TBAs, WDGs, Community leaders and elders, clan leaders, religious leaders and kebele administrators.• MNCH messaging• Service promotion and provision through health corner with portable ultrasound• Address cultural barriers• Community recognition (champions/models rewarding)Pregnant Women Conferences (monthly event)About 30 pregnant women identified through TBAs and WDGs and facilitated by HEWs• Peer learning• Services (ANC, delivery and PNC) promotion• Experience sharingWomen Empowerment & SBCC (quarterly)Pregnant and lactating women, husbands and partners, mothers-in-law and other influential family members, women leaders and volunteers, health extension workers and facility staff• Health education• Multimedia communication• Household engagement Community engagement intervention strategies in afar and Benishangul Gumuz Region, Ethiopia, Feb 2021-April 2024 Community Quality improvement (QI) (monthly meeting) • Identify & refer cases (pregnant women, sick infants and postnatal mothers) • Local problem-solving (identify community problems and design community solutions focusing on care seeking • They use health post data for their actions • Awareness creation • Behavior changes messaging • Misconceptions clarifying • Community accountability • MNCH messaging • Service promotion and provision through health corner with portable ultrasound • Address cultural barriers • Community recognition (champions/models rewarding) • Peer learning • Services (ANC, delivery and PNC) promotion • Experience sharing • Health education • Multimedia communication • Household engagement The interventions were aligned with the World Health Organization’s (WHO 2020) four levels of community engagement: community-oriented, community-based, community-managed, and community-owned [5]. This approach ensured that the interventions were tailored to the local context, addressing specific community needs and involving key stakeholders. To make them context-specific and enhance relevance, a people-centered design approach was employed, allowing the community to identify problems and cocreate solutions. Cultural aspects, including local languages and health practices, were integrated. Influential people, such as clan leaders, religious leaders, traditional birth attendants, and husbands, participated throughout the process. Local quality improvement team members were fully engaged in community engagement interventions, to promote ownership and accountability, and were rewarded for their efforts to promote positive behavioral change via locally sourced materials. The main roles of community-level actors were to raise awareness, promote services, identify and link pregnant women, sick neonates and infants not immunized per their age (Table 2). Table 2Four Levels of community engagement continuum adapted from the 2020 World Health Organization’s guideline for the study in Afar and Benishangul Gumuz, Ethiopia, Feb 2021-April 2024Community-Oriented ◊Community-Based ◊Community-Managed ◊Community-Owned• Designed by experts• Community as target population• Minimal input from community• Implemented in community settings• Limited community participation• Shared decision-making with community• Community involved in planning & implementation• Fully led by community• External actors facilitate• High local ownership Four Levels of community engagement continuum adapted from the 2020 World Health Organization’s guideline for the study in Afar and Benishangul Gumuz, Ethiopia, Feb 2021-April 2024 • Designed by experts • Community as target population • Minimal input from community • Implemented in community settings • Limited community participation • Shared decision-making with community • Community involved in planning & implementation • Fully led by community • External actors facilitate • High local ownership Study design A non-randomized pre-post quasi-experimental study design was employed. At the beginning of the Optimizing Care for Newborns and Children project, a formative assessment was conducted to establish baseline data, measuring key maternal, neonatal, and child health (MNCH) indicators, in January 2021. However, detailed data on childhood immunization were not collected at that stage. Consequently, baseline immunization data for the two regions were estimated from the 2019 Ethiopian Mini-Demographic and Health Survey [2]. As this was conducted two years before the start of the project, it was further cross-checked and validated with the changes in the region with the DHIS 2 facility data of the respective region. After more than three years of intervention, beginning in February 2021, an endline cross-sectional survey was conducted in April 2024 to assess key MNCH services, including data on illness- and health-seeking behavior and the immunization status of infants. The effectiveness of the intervention was assessed by comparing the endline data with the baseline data, as well as examining differences between participants who engaged in the interventions and those who did not (internal comparisons). Population, sample size The study population (respondents) consisted of mothers with children under 18 months of age. Children under 18 months of age were included in this study to capture the full continuum of maternal, newborn, and child health services, including antenatal care, delivery, postnatal care, early infancy care, and completion of the routine immunization schedule that had happened during the intervention period. This age range also helps minimize recall bias, as mothers are more likely to accurately report recent health service utilization and events compared to those with older children. This study analyzed larger data from the optimizing care project evaluation, for which the sample size calculations were based on institutional delivery service utilization rates from the 2019 M-EDHS [2]: 64.9% in Benishangul Gumuz and 30.6% in Afar. With a 95% confidence level, a 5% margin of error, a design effect of 2 and a 10% nonresponse rate, a total of 745 mothers were sampled, 385 from Benishangul Gumuz and 360 from Afar. Sampling technique To ensure representativeness, separate sample sizes were calculated for each region, and 14 out of 21 woredas (7 from each region) were randomly selected, with two kebeles chosen per woreda, resulting in a total of 28 kebeles. Although woredas had over ten kebeles on average, two kebeles per woreda were randomly selected for logistical feasibility and balanced representation. A quota-based approach ensured equal contribution from each woreda, avoiding the underrepresentation of those with fewer kebeles that could occur with proportional allocation. Then, sample sizes were distributed accordingly to have a comparable number of mothers per kebele, with approximately 28 mothers per kebele in Benishangul Gumuz and 26 in Afar. House-to-house surveys were conducted to identify eligible mothers (having children under 18 months). When the number of eligible mothers in a kebele was close to the allocated sample, all were included; when it exceeded the allocation, simple random sampling was applied (Additional file 1). Data collection process, tools and quality assurance Structured, pre-coded and pretested questionnaire, which was developed by the research team (Additional File 2) was used to gather data on child health care knowledge, attitudes, maternal health care utilization, sick child treatment and full immunization by interviewing mothers. The data on each antigen of the immunization and full immunization for all basic vaccines were taken from the mothers’ oral reports and cross-checked with the child’s immunization card where available. The questionnaire was initially developed by the research team in English, then translated into Amharic by language expert and administered by experienced data collectors fluent in local languages. The data were collected via house-to-house surveys via face-to-face interviews with BSc nurses and supervised by MPH holders. To ensure data quality, experienced data collectors were recruited and trained intensively on the study’s purpose, methods, and ethical considerations. Data collection processes were closely supervised to ensure adherence to procedures and quality. Data processing and analysis The collected data were entered into EPI-INFO V7.2 and analyzed with STATA 17. Descriptive statistics, including means and percentages, were used to describe the respondents’ background. From a total sample of 745 mothers of children under 18 months, 446 mothers with children aged 9 to 18 months were selected to evaluate the effect of the community segment interventions on full immunization coverage and identify other key determinants. To evaluate the effect of the community engagement intervention on full immunization coverage, initial bivariate analysis was conducted to test the significance and estimate crude odds ratios and identify potential confounding variables, which were selected as candidate covariate at a significance level of p < 0.25. Subsequently, multivariate analysis was performed to adjust for potential confounders, with statistical significance assessed at p < 0.05. A multistage cluster sampling design was applied to select study participants, with woredas serving as primary clustering units and mothers nested within kebeles and kebeles nested within woredas. Given this hierarchical data structure, observations within clusters are likely to be correlated due to shared contextual factors such as access to health services, geographic proximity, and residence characteristics. This clustering violates the independence assumption underlying standard logistic regression and conventional chi-square–based analyses, making such approaches inappropriate for valid inference. To address this issue, a mixed-effects multilevel logistic regression model was employed to account for intra-cluster correlation and unobserved heterogeneity at the woreda level while simultaneously adjusting for individual- and cluster-level covariates. Model evaluations supported the appropriateness of this approach: the null model ICC was 9.8%, indicating meaningful clustering, and increased to 26.5% in the full model, suggesting that a substantial proportion of variability in full immunization is attributable to cluster-level factors. Further evidence is provided by model fit statistics and variance decomposition measures. The log likelihood improved from the null model (−274.2) to Model 1 (−265.3), indicating better fit with cluster-level variables. A larger improvement occurred in Model 2, when individual-level variables included (−171.8), and the full model (Model 3) achieved the highest log likelihood (−167.7), confirming the best overall model fit (p < 0.001). The Akaike Information Criterion (AIC) and Bayesian Information Criterion (BIC) decreased across successive models, with the lowest values observed in the full model, indicating improved goodness of fit. Although proportional change in variance (PCV) reflects partial reduction in unexplained cluster-level variance with the inclusion of covariates, residual heterogeneity remains. Collectively, these model evaluation findings indicated that mixed-effects multilevel logistic regression is more appropriate than ordinary logistic regression, as it accounts for clustering, improves model fit, and yields more reliable and unbiased estimates (Table 3). List and description of the variables at different levels is indicated in the table below (Table 4). Table 3Model fit statistics and random effects estimates for mixed-effects multilevel logistic regression analysis to on the effect of community engagement intervention on full immunization in afar and Benishangul Gumuz regions, Ethiopia, April 2024ParametersModel 0: Empty ModelModel 1: Higher level variables includedModel 2: Individual level variables includedModel 3: Full model (all variables included)Intraclass Correlation Coefficient(ICC)9.8%16.5%21.0%26.5%Log Likelihood Ratio (LR)−274.2−265.3−171.8−167.7Significance of Log Likelihood (LR) Vs logistic regression (p-value)< 0.001< 0.001< 0.001< 0.001Akaike Information Criterion (AIC)552.3538.5380.2376.5Bayesian Information Criterion (BIC)560.5554.9429.4433.9Cluster-level variance0.420.650.881.18Median Odds Ratio (MOR),1.862.162.452.82Proportional Change in Variance (PCV)Reference0.551.101.81 Model fit statistics and random effects estimates for mixed-effects multilevel logistic regression analysis to on the effect of community engagement intervention on full immunization in afar and Benishangul Gumuz regions, Ethiopia, April 2024 Table 4Description of variables and measurements for the effects of community engagement interventions on full immunization in Afar and Benishangul Gumuz Regions, Ethiopia, April 2024VariablesDescription of measurementOutcome variablesTreatment seekingThe place where treatment was sought for sick children in the two weeks period before the survey was collected from mothers’ interview response and categorized as: “treated at Health facility”, “drug purchased from private drug vendor/pharmacy without order” and “no-treatment except home-based care by mothers”Full immunizationA child who took all the basic vaccines as per the schedule as reported by mothers and confirmed from immunization card where available: BCG, three doses of Polio vaccine, three doses of Pentavalent vaccine (DPT + Hep B + Hib), three doses of PCV, two doses of Rota virus vaccine and one dose of measles vaccine. Cluster variable WoredaWoreda, also called Woreda, was taken as a cluster variable from administrative labelingRegionRegion was labeled as “Afar” and “Benishangul Gumuz” administrative labelingPlace of residenceActual place of residence of the mothers was categorized as “Urban” and “Rural” Lower (individual-level) variables Wealth indexThe wealth index was measured by using household assets (utensils and utilities), communication/information materials (TV, radio, phone), bank account/book, agricultural farming lands, livestock (farm animals) as customized from the EDHS. Then, composite index was produced by using PCA and later categorized into three (poor, middle and rich).AgeAge of mother was measured in full years from own report during interview and categorized in 5 years interval and later recategorized as 18–24, 25–34 and ≥ 35.Educational statusThe educational status was also measured by mother’s own report and categorized unable to read and write, able to read and write but no formal education, primary (Grade 1–8), secondary (Grade 9–12), higher education/college (Grade 12+) and later merged and recategorized for regression.ParityThe number of births a mother ever had (including still births) were asked and recategorized as primi (1 birth), multipara (2–3) and grand multipara (4 or more).Knowledge of mothers on child immunizationThe knowledge of mothers on child immunization was measured by using 7 “Yes” or “No” questions related to vaccine preventable diseases, available vaccines, place to get vaccine, age of the child? and schedule of each immunization. Then, composite score was produced and those scored above or equal to the mean score was labeled as having good knowledge and otherwise poor knowledge.Attitude of mothers toward child immunizationThe attitude of mothers toward child immunization was measured by using 9 Likert scale questions related to vaccine safety, number of doses and frequency, advantages (benefits), side effects and adequacy of health information given. Then, composite score was produced and those scored above or equal to the mean was labeled as having “Favorable attitude” and otherwise, “non-favorable attitude”.ANCMothers were asked about ANC visits during their pregnancy and categorized as “No ANC visit at all”, “1–3 ANC visits” and Four or more ANC visits”DeliveryThe place where the woman had given her last birth was asked and categorized as: Health facility delivery” or “Home delivery”PNCMothers were asked about PNC follow up and categorized as “Yes” or “No”InterventionCommunity engagement intervention in this study was measured by using 4 activities: attending community festivals, Community dialog and learning session, pregnant women conference and Social and behavioral Change Communication (SBCC) session. A composite score was produced for the four indexes and categorized as: “Not attended any one of the interventions”, “Attended only 1–2 of the interventions” and “Attended 3–4 of the interventions” Description of variables and measurements for the effects of community engagement interventions on full immunization in Afar and Benishangul Gumuz Regions, Ethiopia, April 2024 Results Sociodemographic characteristics For the end-line household survey, from the initial planned sample size of 745, a total of 726 mothers of children under the age of 18 months participated in the study, resulting in a response rate of 97.5%. The response rates in the two regions were similar. From the planned sample of 385 participants from the Benishangul Gumuz region, 376 participated (response rate of 97.7%), and from a sample of 360 participants from the Afar region, 350 participated (response rate of 97.2%). Majority of the respondents, 441 (60.8%), were in the age range of 25–34 years, and the age distribution was similar in both regions. Overall, 585 (80.6%) were from rural residences, and the rural‒urban distribution was similar in both regions. Nearly half of the respondents, 344 (47.4%), have neither attended formal education nor can read or write. The proportion of mothers who could not read and write in the Afar region, 281 (80.3%), was substantially greater than that in the Benishangul Gumuz Region, 63 (16.8%) (Table 5). Table 5Sociodemographic and economic characteristics of the study participants in the Afar and Benishangul Gumuz Regions, Ethiopia, April 2024 (n = 726)CharacteristicsAfar (n = 350)n(%)Benishangul Gumuz (n = 376) n(%)Total (n = 726)n(%) Age 15–196(4.0)38(10.1)44(6.0)20–2468(19.4)95(25.3)163(22.5)25–29130(37.1)134(35.6)264(36.4)30–34109(31.1)68(18.1)177(24.4)≥3537(10.6)41(10.9)78(10.7) Residence Rural280(80.0)305(81.1)585(80.6)Urban70(20.0)71(18.9)141(19.4) Educational status Unable to read and write281(80.3)63(16.8)344(47.4)Read and write only35(10.0)45(12)80(11.0)Primary (Grades 1–8)20(5.7)142(37.8)162(22.3)Secondary school (Grades 9–12)10(2.9)76(20.2)86(11.8)College and above (12+)4(1.1)50(13.3)54(7.4) Marital status Married337(96.3)370(98.4)707(97.4)Others*13(3.7)6(1.7)19(2.6) Wealth Index Lowest64(18.3)176(46.8)240(33.1)Middle106(30.3)142(37.8)248(34.2)Highest180(51.4)58(15.4)238(32.8)*Single mother or divorced Sociodemographic and economic characteristics of the study participants in the Afar and Benishangul Gumuz Regions, Ethiopia, April 2024 (n = 726) *Single mother or divorced Mothers’ knowledge of child health care The majority, 666 (91.7%), of the mothers (95.5% in Benishangul Gumuz and 87.7% in Afar) knew at least one childhood illness. Overall, diarrhea, 624 (93.8%), fever, 463 (69.6%) and cough, 426 (64.1%) were the most common childhood illnesses known by mothers. The proportions of mothers who mentioned these childhood illnesses were comparable in the two regions. The great majority of the mothers, 620 (94.2%) (90.5% in Benishangul Gumuz and 98.7% in Afar), knew that sick children must be taken to health institutions. Most, 670 (92.3%), knew at least one disease that could be prevented by vaccines. A total of 515 (76.6%) measles cases were known by mothers, followed by those known by mothers, with 460 (68.5%) cases (Table 6). Table 6Knowledge of mothers regarding childhood illnesses and immunization: the Afar and Benishangul Gumuz Regions, Ethiopia, April 2024VariableCategoryBG (N = 376)n(%)Afar (N = 350)n(%)Total (N = 726)n(%)Heard childhood illnessYes359(95.5)307(87.7)666(91.7)No17(4.5)43(12.3)60(8.3)Childhood illnesses known (multiple responses) (n = 666)Diarrhea333(92.8)291(95.1)624(93.8)Fever262(73.0)201(65.7)463(69.6)Cough252(70.2)174(56.9)426(64.1)Difficulty of breathing168(46.8)44(14.4)212(31.9)Lack of growth92(25.6)48(15.7)140(21.1)Convulsion58(15.4)37(12.1)95(14.3)Known place to take sick child (n = 666)Health facilities323(90.5)297(98.7)620(94.2)Home remedies27(7.2)1(0.15)28(4.3)Others (religious or traditional)7(1.4)3(0.45)10(1.6)Know at least one vaccine-preventable diseaseYes367(97.6)303(86.9)670(92.3)No10(2.4)46(13.1)56(4.6)Vaccine-preventable disease known (multiple response) (n = 670)Measles261(71.1)254(83.6)515(76.6)Polio321(87.5)139(45.7)460(68.5)Pneumonia266(72.5)116(38.2)382(56.8)Diarrheal disease220(59.9)149(49)369(54.9)Tetanus197(53.7)66(21.7)263(39.1)Pertussis93(24.7)158(52)251(37.4)Tuberculosis176(48.1)47(15.5)223(33.2)Meningitis96(26.2)59(19.4)155(23.1)Diphtheria75(20.4)24(7.9)99(14.7)Hepatitis61(16.6)10(3.3)71(10.6) Knowledge of mothers regarding childhood illnesses and immunization: the Afar and Benishangul Gumuz Regions, Ethiopia, April 2024 Attitudes of mothers toward child health Overall attitudes toward the appropriate child health care services (health seeking and immunization) computed from 9 relevant key attitudes via Likert scale questions indicated that 513 (70.7%) mothers, 281 (74.3%) in Benishangul Gumuz and 232 (66.3%) in the Afar region, reflected favorable attitudes toward appropriate child health care services (Fig. 1). Mothers’ attitudes toward child health in the Afar and Benishangul Gumuz Regions, Ethiopia, April 2024 Treatment seeking practice for sick children Overall, 159 (21.9%) of the total 726 children under 18 months of age (17.6% in Benishangul Gumuz and 26.6% in Afar) had any one of the symptoms of common childhood illness in the last two weeks. The prevalence of any of the symptoms decreased from 43.3% at baseline to 21.9% at the end of the study (p < 0.001). A decline was observed in the prevalence of fever (from 19.7% to 5.5%) and diarrhea (from 15% to 6.6%), whereas no decline was observed in the prevalence of cough. Treatment seeking for sick children increased from 83.4% at baseline to 90.4% at the end of the study. Specifically, the prevalence of fever increased from 79.4% to 90.9%, that of diarrhea increased from 85.4% to 89.6%, and that of cough increased from 89.2% to 91.2%. Regional disaggregation revealed a notable increase in treatment seeking in the Afar region, where there was a significant gap at baseline, increasing from 75.1% to 89.0%. In contrast, in Benishangul Gumuz, treatment seeking remained almost unchanged, as it was already higher at baseline than at the endline (Table 7). Table 7Two-week prevalence of common symptoms of childhood illnesses and treatment-seeking behavior in the Afar and Benishangul Gumuz Regions, Ethiopia, April 2024RegionFeverDiarrheaCoughAnyone symptomTwo weeks prevalenceBaselineEndlineBaselineEndlineBaselineEndlineBaselineEndlineAfar23.78.017.28.07.910.048.826.6Benishangul Gumuz15.63.212.85.39.48.837.817.6Total19.75.515.06.68.69.443.321.9Treated at health facilityAfar70.889.379.885.778.091.475.189.0Benishangul Gumuz92.491.792.895.098.690.994.192.3Total79.490.985.489.689.291.283.490.4 Two-week prevalence of common symptoms of childhood illnesses and treatment-seeking behavior in the Afar and Benishangul Gumuz Regions, Ethiopia, April 2024 The findings from the internal comparison revealed that treatment-seeking behavior improved in line with the number and types of interventions a mother participated in. Seeking treatment at health facilities varied from 89.5% among mothers who never participated in any of the community engagement interventions to 97.9% among those who participated in three or more of the interventions (p < 0.01). However, purchasing drugs from private drug vendors/pharmacy and giving home-based care without taking them to health facilities were in the reverse order (Fig. 2). Relationship between participation in community engagement interventions and seeking treatment for sick children in the Afar and Benishangul Gumuz Regions, Ethiopia, April 2024 Full immunization coverage for all basic childhood vaccines This section assessed full immunization coverage and the effect of community engagement interventions among children aged 9–18 months. Of 446 children, 296 (66.4%) were fully immunized, with regional variation (71.0% in Benishangul Gumuz and 61.1% in Afar). Full immunization coverage increased from 40% at baseline to 66.4% at endline overall, rising from 19.7% to 61.1% in Afar and from 66.1% to 71.0% in Benishangul Gumuz (Fig. 3). The full immunization rate before and after the community engagement intervention in the Afar and Benishangul Gumuz Regions, Ethiopia, April 2024 The findings also show that as the number and type of community engagement interventions attended by mothers increase, the full immunization rate increases, ranging from 40.2% among those who never participated in any of the interventions to 80% among those who participated in all the intervention types. The full immunization rate among those who never participated in any of the interventions was similar to that of the baseline data (40.2% vs. 40%) (Fig. 4). Comparison of full immunization coverage between mothers who attended and not attending community engagement interventions in the Afar and Benishangul Gumuz Regions, Ethiopia, April 2024 Mixed-effects multilevel analysis of the effects of community engagement interventions on the full immunization of children To assess the effect of community engagement interventions on full immunization coverage, a mixed-effects multilevel logistic regression was conducted with Woreda taken as the cluster variable. Cluster-level variables (region and residence) and individual-level factors (wealth index, maternal age, education, and parity) were included as covariates based on p-values < 0.25 in bivariate analysis. Additionally, maternal knowledge, attitudes toward immunization, and use of ANC, facility delivery, and PNC services met the inclusion criterion of p < 0.25 and were entered into the final multilevel model (Table 8). Table 8Bivariate analysis of the community engagement interventions and other covariates on full immunization in Afar and Benishangul Gumuz regions, Ethiopia, April 2024VariablesFully immunized (n = 446)Crude OR(95%CI)Yes (n = 296) n(%)No (n = 150) n(%) Community engagement interventions Not attended any of the interventions33(37.9)54(62.1)1.00Attended only 1–2 of the interventions114(68.3)53(31.7)3.52(2.05, 6.05)Attended 3 or more of the interventions149(77.6)43(22.4)5.67(3.27,9.83) Higher level (Level 1) Covariates RegionAfar127(61.1)81(38.9)1.00Benishangul Gumuz169(71.0)69(29.0)1.56(1.05, 2.32)Place of residenceRural230(63.9)130(36.1)1.00Urban66(76.7)20(23.3)1.87(1.08, 3.22) Lower-level (Level 2) covariates Wealth indexLowest92(63.0)54(37.0)1.00Middle class100(63.7)57(36.3)1.03(0.65, 1.64)Highest104(72.7)39(27.3)1.57(0.95, 2.58)Age of mother (years)18–2477(64.7)42(35.3)1.0025–34180(65.9)93(34.1)1.06(0.67, 1.66)≥ 3539(72.7)15(27.8)1.42(0.70, 2.87)Educational status of motherUnable to read and write109(53.7)94(46.3)1.00Able to read and write41(73.2)15(26.8)2.36(1.23, 4.53)Primary or above156(78.1)41(21.9)3.07(1.97, 4.78)ParityPara 1 (1st child)33(44.6)41(55.4)1.00Para 2–3 (multipara)106(65.4)56(34.6)2.35(1.34, 4.12)Para 4 or above (Grand multipara)157(74.8)53(25.2)3.68(2.12, 6.41)Knowledge of mother on child immunizationPoor knowledge22(32.8)45(67.2)1.00Good knowledge274(72.3)105(27.7)5.34(3.07, 9.32)Attitude of mother toward child immunizationUnfavorable attitude123(55.4)99(44.6)1.00Favorable attitude173(77.2)51(22.8)2.73(1.81, 4.11)ANC follow upNo ANC visit17(29.8)40(70.2)1.001–3 ANC visits76(68.5)35(31.5)5.11(2.55, 10.23)4 or more ANC visits203(73.0)75(27.0)6.37(4.41, 11.91)Place of deliveryHome birth38(43.2)50(56.8)1.00Health facility birth258(72.1)100(27.9)3.40(2.01, 5.49)At least 1 PNC follow upNo37(34.3)71(65.7)1.00Yes259(76.6)79(23.4)6.29(3.93, 10.07) Bivariate analysis of the community engagement interventions and other covariates on full immunization in Afar and Benishangul Gumuz regions, Ethiopia, April 2024 After adjusting for confounding factors in the mixed effects multilevel logistic regression analysis, community engagement interventions had significant association with full immunization. Mothers who participated in one or two of the interventions were more than twice as likely to have their children fully immunized (AOR = 2.56; 95% CI: 1.13, 5.78), whereas those involved in three or more of the interventions were nearly four times more likely to have their children fully immunized (AOR = 3.96; 95% CI: 1.70, 9.20). With respect to the cluster-level variables, place of residence was significant; urban mothers were more than three times more likely to have their children fully immunized than rural mothers (AOR = 3.58; 95% CI: 1.41, 9.09). Maternal education was also a significant predictor of full immunization: mothers who were able to read and write (AOR = 4.10; 95% CI: 1.39, 12.05) and those with primary education or above (AOR = 3.52; 95% CI: 1.55, 7.98) had increased odds. Multiparous mothers were also more likely to fully immunize their children than primiparous mothers (AOR = 2.86; 95% CI: 1.17, 7.01 for para 2–3; AOR = 6.74; 95% CI: 2.38, 19.07 for para ≥ 4). Maternal knowledge (AOR = 4.65; 95% CI: 1.84, 11.78) and favorable attitudes (AOR = 5.42; 95% CI: 2.59, 11.36) were strongly associated with full immunization. Health service utilization also showed significant effects: 1–3 ANC visits (AOR = 5.48; 95% CI: 2.05, 14.64) and ≥ 4 visits (AOR = 4.95; 95% CI: 1.99, 12.29), facility delivery (AOR = 3.57; 95% CI: 1.56, 8.16), and at least one PNC visit (AOR = 7.95; 95% CI: 3.92, 16.09) all increased the likelihood of full immunization (Table 9). Table 9Mixed effects multilevel analysis of the effect of Community engagement intervention on full child immunization in Afar and Benishangul Gumuz Regions, Ethiopia, April 2024VariablesModel 0: Empty modelModel 1: Higer- level factors AOR(95%CI)Model 2: Lower- level factors AOR(95%CI)Model 3: Final full modelAOR(95%CI) Higher level (Level 1) RegionAfar1.001.00Benishangul Gumuz0.63(0.23, 1.69)0.80(0.18, 3.51)Place of residenceRural1.001.00Urban3.67(1.85,7.26)3.58(1.41, 9.09) Lower-level variables (Level 2) Community engagement interventionsNot attended any one 1.00 1.00Attended only 1–22.70(1.23, 5.92)2.56(1.13, 5.78)Attended 3 or more4.34(1.92, 9.80)3.96(1.70, 9.20)Wealth indexLowest1.001.00Middle class0.80(0.37, 1.78)0.77(0.33, 1.82)Highest0.96(0.39, 2.34)0.98(0.37, 2.60)Age of mother (years)18–241.001.0025–340.43(0.20, 0.92)1.34(0.15, 1.76)>/=350.46(0.14, 1.47)1.37(0.11, 2.22)Educational status of motherUnable to read and write1.001.00Able to read and write4.18(1.49,11.7)4.10(1.39, 12.05)Primary or above4.04(1.89, 8.65)3.52(1.55, 7.98)ParityPara 1 (1st child)1.001.00Para 2–3 (multipara)2.67(1.13, 6.31)2.86(1.17, 7.01)Para 4 or above (Grand multipara)6.51(2.39, 17.77)6.74(2.38, 19.07)Knowledge of mother on immunizationPoor knowledge1.001.00Good knowledge4.73(1.92, 11.63)4.65(1.84, 11.78)Attitude of mother toward immunizationUnfavorable attitude1.001.00Favorable attitude4.66(2.33, 9.34)5.42(2.59, 11.36)ANC follow upNo ANC visit1.001.001–3 ANC visits5.60(2.16, 14.52)5.48(2.05, 14.64)4 or more ANC visits4.66(1.94, 11.20)4.95(1.99, 12.29)Place of deliveryHome birth1.001.00Health facility birth3.64(1.65, 8.04)3.57(1.56, 8.16)At least 1 PNC follow upNo1.001.00Yes7.18(3.64, 14.18)7.95(3.92, 16.09)Random EffectsICC9.8%16.5%21.0%26.5%PCVReference0.551.101.81Model Fit StatisticsLog likelihood−274.2−265.3−171.8−167.7AIC552.3538.5380.2376.5BIC560.5554.9429.4433.9 Mixed effects multilevel analysis of the effect of Community engagement intervention on full child immunization in Afar and Benishangul Gumuz Regions, Ethiopia, April 2024 Discussions Childhood illness had declined almost by half, from the baseline level of 43.3% to 21.9%. However, a substantial burden remained among children under 18 months of age in in the study area with about one-in-five reporting cough, diarrhea, or fever in the two weeks preceding the survey. At both baseline and end-line, fever was the most common symptom, followed by diarrhea, and both symptoms decreased after the intervention. Cough was less prevalent at baseline and showed no reduction. These findings are consistent with studies from Bangladesh [15] and Zambia [16], which reported higher fever in hotter climates and more diarrhea in settings with a limited water supply and poor environmental sanitation, especially in rural areas with minimal facilities. For most sick children (90.4%), care was sought from health facilities. This level is higher than that reported in Zambia (72%) [16] and East Africa (64.4%) [17]. Care seeking increased significantly from baseline to endline. At endline, it was higher among the intervention participants (97.9%) than among the non-participants (89.5%). This shows the effect of enhanced community engagement. Similar findings have been reported in previous studies in Ethiopia [18] and Tanzania [19]. These results support the WHO recommendation [20] for sustained investment in community health interventions to promote maternal engagement through education and outreach. Full immunization coverage among children aged 9–18 months increased from 40.0% before the intervention to 66.4% after the intervention. Coverage was lower among non-participants (40.2%) and higher among those who participated in most interventions (80%), with a statistically significant association in the multilevel analysis, indicating a positive intervention effect. These findings are consistent with studies from developing countries [21–26], including Ethiopia [27, 28], which reported that community engagement interventions significantly improve immunization coverage. The improvements in care-seeking and full immunization are likely driven by several mechanisms of community engagement. Targeted health education and outreach might have increased families’ (caregivers’) knowledge and awareness of child health needs. The interventions also might have promoted positive attitudes and social norms around timely care-seeking and vaccination. Activities such as community dialogues, women’s conferences, and festivals likely enhanced trust in health services and strengthened links between households and health facilities. Together, these factors might have created an enabling environment that encouraged caregivers to seek prompt treatment for sick children and immunizations. Although not statistically significant after adjustment, full immunization coverage showed slight regional variation: 71.0% in Benishangul Gumuz and 61.1% in Afar. This may reflect pastoral mobility, conflict, and environmental challenges limiting service access and record keeping in the Afar region. Similar findings were reported in prior studies in Ethiopia [29] and other countries [30, 31], highlighting the need for context-specific strategies to address logistical barriers in mobile populations. In addition to the intervention, sociodemographic factors, knowledge, attitudes, and maternal health care utilization significantly influenced full immunization. Urban mothers were more than three times more likely to fully immunize their children than rural mothers were, which is consistent with previous similar studies [29, 32] and reflects better urban healthcare access. Parity and maternal education were significantly associated with full immunization. Multipara mothers and those with at least primary education were more likely to fully immunize their children. This aligns with studies from Ethiopia and other countries [33–37]. Multiple mothers are likely to draw on prior experience with immunization, whereas educated mothers can access, read, and understand health information from materials and media, increasing immunization uptake. Mothers with good knowledge and positive attitudes toward child immunization were significantly more likely to fully vaccinate their children. This aligns with previous studies emphasizing the crucial role of maternal knowledge and attitudes in improving immunization and child health outcomes [38–41]. It also highlights the impact of community engagement interventions in shaping maternal knowledge and attitudes, promoting positive health-seeking behaviors. Maternal healthcare utilization strongly supported full immunization. Mothers receiving antenatal care, delivering at health facilities, and using postnatal services had higher full immunization rates. Similar findings have been reported in Ethiopia and elsewhere [42–45]. Community engagement interventions enhanced maternal healthcare use, providing opportunities to deliver information on child immunization and related health services. These findings highlight the programmatic implications for a multifaceted approach to improve treatment seeking for sick children and full immunization coverage. Community engagement plays a critical role, with strategies tailored to geographic location, maternal parity, education, knowledge, attitudes, and healthcare utilization to achieve better child health outcomes. The strengths of this study include its focus on grassroots communities in remote, pastoral, and conflict-affected areas with weak health systems, promoting equity in child healthcare. Context-specific community engagement interventions enhanced participation, ownership, and sustainability. The study addressed top national priorities, sick child treatment and immunization by using advanced statistical methods to control for potential confounders and enhance rigor. Limitations of the study may include potential influence from time trends and routine government programs, which may affect attribution solely to the community engagement intervention. Baseline immunization data were drawn from the 2019 Mini Ethiopian Demographic and Health Survey, conducted two years before the project began. Changes in coverage during this two-year period might have introduced bias in estimating the baseline status. The small number of children not taken to health facilities for treatment limited statistical power, and multivariate analysis was not applied for treatment-seeking outcome. For some children, particularly in Afar region where vaccination cards were unavailable, immunization information were collected from mothers’ reports, which might have introduced some recall bias. These limitations should be considered when the findings and programmatic implications are interpreted. Conclusions This study revealed that community engagement interventions significantly improved treatment seeking and full immunization among children in the Benishangul Gumuz and Afar regions. Additional predictors of full immunization included residence, parity, maternal education, knowledge, attitudes, and maternal healthcare utilization. To further enhance outcomes, programs should prioritize context-specific community engagement, empower caregivers, integrate existing community structures, link immunization with broader maternal health services, and target mobile and underserved populations to ensure timely interventions and identify defaulters. Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1. Supplementary Material 2. Acknowledgements The authors would like to thank the Korean International Cooperation Agency (KOICA) for funding the community engagement interventions and this study through UNICEF Ethiopia. We would also like to sincerely thank the Ethiopian Ministry of Health, Afar and Benishangul Gumuz Regional Health Bureaus, the Woreda Health Offices, service providers, and the communities for their cooperation and support in facilitating the implementation of the community engagement interventions and the conduct of this study. Abbreviations Health Sector Transformation Plan Mini-Ethiopian Demographic and Health survey Maternal, Neonatal and Child Health Neonatal Mortality Rate Quality Improvement Social and Behavior Change Communication Sustainable Development Goal Under-Five Mortality Rate Village Health Leader Women’s Development Group Authors’ contributions All the authors contributed at all stages of the study. AGG, JNC, GTD, TKD, AHF, DHK, MA, MS, DN, TC and BBG participated in the conceptualization, design, data accusation, analysis, interpretation and drafting of the manuscript. All the authors read and approved the final manuscript for submission. Funding This study was funded by the Korean International Cooperation Agency (KOICA) through UNICEF from the project grant, “Optimizing Care for Newborns and Children in Afar and Benishangul Gumuz Regions.” The funder, KOICA, has no specific role in the conceptualization, design, data collection, analysis, decision to publish, or preparation of the manuscript. Data availability The data used for this study are included in the manuscript as texts, tables and graphs. The raw dataset analyzed for the current study is available from the corresponding author upon reasonable request. Declarations Ethics approval and consent to participate This study was conducted in accordance with internationally and nationally recognized research ethics standards, including the Declaration of Helsinki and the Ethiopian National Health Research Ethics Guidelines. The study protocol was initially reviewed and approved by the Institutional Review Board (IRB) of the Afar Public Health and Research Institute (APHRI) on January 20, 2021 with IRB reference number of APHRI/IRB/721/2021. Ethical clearance for the endline evaluation was also subsequently obtained from the same IRB on March 7, 2024 (IRB reference number: APHRI-IRB/5/2024), as well as from the Institutional Research Ethics Review Committee of the Benishangul Gumuz Regional Health Bureau on March 5, 2024 (IRB reference number: 1765/2024/RHB-01). All participants were informed about the study objectives and the potential benefits and risks of participation, and informed oral consent was obtained prior to data collection. Participation was entirely voluntary, and respondents retained the right to decline or withdraw at any time without consequence. To ensure confidentiality, data were collected anonymously and reported in aggregate form. 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. Data Availability Statement The data used for this study are included in the manuscript as texts, tables and graphs. The raw dataset analyzed for the current study is available from the corresponding author upon reasonable request.