Propensity score matching analysis of the effect of four or more antenatal care visits on basic childhood immunization in Ethiopia

Abstract

Globally, immunization prevents 3.5 to 5 million deaths annually from common childhood diseases. Several observational studies evidenced that childhood immunization was linked to mothers’ Antenatal Care (ANC) visits. However, the effect of the number of ANC visits on childhood immunization in Ethiopia using propensity score matching has not been investigated. Therefore, this study aimed to assess the effect of four or more ANC visits on childhood immunization in Ethiopia using propensity score matching analysis. This secondary data analysis used the 2019 Ethiopian Mini Demographic and Health Survey, which included mothers whose most recent (index) birth occurred within the five years preceding the survey, with children aged 12–23 months. A propensity score matching (PSM) analysis was performed using a probit model with the psmatch2 command in STATA to estimate the average treatment effect on the treated (ATT), untreated (ATU), and the population (ATE). Common support assumptions were checked, and sensitivity analysis was done by Mantel-Haenszel bounds. A total of 972 mothers were included in this study. The overall basic immunization coverage was 39.72% (95% CI: 36.67, 42.82), and it was significantly higher among those mothers who received ANC 4 + visits (56.93%) when compared to mothers with < 4 ANC visits (27.09%). In the PSM analysis, the ATT values in the treated and control groups were 59% and 48%, respectively, indicating that the basic immunization coverage was increased by 11% (95% CI: 3.00, 17.00) because of the number of ANC visits. The ATU in the intervention and control groups were 27% and 45%, respectively, indicating that those mothers who had < 4 ANC visits, the chance of getting their children immunized would have increased by 18%, if they had made ANC 4 + visits. The final ATE estimate was 13% (95% CI: 5.00, 18.00) among the study participants. After matching, there was no significant difference in baseline characteristics between the treated and control groups (p-value > 0.05), which indicates the quality of matching was satisfactory. Enhancing mothers to have four or more ANC visits could effectively increase the uptake of basic childhood immunization in Ethiopia. Therefore, we recommend policy makers to enhance interventions targeting improving number of ANC visits to increase childhood immunization in Ethiopia.

Introduction

Childhood immunization is one of the most cost-effective public health interventions for improving child survival and the prevention of morbidity and mortality associated with common childhood illnesses globally^1,2. The global number of deaths among under-five children significantly decreased from 12.5 million in 1990 to 4.9 million in 2022 after the introduction of the Expanded Program on Immunization (EPI)³. Currently, immunization prevents 3.5 to 5 million deaths every year from diseases like diphtheria, tetanus, pertussis, influenza, and measles⁴. Although the child mortality rate significantly declined by more than half from the 2000 report, further efforts are needed to accelerate progress towards the Sustainable Development Goal (SDG) target of 25 child deaths per 1000 live births by 2030⁵. In Ethiopia, the EPI was launched in 1980 with an ambitious goal of achieving universal childhood immunization coverage for children less than 2 years of age by 1990⁶. The program has been freely provided by the public health sectors in collaboration with non-governmental organizations and donors in all regions and districts through facility-based and long-term outreach service strategies to achieve the national targets⁶. However, the 2019 Ethiopian Mini Demographic and Health Survey (EMDHS) report revealed that only 44% of children received all recommended vaccinations⁷.

The effectiveness of childhood immunization depends on factors like family and society characteristics, parental attitudes and knowledge, and immunization-related communication and information⁸. Several studies evidenced that childhood immunization coverage was associated with their mothers’ Antenatal Care (ANC) visits^8–14, and tried to link the utilization of prenatal care services with child immunization^9,10. However, the ANC utilization itself may not directly influence full childhood immunization coverage rather the mothers may differ across a set of unobserved characteristics (such as beliefs and attitudes) and known factors of socioeconomic and demographic states (such as residence, education, wealth status, or access to health information) to influence full childhood immunization. Therefore, traditionally, in statistical analysis to control for such confounding associations between ANC utilization and childhood immunization, regression analysis has been performed. However, since the regression analysis only performs adjustments for the observable variables, bias still exists. For instance, even when such factors are controlled within the regression model, the variation in the distribution of factors influencing ANC visits among mothers who had less than four visits and mothers who had four or more visits could bias the effects of ANC visits on the subsequent full childhood immunization^15,16. In addition, due to unobserved variables that could create bias, mothers within the control group are more susceptible to incomplete immunization than mothers within the treatment group.

Therefore, propensity score matching (PSM) is an appropriate approach to estimating the actual impact of ANC visits on full childhood immunization. PSM is a methodological approach that aimed to remove bias by matching treated (ANC ≥ 4 visits) and untreated (ANC < 4 visits) mothers with similar conditional probability to receive the treatment (ANC ≥ 4 visits)¹⁷. In this study, we matched the mothers with less than four ANC visits to mothers with four or more ANC visits with similar propensity score values for ANC visits. Then, it can be reasoned that any difference in full immunization is attributed to ANC visits only. However, as to our search of the literature, there was no study done in Ethiopia that used propensity scores to assess the impact of ANC 4 + visits on full childhood immunization. Hence, this study aimed to assess the impact of four or more ANC visits on childhood immunization by using propensity score matching analysis. The findings of the study will help to develop the necessary strategies and interventions to help improve childhood immunization and, in return, to reduce child mortality for the achievement of SDGs.

Methods

Study design and setting

This study used secondary data from the Ethiopian Mini Demographic and Health Survey (EMDHS) of 2019. EMDHS is a nationally representative cross-sectional household survey and was conducted from March 21 to June 28, 2019. The survey was conducted to provide updated information on selected maternal, child, and neonatal health outcomes.

Data source and population

The analysis was based on secondary data from the 2019 mini EDHS. The EMDHS Uses a two-stage stratified cluster sampling technique, the regions was divided into urban and rural areas. Accordingly, 21 sampling strata were created, and an equal allocation of samples was done in each region, wherein 25 Enumeration areas /EAs were selected from eight regions, whereas, 35 EAs were selected from three larger regions. In the first stage, 305 clusters (93 urban and 212 rural) were selected with probability proportional to EAs size and with independent selection in each sampling stratum. In the second stage, a fixed number of 30 households per cluster were selected. Finally, women aged 15–49 in 9,150 (6360 rural and 2,790 urban) households from 305 clusters were selected. Detail information about the sampling procedure is available¹⁸. Data on childhood immunization (12–23 months) and the history of mothers’ antenatal visits were considered for 972 children (Fig. 1). Figure 1 represents a flowchart that illustrates how the final analytical sample (n = 972) was derived from the initial group of eligible women interviewed (n = 8885).

Fig. 1.

Final sample size and schematic presentation of how the study sample size was selected.

Variables of the study

Outcome variable

The outcome variable was the childhood immunization status within 12–23 months of age. Vaccination data was obtained by enquiring mothers who had given birth in the past 5 years preceding the survey whether they have a vaccination card for their child, and if available, they were asked to show that card to the data collector for obtaining the information about the vaccination such as date, month, number of vaccine doses given, etc. If the child’s vaccination card was unavailable or the vaccine was not recorded on the card, the respondents were asked to recall whether their child had received the vaccine.

WHO defines “basic immunization” as a child that has received one dose of BCG, three doses of pneumococcal conjugate (PCV), pentavalent, oral polio vaccines (OPV); two doses of Rota virus and one dose of measles vaccine. We recoded each variable as “0” for children who did not receive the recommended doses and “1” for those who did, based on the mothers’ reports and the information in the child’s vaccination card^19,20. The final model was executed with the variable coded as a binary outcome by categorizing children as either “immunized” or “not immunized.”

Treatment variable

Data on the number of ANC visits made during the last pregnancy was obtained using the following question during the survey: “How many times did you receive antenatal care during pregnancy?” Mothers who have attended ≥ 4 ANC visits were considered to have adequate ANC visit coverage. The final model was employed with the variable coded, as “less than 4 ANC visits” or “≥4 ANC visits.”

Matching variables

Variables which can affect the treatment variable ANC utilization and childhood immunization were selected based on the available literature^{8,11,12,14,21–23}. The selected variables were the mother’s age, sex of the child, residence, maternal educational status, religion, wealth index, place of delivery, birth order, family size and region.

Data management and analysis

In observational studies, researchers cannot randomly assign study participants to either of the groups (treatment or control) the imbalance of the observed variables introduces bias and affects the exposure’s causal effect. Therefore, propensity score matching (PSM) is an appropriate approach to estimating the actual impact of four or more ANC on Full immunization.

PSM is a methodological technique that aimed to remove bias by matching treated (four or more ANC visits) and untreated (less than four ANC visits) mothers with similar conditional probability to receive the treatment (four or more ANC visits). In this study, we matched the mothers with less than ANC visits to mothers with four or more ANC visit with similar propensity score values for four or more ANC visits. Then, it can be reasoned that any difference in full childhood immunization is attributed to the number of ANC visits only.

In this study, we estimated the ATT of having 4 + ANC visits on childhood immunization. The probit regression model was used to estimate the propensity score and kernel as a matching algorithm^12,24. At the first step of the analysis, basic socio-demographic and obstetric characteristics of the participants were presented as percentages. Following descriptive analysis, a chi-square (χ2) test was performed to check for the significant associations between the explanatory variables and childhood immunization. Several iterations of the propensity score (PS) estimate were done by recoding variables until the balancing property was fulfilled. We have performed a common support option to increase matching quality. Multiple matching algorithms were tested, and the one resulting in the lowest standard bias (B) was selected. Quality of matching was assessed through standardized bias, pseudo R-squared, and likelihood ratio (LR) tests. Percentage bias and percentage reduction in bias for each covariate were calculated to assess the effectiveness of matching.

Following matching, confidence intervals (95% CI) were computed to accompany all effect estimates to enhance the interpretability and robustness of the findings. Bootstrapping with 1000 replications was employed to derive robust estimates of the Average Treatment Effect on the Treated (ATT) by taking into account sampling variability and potential biases in the treatment effect estimates.

Finally, a sensitivity analysis was performed using the Mantel-Haenszel bounds approach to assess the robustness of treatment effects estimated using the unconfoundedness assumption (also known as selection on observables or conditional independence). We checked whether the final PSM estimates are sensitive to the hidden bias²⁵.

As the outcome variable is binary, we have used the Mantel–Haenszel (MH) test statistic²⁶. The gamma coefficient (Γ) was the factor by which hidden bias or unobserved confounders would affect the assignment into intervention for a treated participant compared to an untreated participant with matching covariates. The range of gamma specified was between 1 and 2 with 0.05 increments.

Software application

Data were analyzed by using STATA version 17 (StataCorp, College Station, Texas 77,845 USA) statistical software which is available at https://www.stata.com. Propensity scores were estimated using the pscore command with a probit regression model, and treatment effects including the Average Treatment Effect on the Treated (ATT), the Untreated (ATU), and the overall population (ATE) were estimated using the psmatch2 command with kernel matching. The quality of matching was evaluated using the pstest command, based on standardized bias, pseudo R², t-tests, and likelihood ratio tests. To enhance robustness, bootstrapping with 1,000 replications was applied. Finally, a sensitivity analysis to assess the potential influence of hidden bias was conducted using the mhbounds command^26,27, specifying gamma (Γ) values from 1.0 to 2.0 in 0.05 increments.

Result

Descriptive characteristics of the study participants

A total of 972 participants were included in this study. The analysis was based on the 561 (57.72%) mothers who did have less than four ANC visits and 411 (42.28%) mothers who received ≥ 4 ANC visits for their youngest children.

The descriptive analysis result showed that the overall basic vaccination coverage in the study participants was 39.72% (95% CI: 36.67, 42.82) and this coverage was significantly higher among those mothers who received adequate ANC visits (56.93%) when compared to mothers with less than four ANC visits (27.09%). All the baseline characteristics (except for the sex child) showed significant differences (p < 0.05) across the number of ANC visits (< 4 visit vs. ≥4 visit) before matching (Table 1).

Table 1.

Baseline socio-demographic and obstetric characteristics of participants by antenatal care (ANC) before matching.

Variables	Before matching
	Less than four ANC visits	Four or more visits	P value
Age of the mother
15_24 years	177 (60.62)	115 (39.38)	0.037
25_34 years	283 (55.83)	225 (44.17)
≥ 35 years	101 (59.41)	71 (40.59)
Residence
Urban	89 (34.77)	170 (65.23)	< 0.001
Rural	472 (66.20)	241 (33.80)
Religion
Orthodox	139 (44.41)	177 (55.59)	< 0.001
Muslim	296 (64.63)	162 (35.37)
Protestant	105 (61.05)	67 (38.95)
Others	21 (80.77)	5 (19.23)
Maternal education
No formal education	339 (71.37)	136 (28.63)	< 0.001
Primary	175 (53.19)	154 (46.81)
Secondary/higher	47 (28.48)	121 (71.52)
Household wealth index
Poor	328 (73.38)	119 (26.62)	< 0.001
Middle	80 (60.15)	53 (39.85)
Rich	153 (39.33)	239 (60.67)
Place of delivery
Home	341 (82.77)	71 (17.23)	< 0.001
Health facility	220 (39.29)	340 (60.71)
Birth order
First	108 (48.87)	113 (51.13)	0.001
Second	115 (54.76)	95 (45.24)
Third or above	338 (62.48)	203 (37.52)
Sex of the child
Male	285 (58.40)	203 (41.60)	0.063
Female	276 (57.02)	208 (42.98)
Family size
Less than 4	73 (50)	73 (50)	0.050
4–6	292 (58.40)	209 (41.60)
Greater than 6	196 (60.31)	129 (39.69)
Regions
Large central	234 (55.32)	189 (44.68)	< 0.001
Small peripheral	240 (71.01)	98 (28.99)
Metropolitan	87 (41.23)	124 (58.77)

Other: Catholic, traditional, and others category of religion.

Estimation of propensity score

The mean propensity score was 0.43, with a standard deviation (SD) of 0.25, indicating variability between the intervention and control groups. The number of blocks was 5. Figure 2 shows the balance of the propensity score distributions between the treatment and control groups. The bars above the line for mothers in the treated group and ones below the line show propensity scores for mothers in the control group. The figure indicates adequate overlap in the propensity score distributions between the two groups, which included a total of 969 observations. Mothers in both intervention and control groups with PS outside the region of common support observations were discarded in the graph and in the final analysis. The balancing property was satisfied and the region of common support between the intervention and control group was high ranging from 0.06 to 0.92 of the PS.

Fig. 2.

Propensity score histogram by treatment status (number of ANC visits).

Impact of four or more ANC visits on childhood immunization

We used Propensity Score Matching (PSM) to reduce selection bias and ensure comparability between groups with and without ≥ 4 ANC visits. This allowed us to estimate the effect on childhood immunization more reliably.

The unmatched estimate indicated that mothers who attended four or more ANC visits had a 29% higher likelihood of immunizing their children compared to those who attended fewer than four visits. The kernel matching method had the best quality of matching, and it was used to estimate the ATE, ATT, and ATU. The calculated ATT values in the treatment and control groups were 0.59 and 0.48, respectively, indicating that the immunization coverage was increased by 11% (95% CI: 3.00, 17.00, p < 0.01) because of four or more ANC visits. Similarly, the calculated ATU values in the treatment and control groups were 0.27 and 0.45, respectively. This means that if mothers who had less than four ANC visits had made four or more visits, the chance of getting their children immunized would have increased by 18%. To obtain a robust estimate of standard error, bootstrapping with 1000 replicates was employed. The final ATE value was found to be 0.13 (95% CI: 0.05, 0.18, p < 0.001) among the total study participants, highlighting that four or more ANC visits may have contributed to increasing childhood immunization across both treatment and control groups (Table 2).

Table 2.

The impact of four or more ANC visits on basic childhood immunization using PSM method in Ethiopia obtained from Ethiopian mini DHS 2019.

Impact of ANC visit on childhood immunization	Treated	Controls	Difference	SE	t-stat	Average treatment effect with cluster-robust standard errors
						Treatment effect	Robust SE	Z	p > |z|	95% CI
Unmatched	0.56	0.27	0.29	0.03	9.84	–	–	–	–	–
ATT	0.59	0.48	0.11	0.04	1.88	0.10	0.04	2.85	0.004	0.03, 0.17
ATU	0.27	0.45	0.18			–	–	–	–	–
ATE			0.13			0.12	0.03	3.66	0.000	0.05, 0.18

ATT: Average treatment effect on the treated, ATU: Average treatment effect on the untreated, ATE: Average treatment effect on the population.

Quality of matching

Common support

Of the 972 observations, only 3 observations were discarded due to off-support (Table 3). We plot the distributions of propensity score and the distribution is almost similar for both treatment and control groups after matching on PS (Fig. 3). The Fig. 3 demonstrates the effectiveness of Propensity Score Matching (PSM) in balancing covariates between the treatment and control groups. Before matching, significant differences (standardized biases) likely existed across variables. After matching, the plot reveals that biases for all covariates have been substantially reduced, with most clustered near 0% standardized bias, indicating successful balance. The presence of significant overlap between the characteristics of the treated and control groups proves the validity of the common support assumption (Fig. 3). The common support table (Table 3) displays the distribution of propensity scores for both the treated (less than 4 ANC visits) and control (≥ 4 ANC visits) groups. It identifies the region where the scores of both groups overlap, ensuring that individuals from both groups with similar covariate profiles are available for matching. Observations falling outside the common support region were excluded from the final analysis to improve the comparability between groups and reduce bias (Table 3).

Table 3.

Common support summary table for propensity score matching analysis of assessing the effect of four or more antenatal care visits (ANC) on basic childhood immunization in ethiopia.

Assigned treatment	Off support	On support	Total
Untreated	0	561	561
treated	3	408	411
Total	3	969	972

Fig. 3.

Quality of matching by propensity score distribution.

Balancing test

The difference between the unmatched and matched pairs was assessed by a t-test and the significance level of the test. Even though there was a significant mean difference before matching across the covariates, there was no significant mean difference across almost all the covariates after matching (Table 4). This indicated that the treated and control group are adequately balanced for all the variables included in the model.

Table 4.

Performance of the propensity score matching: quality measurements.

Variable	Sample	Mean		% bias	% Bias reduction	t-statistic	P-value
		Treated	Control
Maternal age	Unmatched	1.88	1.86	3.4		0.51	0.607
	Matched	1.88	1.84	1.3	62.2	0.18	0.855
Residence	Unmatched	1.59	1.84	− 57.8		− 9.09	< 0.001
	Matched	1.59	1.58	1.6	97.2	0.20	0.839
Sex of child	Unmatched	1.51	1.49	3.1		0.47	0.637
	Matched	1.51	1.51	− 0.4	86.0	− 0.06	0.951
Religion	Unmatched	1.76	2.01	− 32.9		− 5.06	< 0.001
	Matched	1.76	1.81	− 6.8	79.4	− 0.97	0.331
Maternal education	Unmatched	0.96	0.48	66.1		10.31	< 0.001
	Matched	0.95	0.96	− 1.1	98.3	− 0.15	0.882
Wealth index	Unmatched	2.28	1.68	68		10.46	< 0.001
	Matched	2.28	2.27	0.1	99.9	0.01	0.990
Place of delivery	Unmatched	0.83	0.39	99.2		14.95	< 0.001
	Matched	0.83	0.81	1.8	98.2	0.30	0.766
Birth order	Unmatched	2.21	2.41	− 23.9		− 3.70	< 0.001
	Matched	2.21	2.26	− 6.6	72.2	− 0.91	0.361
Family size	Unmatched	2.13	2.22	− 12.9		− 1.99	0.047
	Matched	2.13	2.14	− 1.3	90.2	− 0.18	0.858
Region	Unmatched	1.84	1.73	14		2.18	0.030
	Matched	1.84	1.83	2.4	82.9	0.33	0.745
Matching approach Kernel matching.
Mean bias Unmatched = 38.1; Matched = 2.3.
Median bias Unmatched = 28.4; Matched = 1.4.
Pseudo R2 Unmatched = 0.198; Matched = 0.002.
LR χ 2 Unmatched = 261.83; Matched = 2.81.
P-value Unmatched = < 0.001; Matched = 0.986.
Standardized bias Unmatched = 116.1; Matched = 11.7.
R Unmatched = 1.01 ; Matched = 1.00.
Percentage of variance Unmatched 33; matched 0.

Standardized bias

The pstest command indicates that the mean and median bias values were significantly reduced after matching between the treatment and control groups. The mean absolute bias decreased from 38.1% in the unmatched sample to 2.3% after matching between the treated and control groups, which is less than the threshold (5%), showing the quality of matching in the model. The standardized biases for all the covariates after matching were < 5%, which reduced from 28.4% in the unmatched sample to 1.4% after matching (Table 4).

Significance of the model

The model significance was evaluated by LR and pseudo-R² tests. The LR-test was insignificant (p = 0.986), and the pseudo-R² was 0.002, indicating that there was no systematic difference in covariate distribution between the treated and control groups (Table 4).

Sensitivity analysis

The sensitivity analysis was performed using the Mantel Haenszel statistic to estimate the extent of unobservable covariates biases (hidden biases) on our inferences about the effect of 4 + ANC visits. The ATE of adequate ANC visits on childhood immunization remained statistically significant at all the specific values of gamma (p_mh < 0.001). This indicates that the positive treatment effect observed for childhood immunization was insensitive to the hidden bias within the determined range by gamma (Γ) (Table 5).

Table 5.

Sensitivity analysis using Mantel-Haenszel bounds.

Gamma (Γ)	Test statistics		Significance level
	Over-estimation (Q_mh+)	(Under-estimation) Q_mh−	(Over-estimation) p_mh+	(Under-estimation) p_mh−
1	9.20956	9.20956	< 0.001	< 0.001
1.05	8.84108	9.5892	< 0.001	< 0.001
1.1	8.48668	9.94826	< 0.001	< 0.001
1.15	8.1493	10.2929	< 0.001	< 0.001
1.2	7.8274	10.6243	< 0.001	< 0.001
1.25	7.51962	10.9436	< 0.001	< 0.001
1.3	7.22477	11.2517	< 0.001	< 0.001
1.35	6.94181	11.5495	< 0.001	< 0.001
1.4	6.66982	11.8376	< 0.001	< 0.001
1.45	6.40798	12.1168	< 0.001	< 0.001
1.5	6.15555	12.3878	< 0.001	< 0.001
1.55	5.91188	12.6509	< 0.001	< 0.001
1.6	5.67637	12.9067	< 0.001	< 0.001
1.65	5.4485	13.1556	< 0.001	< 0.001
1.7	5.22779	13.3982	< 0.001	< 0.001
1.75	5.01378	13.6346	< 0.001	< 0.001
1.8	4.80609	13.8653	< 0.001	< 0.001
1.85	4.60434	14.0905	< 0.001	< 0.001
1.9	4.40821	14.3106	< 0.001	< 0.001
1.95	4.21738	14.5258	< 0.001	< 0.001
2	4.03158	14.7364	< 0.001	< 0.001

Discussion

In public health, healthcare policymakers attempted to assess how the health interventions among the treated populations have changed. This study was aimed to investigate the impact of four or more ANC visits on basic childhood immunization using propensity score matching analysis. PSM reduces selection bias and provides an alternative for measuring treatment effects in observational/non-experimental studies when randomized clinical trials are not possible or unethical.

This study found that from 972 participants, 42.28% of mothers received ≥ 4 ANC visits for their youngest children. The overall basic immunization coverage in the study participants was 39.72% (95% CI: 36.67, 42.82) and this immunization coverage was significantly higher among those mothers who received adequate ANC visits (56.93%) when compared to mothers with less than four ANC visits (27.09%). This study finding is consistent with a study conducted with 2016 Ethiopian DHS, 38.3%¹⁴. However, the finding of this study is lower than a meta-analysis conducted in Ethiopia, 65%⁸. The difference may be due to the methodology used in this study, which relies on survey data, whereas the higher magnitude in previous findings may result from pooled data extracted from multiple published studies^28,29. In addition, the finding was lower than studies conducted in Tanzania, 52.5%³⁰ and Ghana, 56.5%³¹. The observed difference may be attributed to sociocultural factors, policy variations, and differences in health systems across countries³².

In unmatched analysis, women who made ≥ 4 ANC visits had 29% more chance of immunizing their children, compared to women who did not have adequate ANC visits. To evaluate the exclusive impact of the intervention, PSM was performed, and it was effective in improving the balance for most of the covariates. The basic immunization coverage was increased by 11% because of adequate ANC visits, and mothers who did not have adequate ANC visits, if they had made ≥ 4 visits, the probability of getting their children immunized would have increased by 18%. This study finding was consistent with a study conducted in India that reported immunization coverage was significantly higher among those mothers who received ≥ 4 ANC visits (71.8%) when compared to women with < 4 ANC visits (39.9%), and the immunization coverage was increased by 24% due to the number of ANC visits¹². A similar study in India also found that mothers who had at least two ANC visits had a 43% higher chance to immunize their children compared to mothers who did not make any visits³³.

In this study, we used the kernel method to match the treated and control groups. The model indicates that when treated mothers were compared with their matched counterparts, who were similar in every observed pre-existing characteristic except for ANC visits, it demonstrated the helpfulness of four or more ANC visits for subsequent child immunization. This also implicates that antenatal care follow ups are the key benchmark platforms for educating and counseling pregnant women on the benefits of child immunization³⁴. This positive interaction between the number of ANC visits and immunization services can be attributed to the fact that mothers who exposed to health facilities for ANC follow-up become more informed about the benefits and schedule of vaccination and encouraged to seek healthcare^11,13,35. Women have different benefits through ANC visits, such as counseling on healthy lifestyles, institutional delivery, and education about the benefits of childhood vaccination. In addition, consistent ANC visits establish good relationships between the pregnant woman and her health care provider³⁶. Women’s interactions with healthcare providers can build trust, strengthening their relationship by encouraging the development of personal disclosure, and it may affect women’s health care-seeking behavior^37,38.

In developing countries like Ethiopia, although the first ANC visit is encouraging (74%), a significant gap exists in the coverage of full childhood immunization (44%)⁷. Women have raised many reasons for not taking the benefit of free immunization, such as lack of awareness about its benefits, while others refuse it on ethical or religious issues³⁹. Reporting no trust in vaccination, fear of side effects, and unfamiliar place and time of vaccination could be the other reasons for non-immunization^21,39. These barriers can be overcome by enhancing interactions between mothers and healthcare workers during antenatal care (ANC) visits. Women who regularly visit health centers may develop trust in the healthcare system, gain satisfaction, and become more informed about the benefits of vaccination, which could increase their likelihood of returning for their children’s vaccinations³³. Furthermore, it is important to enhance the communication, education, and information-sharing skills of health workers to improve childhood vaccination rates. Family support and engaging husbands during antenatal care visits is also essential for informing and motivating them, so they feel confident and assured in supporting their children’s vaccination^15,40. Since people adopt new behaviors not only through personal experiences but also by observing the actions of the majority within their social network⁴¹. Participating communities, especially through group-based antenatal care (ANC), can help identify local barriers to ANC services and encourage more frequent visits¹².

Proximity to health facilities is one of the major determinants for ANC utilization. Several studies have reported that increased distance to health facilities is associated with fewer ANC visits and delayed initiation of care^42–44. Women living closer to health facilities are more likely to attend ANC visits due to reduced travel time, transportation costs, and physical effort required to access services. In contrast, women residing in remote or rural areas often face significant geographic and infrastructural barriers, which discourage timely and frequent ANC attendance^45,46. Hence, reducing the physical distance to health facilities through service expansion, outreach, or transportation support is crucial to improving frequent ANC visits.

Limitations

Although we included a large number of covariates to estimate the propensity score, there might still be some remaining residual confounding/bias due to unobserved variables, which could lead to overestimating the effects of the treatment on outcome variables. Additionally, since we used secondary data, cultural factors and differences in health system infrastructure across regions could not be addressed. Furthermore, the respondents were asked about events that occurred within the five years prior to the survey, which may lead to recall bias.

Conclusion

Our study confirmed that ANC 4 + visits significantly increased the likelihood of basic childhood immunization in Ethiopia. Health policymakers and healthcare providers should focus on the pregnant women to attend adequate ANC visits, as our study indicates that ANC 4 + visits significantly improved the chance of subsequent childhood vaccination.

Abbreviations

ATE

Average treatment effect

ATT

Average treatment effect on the treated

ATU

Average treatment effect on the untreated

ANC

Antenatal care

EMDHS

Ethiopian mini Demographic and health Survey

EPI

Expanded program immunization

PSM

Propensity score matching

Author contributions

M.G.T. conceived the idea. M.G.T. led the methodology, extracted the data, conducted analysis, and wrote the original draft of the manuscript, K.A.D., W.D.N., M.J., G.T., M.M.T., T.Z.T. and A.H. were involved in the methodology, statistical analysis, and interpretation. All of the authors read and approved the final manuscript.

Funding

No funding was received.

Data availability

Data used in our study are publicly available upon request from the DHS program website. (https://dhsprogram.com/).

Declarations

Competing interests

The authors declare no competing interests.

Ethical approval and consent to participate

The data were accessed from the DHS website https://dhsprogram.com/data/available-datasets.cfm after getting registered and permission. The retrieved data were used for this registered research only. The data were kept confidential and no identifier was made to identify any household or individual respondent.