Association between the continuum of maternal healthcare services and child immunisation in East Africa: a propensity score matching analysis Authors: Alemayehu Kasu Gebrehana, Angwach Abrham Asnake, Beminate Lemma Seifu, Bezawit Melak Fente, Meklit Melaku Bezie, Zufan Alamrie Asmare, Sintayehu Simie Tsega, Yohannes Mekuria Negussie, Hiwot Altaye Asebe, Mamaru Melkam Published: 2025 ABSTRACT AbstractObjectiveTo assess the association between the maternal continuum of healthcare and child immunisation in East Africa using propensity score matching (PSM).DesignCross-sectional study using Demographic and Health Survey data.SettingThis study was conducted in East African countries.ParticipantsThis study included a weighted sample of 13 488 women with children aged 12–23 months.Outcome measureChild immunisation was the outcome variable of this study.ResultsThe PSM estimates indicate that the average treatment effect on the treated for complete child immunisation was 0.0583, meaning that children of mothers who received a complete maternal continuum of care had a 5.83% higher probability of being fully immunised compared with children of mothers with incomplete care. Expressed relative to the treated group’s mean, this corresponds to a 7.48% increase. Additionally, our results indicated that the population average treatment effect was 0.0629. This means that, on average, a complete continuum of maternal healthcare increases the probability of full child immunisation by approximately 6.29% across the entire population.ConclusionThe study highlights that children whose mothers receive comprehensive maternal healthcare are more likely to complete their childhood immunisations. This finding underscores the need to integrate immunisation services into maternal healthcare programmes to enhance vaccination coverage and promote better child health. To maximise this connection, improving access to maternal healthcare, especially in underserved regions, is crucial, along with ensuring that immunisation is a regular part of maternal care. STRENGTHS AND LIMITATIONS OF THIS STUDY This study used large, nationally representative Demographic and Health Survey data from 11 East African countries, enhancing the generalisability of the findings. This study used large, nationally representative Demographic and Health Survey data from 11 East African countries, enhancing the generalisability of the findings. Inclusion of key socio-demographic and household variables as confounders strengthened the validity of the matching process. Inclusion of key socio-demographic and household variables as confounders strengthened the validity of the matching process. Propensity score matching can only account for measured confounders, so residual confounding from unmeasured variables cannot be fully excluded. Propensity score matching can only account for measured confounders, so residual confounding from unmeasured variables cannot be fully excluded. Matching criteria may reduce the sample size and statistical power, potentially limiting the precision of the estimates. Matching criteria may reduce the sample size and statistical power, potentially limiting the precision of the estimates. The dichotomisation of both the treatment variable and the outcome inevitably reduces the complexity of these constructs and may obscure meaningful variation across intermediate categories. The dichotomisation of both the treatment variable and the outcome inevitably reduces the complexity of these constructs and may obscure meaningful variation across intermediate categories. Introduction From 2000 to 2017, maternal deaths worldwide significantly decreased, with global estimates for 2017 showing 295 000 maternal deaths, a 35% reduction from 2000. Sub-Saharan Africa (SSA) and Southern Asia accounted for 86% of these deaths, with SSA alone responsible for about 66% and Southern Asia nearly 20%.1 Similarly, significant progress has been made in child survival. The total number of deaths among children and young adolescents decreased by 56%, from 14.2 million in 1990 to 6.2 million in 2018.2 Despite a notable decrease in maternal and child mortality, approximately 800 women still die every day from preventable causes related to pregnancy and childbirth. This continues to pose a challenge to achieving Sustainable Development Goal targets 3.1 (reducing the global maternal mortality ratio to below 70 per 100 000 live births) and 3.2 (reducing neonatal mortality to no more than 12 per 1000 live births and under-5 mortality to no more than 25 per 1000 live births).35 The continuum of care during pregnancy, childbirth and the postnatal period refers to a comprehensive approach aimed at improving maternal and neonatal health while also preventing mortality and morbidity among mothers and infants.6 7 The prevalence of completing the maternal continuum of care varies across Africa. A systematic review and meta-analysis conducted in the region found that the pooled prevalence of completing the maternal continuum of care was 20.9%.8 The prevalence ranged from 6.9% in Ethiopia9 to 50.4% in Egypt.10 Studies identified that factors such as women’s age,11 women’s educational status,7 11 12 place of residence,8 12 13 wealth index,7 8 11 13 media exposure,81113 parity,8 plan of pregnancy,8 number of children11 and timing of antenatal care (ANC) initiation12 13 were significantly associated with maternal continuum of care. Additionally, a study also identified that the continuum of maternal healthcare was found to be associated with an increase in the immunisation level of children, with about 76% of the children whose mothers had complete continuum of care being fully immunised.7 Similarly, higher levels of women’s education,1416 higher levels of father’s education,14 urban residence,1618 being married,16 delivered in a health facility,141618 ANC follow-up during pregnancy,17 18 being in the rich household wealth index category,17 parity17 and maternal healthcare utilisation14 were identified as significant and positive predictors of complete childhood immunisation. In contrast, a higher family size15 and occupation as a farmer15 were identified as negative predictors of complete childhood immunisation. Numerous studies have been conducted in the past to examine factors that influence the continuum of maternal healthcare and how this continuum of care might affect childhood immunisation. These studies often use statistical methods, such as classical logistic regression, which help identify associations between variables (in this case, the maternal continuum of healthcare and childhood immunisation). This study, however, takes a different approach by employing PSM analysis, a more advanced statistical technique. This method aims to control for confounding factors, variables that might influence both the maternal continuum of care and childhood immunisation. It does so by matching participants with similar characteristics but differing in the key factor (whether they received the continuum of maternal healthcare). This means that the study seeks to examine whether there is an association between the continuum of maternal healthcare and childhood immunisation while controlling for other confounding variables, rather than implying a direct causal relationship between the two factors, due to the cross-sectional nature of our data. Moreover, using PSM to examine the association between the continuum of maternal healthcare utilisation and childhood immunisation can help planners, policymakers and health managers make more informed decisions. For planners, it aids in identifying key maternal health services that drive higher immunisation rates, ensuring efficient resource allocation. Policymakers benefit by gaining evidence of the link between maternal healthcare and immunisation, enabling the creation of targeted, equitable policies. Health managers can assess and refine maternal health programmes to improve immunisation outcomes. Overall, PSM offers a reliable method for understanding and enhancing the relationship between maternal healthcare and childhood immunisation. Methods Study design, setting, data source, population and sampling procedures The Demographic and Health Survey (DHS) programme uses a cross-sectional study design. This study used the recent DHS dataset from 2015 to 2023, which covers 11 East African countries: Burundi, Ethiopia, Kenya, Madagascar, Malawi, Mozambique, Rwanda, Tanzania, Uganda, Zambia and Zimbabwe. The DHS is conducted by the US Agency for International Development through its DHS programme. The survey gathers and analyses data on population, health and nutrition in developing countries, covering various topics such as fertility, family planning, maternal and child health, gender, HIV/AIDS and other health-related issues. The data were downloaded from the DHS website (https://dhsprogram.com/)on request. The source population included all children aged 12–23 months and their mothers. The study population consisted of children aged 12–23 months and their mothers from selected enumeration areas across all 11 East African countries. Datasets from individual countries were combined to assess the association between maternal continuum of care and childhood immunisation in these countries. The DHS used a two-stage stratified sampling method to select study participants. Our study included a weighted sample of 13 488 women with children aged 12–23 months. Outcome variable The outcome variable in this study was child immunisation, measured by whether a child in the age group of 12–23 months had received all basic childhood vaccinations. This variable was dichotomised into a binary variable (0, incomplete childhood immunisation; 1, complete childhood immunisation). Children who received one dose of the Bacillus Calmette–Guérin vaccine; three doses of the diphtheria, pertussis and tetanus vaccine; three doses of the polio vaccine (excluding the polio vaccine given at birth); and one dose of the measles vaccine were considered to have completed childhood immunisation.7 19 20 Treatment variable The treatment variable in this study was the maternal continuum of healthcare, defined using three maternal health indicators: ANC, institutional delivery and postnatal care (PNC). This variable was categorised into two groups: 1, complete continuum of care (when a woman reported having at least four ANC contacts during her most recent pregnancy, delivering in a health facility and receiving at least one PNC check-up within 6 weeks after childbirth), and 0, incomplete continuum of care (when women reported fewer than four ANC contacts, delivering outside a health facility or having no PNC check-up).101221 22 In addition to categorising maternal continuum of care as complete or incomplete, we further examined the components of the continuum to understand where most women fall short. The treatment variable was constructed from three key services: attending at least four ANC visits, delivering in a health facility and receiving at least one PNC check-up within 6 weeks after childbirth. Women were considered to have a complete continuum of care if they received all three services and incomplete if they missed one or more. We also explored the proportion of women who did not meet each component to provide more context on the distribution of incomplete care. Confounding variables In the PSM approach, the choice of confounding variables is based on previous literature regarding the determinants of treatment (ie, maternal continuum of healthcare) and outcome (ie, child immunisation) variables. A confounding variable is a factor that influences both the independent and dependent variables, but it should not lie on the pathway between the predictor and the outcome. In this study, women’s age, educational level, marital status, place of residence, sex of the household head, sex of the child, media exposure, household wealth index and countries in East Africa were considered as confounding variables. Parity (birth order) was evaluated as a potential covariate for inclusion in the propensity score model based on prior evidence suggesting its association with maternal healthcare utilisation and childhood immunisation. Although balance diagnostics demonstrated acceptable balance in the means (standardised mean difference –2.6), substantial residual imbalance remained in the variance (variance ratio 1.22, exceeding the commonly accepted threshold). Due to this unresolved imbalance, parity was not included in the final propensity score model. Statistical analysis In this study, we used PSM analysis. PSM is a statistical method that creates an artificial control group by pairing treated units with non-treated units based on similar characteristics and propensity scores. It aims to simulate randomisation and address selection bias issues that often affect observational studies.23 24 There are five key steps involved in estimating the effect using PSM: (1) propensity scores estimation, (2) choosing matching algorithm, (3) checking overlap or common support, (4) matching quality or effect estimation and (5) sensitivity analysis.25 26 Propensity score is a score that determines the probability of a subject’s treatment (or exposure) group as a function of the measured covariates for that subject and is expressed as follows27 : where Z is the treatment (exposure) variable (ie, the continuum of maternal healthcare in our case) and X represents the background variables (covariates). The propensity score is commonly estimated using a logistic regression model, where it represents the predicted probability based on the fitted model.27 We used a logit model to estimate the propensity score, with the continuum of maternal healthcare as the dependent variable and the other confounding variables in this study as predictors. The matching algorithms commonly used in PSM include Nearest Neighbour Matching, Calliper and Radius Matching, Stratification and Interval Matching, and Kernel and Local Linear Matching.25 In this study, we employed nearest neighbour matching, where individuals from the comparison group were selected as matching partners for treated individuals based on the closest propensity scores. However, nearest neighbour matching can result in poor matches if the nearest neighbour is significantly different. To address this issue, we set a tolerance threshold for the maximum allowable propensity score distance (calliper). Using this approach, we matched women with a complete continuum of care to those without, ensuring their propensity scores were within a 0.023 calliper. The size of the calliper was determined by taking 0.2 times the SD of the propensity score’s logit.28 29 Only three individuals fell outside the common support region and were therefore excluded from the analysis. This suggests that the majority of the treated and comparison individuals have overlapping propensity scores, indicating that common support is well-established in our data. The exclusion of these few individuals ensures that our analysis is based on a valid comparison between treated and untreated groups, minimising potential bias due to a lack of overlap in the covariate distributions. As a result, we can confidently proceed with the matching process, knowing that the treatment and comparison groups are well-aligned in terms of their propensity scores. The average treatment effect (ATE) and average treatment effect on the treated (ATT) were defined using only the region of common support (overlap). The treatment effect on an individual is determined by calculating the difference in outcomes with and without the treatment7 25 : where Di represents the difference in outcomes for individual i with and without the treatment and (Y1i, Y0i) denote the outcomes of individual i with and without the treatment. The population ATE refers to the difference in expected outcomes between participation and non-participation, which is expressed as follows: E(Y1) represents the expected value of Y for all units in the treatment group, whereas E(Y0) represents the expected value of Y for all units in the control group. The ATT is defined as the difference between the expected outcomes with and without treatment for those who actually received the treatment. It is the primary evaluation parameter of interest and is expressed as7 25 : The quality of the matching was assessed using pseudo-R² and standardised bias. This analysis incorporated the p value of the likelihood ratio along with pseudo-R² and standardised bias to evaluate the matching quality. In PSM, sensitivity analysis is employed to evaluate the robustness of inference results against potential unmeasured confounding. The sensitivity analysis was performed using the Mantel–Haenszel statistic. Data analysis was conducted using STATA version 17 software. Assumptions of the PSM method Unconfoundedness/conditional independence/selection on observables assumption There is a set X of observable covariates such that, once these covariates are controlled for, the potential outcomes are unaffected by the treatment status.7 25 30 Common support assumption For every possible value of X, there is a non-zero probability that individuals can be both treated and untreated.7 25 30 This means that, across all values of the covariates X, there should be individuals in both the treated and untreated groups. This overlap is crucial to ensure valid comparisons and proper matching between the groups in inference methods. Ethical considerations Access to the dataset was obtained from the DHS website (https://dhsprogram.com/). As this study used secondary data from publicly available survey data provided by the DHS programme, ethical approval and participant consent were not required. The data files do not contain any personal identifiers, including individual names or household addresses. Patient and public involvement Patients and/or the public were not involved in the design, conduct, reporting or dissemination plans of this research. Results Characteristics of the study participants The weighted sample of 13 488 women with children aged 12–23 months was used in the study. The median age of the women who participated in the study was 28 years (IQR: 23–33). Nearly a quarter, 2995 (22.21%) of the women were illiterate, and about one-third, 1580 (33.62%), of the children with incomplete immunisation had mothers who were not educated. More than three-fourths, 10 287 (76.26%), of the women were rural residents. The majority, 3761 (80.03%) of children with incomplete childhood immunisation were from rural areas. More than one-fifth, 2830 (20.98%) of the participants were from Kenya. Nearly a quarter, 1104 (23.49%) of children with incomplete immunisation were from Ethiopia (table 1). About 813 (5.91%) of children did not receive any vaccination (online supplemental table 1). Table 1Characteristics of the study participants used to assess the association between the maternal continuum of healthcare and child immunisation in East Africa (weighted n=13 488)VariableFrequency (%)Childhood vaccinationCompletefrequency (%)Incompletefrequency (%)Age of women (in years) 15–191013 (7.51)606 (59.80)407 (40.20) 20–243468 (25.71)2297 (66.24)1171 (33.76) 25–293482 (25.82)2288 (65.70)1194 (34.30) 30–342702 (20.03)1798 (66.54)904 (33.46) 35–391885 (13.97)1217 (64.55)668 (35.45) 40–44774 (5.74)486 (62.85)288 (37.15) 45–49164 (1.21)96 (58.68)68 (41.32)Women’s level of education No education2995 (22.21)1415 (47.24)1580 (52.76) Primary education6223 (46.14)4164 (66.92)2059 (33.08) Secondary education3481 (25.81)2607 (74.89)874 (25.11) Higher education789 (5.85)602 (76.33)187 (23.67)Residence Urban3201 (23.74)2263 (70.68)938 (29.32) Rural10 287 (76.26)6526 (63.44)3761 (36.56)Marital status Single1530 (11.34)1041 (68.07)489 (31.93) Married/living together11 486 (85.15)7493 (65.24)3993 (34.76) Widowed/divorced472 (3.50)253 (53.68)219 (46.32)Sex of the household head Male10 411 (77.18)6823 (65.54)3587 (34.46) Female3077 (22.82)1965 (63.85)1113 (36.15)Sex of the child Male6805 (50.45)4451 (65.41)2354 (35.59) Female6683 (49.55)4337 (64.89)2346 (35.11)Wealth index Poorest3285 (24.35)1850 (56.32)1435 (43.68) Poorer2828 (20.96)1799 (63.61)1029 (36.39) Middle2577 (19.10)1745 (67.72)832 (32.28) Rich2495 (18.50)1713 (68.67)782 (31.33) Richest2303 (17.08)1681 (72.98)622 (27.02)Media exposure No5244 (38.88)2866 (54.65)2378 (45.35) Yes8244 (61.12)5922 (71.84)2322 (28.16)East African countries Burundi1201 (8.91)1021 (84.97)180 (5.03) Ethiopia1808 (13.40)704 (38.93)1104 (61.07) Kenya2830 (20.98)1980 (69.95)850 (30.05) Madagascar1120 (8.30)546 (48.75)574 (51.25) Malawi974 (7.22)752 (78.12)222 (22.82) Mozambique722 (5.36)304 (42.08)418 (57.92) Rwanda762 (5.65)734 (96.27)28 (3.73) Tanzania741 (5.49)377 (50.84)364 (49.16) Uganda832 (6.17)478 (57.49)354 (42.51) Zambia1761 (13.06)1336 (75.84)425 (24.16) Zimbabwe737 (5.46)558 (75.75)179 (24.25)ANC <4 ANC visits5811 (43.09)3349 (57.64)2462 (42.36) ≥4 ANC visits7677 (56.91)5439 (70.85)2238 (29.15)Place of delivery Home delivery3510 (26.02)1499 (42.72)2011 (57.28) Institutional delivery9978 (73.98)7289 (73.05)2689 (26.95)PNC No PNC10 194 (75.57)6465 (63.42)3729 (36.58) PNC received3294 (24.43)2323 (70.52)971 (29.48)Continuum of care Complete continuum1791 (13.28)1354 (75.62)437 (24.38) Incomplete continuum11697 (86.72)7434 (63.55)4263 (36.45)ANC, antenatal care; PNC, postnatal care. Characteristics of the study participants used to assess the association between the maternal continuum of healthcare and child immunisation in East Africa (weighted n=13 488) ANC, antenatal care; PNC, postnatal care. Magnitude of continuum of care Among the total weighted sample of 13 488 women, only 13.28% (n=1791) completed the full continuum of care, whereas 86.72% (n=11 697) did not. Institutional delivery showed the highest coverage (73.85%), followed by attending at least four ANC visits (56.91%). PNC within 6 weeks was the least used service, with only 23.64% coverage. Among women with incomplete care, 43.80% received two of the three services, 30.53% received one and 13.55% received none. These findings indicate that incomplete continuum status often reflects partial service uptake, most commonly due to missed PNC follow-up. Association between the continuum of maternal healthcare services and child immunisation The PSM estimates indicate that the ATT for complete child immunisation was 0.0583, meaning that children of mothers who received a complete maternal continuum of care had a 5.83% higher probability of being fully immunised compared with children of mothers with incomplete care. Expressed relative to the treated group’s mean, this corresponds to a 7.48% increase (0.0583 ÷ 0.7794×100). The unmatched results indicated that the probability of complete child immunisation was 14.1% higher among mothers with a complete maternal continuum of care compared with those with an incomplete care (table 2). Table 2Propensity score matching estimates of average treatment effect on the treated, accumulated thermal units and average treatment effect of complete maternal continuum of care on complete child immunisation in East AfricaImpact of incomplete maternal continuum of care on incomplete child immunisationTreatedControlDifferenceSET-testUnmatched0.77950.63860.14090.012411.40ATT0.77940.72120.05830.02652.20ATU0.63850.70200.0635ATE0.0629ATE, average treatment effect; ATT, average treatment effect of the treated; ATU, accumulated thermal units. Propensity score matching estimates of average treatment effect on the treated, accumulated thermal units and average treatment effect of complete maternal continuum of care on complete child immunisation in East Africa ATE, average treatment effect; ATT, average treatment effect of the treated; ATU, accumulated thermal units. Common support There was a large overlap in propensity scores between the treated and untreated groups. Only three individuals (two from untreated and one from treated) were off support and excluded from the analysis, indicating that common support is well established in our data (figure 1 and figure 2). Histogram of propensity score distribution for the continuum of maternal healthcare services. Propensity score distribution of the continuum of maternal healthcare services. Quality of matching The pre- and post-matching mean values and biases for all covariates in both the treated and control groups were presented. The percentage bias reduction, along with the t-test and p values, was also reported for all matched covariates. The significance levels of the t-test indicated that the covariates used in this study were well-balanced (table 3). Table 3Balancing test for covariatesVariableUnmatchedMatchedMeanBias% reduction biasT-testP>tTreatedControlAgeUnmatched3.26253.2890−1.90−0.730.468Matched3.26083.22062.90−51.90.870.382Women educationUnmatched1.60301.051868.0025.750.000Matched1.60221.58900.5099.20.150.878ResidenceUnmatched1.57481.7691−42.30−17.250.000Matched1.57511.5847−2.1095.1−0.560.574Media exposureUnmatched0.77180.568544.3015.970.000Matched0.77160.76980.4099.10.250.805Sex of household headUnmatched1.25891.24333.601.390.166Matched1.25901.2596−0.1096.1−0.400.968Sex of childUnmatched1.49491.48951.100.410.681Matched1.49461.49220.5055.20.140.890Countries of East AfricaUnmatched7.28114.995271.3028.000.000Matched7.27887.3255−1.5097.9−0.430.671 Balancing test for covariates The quality of matching was evaluated using pseudo-R² and standardised bias. The mean and median reductions in standardised bias were 96.4% and 98.8%, respectively. This indicates that the matching process effectively reduced the average standardised bias by approximately 96.4%, demonstrating a significant improvement in covariate balance. Additionally, the pseudo-R² decreased from 0.128 to 0, and the likelihood ratio X² value (1.19) became non-significant (p=0.991), further confirming improved balance and suggesting that covariates no longer strongly predicted treatment assignment after matching. Moreover, the p value for the treatment variable was non-significant post-matching, suggesting that the treatment assignment is no longer significantly related to the covariates, indicating that the groups are now balanced with respect to those covariates (table 4). The comparison of covariate distributions between the treated (complete continuum of care) and control (incomplete) groups shows substantial improvement in balance after propensity score matching. Table 5 presents the distribution of each category of the categorical variables, showing the proportions in the treated and control groups before and after matching. Table 4Quality of matching testsSamplePseudo R2LR X2P value of LRMean biasMedian biasVariance ratio (B)Ratio of variances (R)Proportion of unbalanced (%var)Mean bias reductionUnmatched0.1281296.340.00033.242.399.50.863396.4%Matched0.0001.190.9911.20.53.81.050LR, likelihood ratio. Quality of matching tests LR, likelihood ratio. Table 5Covariate distribution in treated and control groups before and after matchingVariableCategoryContinuum of maternal healthcare servicesComplete (before matching)Incomplete (before matching)Complete (after matching)Incomplete (after matching)Age15–196.497.677.447.8420–2424.8825.8426.6325.8125–2926.4425.7226.6025.3430–3421.8319.7519.0620.0435–3914.9813.8214.4214.0140–444.235.974.785.7845–491.151.221.071.16Women educationNo education5.9524.6920.8924.54Primary38.4647.3247.1644.33Secondary40.8923.5025.9725.77Higher14.704.495.985.35ResidenceUrban47.7920.0526.7525.48Rural52.2179.9573.2574.52Media exposureNo20.2841.7338.3240.68Yes79.7258.2761.6859.32Sex household headMale75.4177.4675.9375.47Female24.5922.5424.0724.53Sex childMale50.7450.4051.1150.94Female49.2649.6048.8949.06CountriesBurundi1.5010.048.848.56Ethiopia1.9215.165.9712.92Kenya26.2720.1830.3122.22Madagascar4.358.915.678.27Malawi5.977.419.297.10Mozambique3.365.666.955.02Rwanda4.355.853.765.76Tanzania3.835.752.125.98Uganda3.426.592.717.18Zambia29.1010.6018.1312.37Zimbabwe15.933.866.254.61 Covariate distribution in treated and control groups before and after matching Sensitivity analysis The sensitivity analysis was conducted using the Mantel–Haenszel statistic to assess the potential impact of unobservable covariate biases on our conclusions regarding the effects of complete continuum of care. In our study, a critical value of ≥1.25 suggests that the result would be insensitive, and the conclusion could be uncertain. Discussion Our study evaluated the association between the maternal continuum of healthcare and child immunisation in East Africa using PSM. In observational studies where randomisation is not feasible, PSM is one of the most effective methods for assessing the impact of an intervention by creating an appropriate comparison group. This research investigates the association between the maternal continuum of healthcare and child immunisation while controlling for confounding variables. We used nearest neighbour matching with a calliper of 0.023 to compare childhood immunisation among women who received complete maternal continuum of care (exposed) and those who received incomplete maternal continuum of care (non-exposed). Our PSM analysis revealed that children of women with a complete continuum of maternal healthcare had a 5.83% higher chance of having complete childhood immunisation compared with children of mothers with an incomplete continuum of maternal healthcare services. Women who had an incomplete continuum of maternal healthcare had a 6.35% higher likelihood of their child to have complete childhood immunisation if they had a complete maternal continuum of healthcare. Furthermore, our results showed that, on average, a complete continuum of maternal healthcare increases the likelihood of complete child immunisation by about 6.29% in the entire population (both treated and untreated). This finding clearly demonstrates that a complete maternal continuum of care significantly contributes to complete child immunisation while controlling for potential confounders. If a woman receives all three maternal healthcare services during pregnancy, delivery and the postnatal period, her child is more likely to be fully immunised, regardless of her socio-demographic, socio-economic, pregnancy or childbirth status. This finding is supported by a study conducted in India, which revealed that women who received a complete continuum of care had a 17% increase in the likelihood of their child being fully immunised.7 Similarly, a study conducted in Benin found that the likelihood of full immunisation was lower among children whose mothers did not attend ANC visits, received assistance from traditional birth attendants and did not have a PNC check-up.31 Studies in Nigeria,32 Pakistan33 and Indonesia have shown that women with fewer or no ANC visits are less likely to have children who receive complete childhood immunisation. Research in Malawi34 and Ethiopia35 found that children whose mothers did not attend PNC check-ups were less likely to be fully immunised. Additionally, a study in Indonesia36 indicated that children whose mothers received assistance from traditional birth attendants during childbirth were less likely to have access to full immunisation. All of the above findings clearly indicate that all variables used to measure the maternal continuum of healthcare in our study, even individually, significantly affect complete childhood immunisation, further supporting our finding that the complete maternal continuum of care has a significant impact on child immunisation, controlling for other covariates, since our treatment variable was generated from those three individual variables. The possible reason why children of women who receive a complete continuum of maternal healthcare are more likely to be fully immunised is that these women have better access to healthcare services, including regular check-ups and health education that emphasises the importance of vaccinations. Continuous care throughout pregnancy, delivery and the postnatal period provides opportunities for healthcare providers to inform and remind mothers about immunisation schedules. Additionally, women who receive comprehensive maternal care tend to have healthier pregnancies and deliveries, resulting in healthier newborns who are more likely to receive timely immunisations. The trust built between mothers and healthcare providers also encourages consistent vaccination, further ensuring that children are fully immunised. In this study, it is important to acknowledge that the continuum of maternal healthcare and child immunisation is inherently multidimensional constructs. By converting these measures into dichotomous variables for analysis, some nuance and gradation in service utilisation and immunisation completeness may be lost. Although this simplification was necessary for propensity score estimation and balance assessment, it may mask variations within categories that could have implications for interpretation. In general, our study established a clear relationship between the maternal continuum of care and childhood immunisation in East Africa. We accomplished this by assessing whether the maternal continuum of care is associated with the likelihood of children receiving full immunisation. This research highlights how continuous care throughout pregnancy, delivery and the postnatal period can positively influence immunisation outcomes for children in the region. Strengths and limitations of the study Our study used nationally representative data and employed the PSM method to explore the relationship between the continuum of maternal care and childhood immunisation. By applying this rigorous approach, we were able to provide valuable insights with significant implications for policymakers, helping to inform the development of evidence-based policies. However, despite these strengths, there are some limitations that should be considered when interpreting our findings. One key limitation is that PSM can only control for confounders that are both measured and included in the model. This limitation may introduce potential bias if important variables are not accounted for. Another consideration is the possible reduction in sample size after matching. If strict matching criteria are applied, this may lead to a decrease in sample size, which can diminish statistical power and introduce greater imprecision in the estimates. Finally, the dichotomisation of both the treatment variable and the outcome, although necessary for the matching framework, inevitably reduces the complexity of these constructs and may obscure meaningful variation across intermediate categories. Conclusion and recommendations The study highlights that children of mothers who receive a complete continuum of maternal healthcare are more likely to have completed their childhood immunisations. This finding underscores the importance of integrating immunisation services into maternal healthcare programmes to improve vaccination rates and overall child health. To leverage this connection, it is crucial to enhance access to maternal healthcare, especially in underserved areas, and ensure that immunisation is a routine part of maternal care. Supplementary material