Healthcare access barriers and maternal healthcare service uptake in 14 sub-Saharan African countries (2017–2021): a propensity score matched analysis

Abstract

Introduction

The increasing inequality and multifaceted barriers to accessing maternal healthcare services play a significant role in the observed high maternal mortality ratios in sub-Saharan African (SSA) countries compared with other regions. Evidence is required to understand the relationship between barriers to healthcare access and the utilisation of maternal healthcare services. This study aims to assess the role of healthcare access barriers in the uptake of maternal health services in SSA.

Methods

We used a cross-sectional, demographic and health survey data from 14 SSA countries. We included 31 940 (31 553 weighted) reproductive-age women who had given birth in the year preceding the survey. Healthcare access barrier was an exposure variable and considered when a woman reported at least one of the following four concerns limiting access to healthcare services during illness: (1) difficulty obtaining permission to visit health facility, (2) financial challenges in covering healthcare expenses or treatment costs, (3) having significant travel distance to reach a health facility and (4) reluctance not wanting to visit facilities alone. We used χ² tests to select covariates and logistic regression to assess the association between healthcare access barriers and maternal healthcare service uptake. Propensity score matching (PSM) reduced selection bias by balancing observed characteristics between women with and without reported barriers. A one-to-one logit nearest neighbour matching method with bootstrap was used to estimate the average treatment effect on the treated (ATT) (women with access barriers) to antenatal care (ANC), health facility birth and postnatal care (PNC) services utilisation. Maternal education level, household wealth status, access to media, parity, COVID-19 pandemic period, geographical residence and country of residence were covariates used for matching. Post-matching sensitivity analysis and balance tests were conducted to assess the robustness of our PSM analysis.

Results

Almost three-fifths (60.2%) of women in the SSA (95% CI; 53.7% to 66.7%) reported barriers to accessing healthcare services. Barriers to accessing healthcare were associated with an 8-percentage point reduction in the likelihood of having a health facility childbirth (ATT: −0.08; SE: 0.016). However, there was insufficient evidence for differences in the uptake of ANC visit and PNC visit between the treatment (women with access barriers) and control groups. Post-matching tests confirmed that covariates were well balanced, and hidden biases were substantially reduced.

Conclusion

The presence of barriers to healthcare access is negatively associated with the utilisation of health facility childbirth in SSA countries. These findings underscore the need for public health programmes that address the primary barriers to healthcare access, including financial constraints and physical accessibility, to enhance childbirth outcomes in health facilities.

WHAT IS ALREADY KNOWN ON THIS TOPIC

• Previous studies have identified factors such as poverty, gender inequality, distance to health facilities and provider-related constraints as key impediments to maternal health service utilisation in sub-Saharan Africa (SSA). However, there remains limited evidence quantifying the role of women’s perceived barriers on the actual uptake of essential maternal health services across multiple SSA countries using robust analytical approaches.

WHAT THIS STUDY ADDS

• A substantial proportion of women in SSA reported encountering perceived barriers to accessing healthcare services. Using a propensity score matching approach, the study demonstrates that such barriers significantly reduce the likelihood of childbirth in health facilities by eight percentage points. However, their association with antenatal and postnatal care utilisation was not statistically significant.

HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY

• Our findings underscore persistent inequalities in maternal health service access across SSA and support prioritising interventions that address structural and social barriers. Enhancing women’s financial empowerment and improving service affordability and transportation are crucial to increasing facility-based childbirth and advancing Sustainable Development Goal targets for universal health coverage and maternal mortality reduction.

Background

The Sustainable Development Goals aim to reduce the maternal mortality ratio to less than 70 per 100 000 live births (Target 3.1) and to end preventable deaths of neonatal mortality to ≤12 per 1000 live births (Target 3.2) globally by 2030.¹ However, recent global estimates show that maternal mortality is still unacceptably high, with more than 2 87 000 women dying every year due to preventable causes related to pregnancy and childbirth complications globally.² While 95% of maternal deaths are from low- and middle-income countries (LMICs), Sub-Saharan Africa (SSA) alone accounted for approximately 70% of global maternal deaths.³ In 2020, the maternal mortality ratio in SSA was 545 per 100 000 live births, compared with a global average of 223 per 100 000 live births.⁴ Moreover, SSA bears the highest neonatal mortality rate globally, accounting for 43% of the total, with 27 deaths per 1000 live births reported in 2019.⁵

The lack of progress in reducing maternal and neonatal mortality in many countries may reflect inequalities in access to quality health services,² and the limited involvement of women in setting public priorities.⁶ A plausible approach to addressing this challenge is to ensure the provision of comprehensive and high-quality maternal and child health services.⁷ However, there is a low uptake of maternal health services in LMICs as well as in SSA.^{8 9} A study in 15 LMICs showed that only 13% of pregnant women received the WHO-recommended antenatal care (ANC) visits during pregnancy (ie, eight or more visits).⁸ Moreover, more than half (52.5%) of women did not receive postnatal care (PNC) in SSA countries.⁹

The low uptake of maternal healthcare by women is attributed to several factors, including sociodemographic and socioeconomic factors, as well as women’s decision-making autonomy.9,11 Limited access to healthcare is a significant factor with potential impact on maternal healthcare services.⁹ Studies showed that women who had barriers to healthcare access and had the poorest wealth index were significantly associated with incomplete utilisation of the maternal continuum of care.¹² Women who have the financial capacity to receive health services and receive permission to visit the health facility are more likely to have earlier initiation of ANC visits.¹³

Nonetheless, previous studies that have investigated the association between barriers to accessing maternal healthcare and utilisation relied on conventional logistic regression¹³ and failed to estimate the true effect of access barriers. The limited available studies assessing these could introduce bias as the populations in the exposure and control groups may possess different underlying characteristics, creating a challenge to accurately estimate the role of the exposure variable (barrier to accessing healthcare) on the outcome. Propensity score-matched analysis offers a more robust approach to controlling for confounding variables by matching on background characteristics.¹⁴ Thus, this study aimed to investigate the role of barriers to accessing healthcare on maternal healthcare utilisation using a propensity score matching (PSM) analysis.

Methods

Study design, population and data

This study used cross-sectional data collected from 14 SSA countries by the Demographic and Health Surveys (DHS).¹⁵ DHS is a nationally representative cross-sectional data collection to measure key policy-relevant and pragmatic indicators across LMICs. The DHS uses standardised questionnaires, coding procedures, data collection protocols and sampling techniques, which provide robust and comparable health outcome measures.¹⁶ We used the most recent standardised data, collected after the 2016 WHO recommendations on ANC.¹⁷ We included SSA countries with publicly available and standardised DHS datasets, resulting in a final analytical sample of the latest data from 14 countries (online supplemental table 1). Typically, DHS samples are stratified by administrative geographic region and by urban/rural areas within each region. The sample designs are usually two-stage probability samples drawn from an existing sample frame. In the first stage of sampling, enumeration areas (EAs) were selected with probability proportional to size within each stratum. In selected EAs, following the listing of households, a fixed number of households is selected from the selected cluster using equal probability systematic sampling. The detailed sampling procedure was outlined in the DHS guidelines.¹⁶ We used the Children’s Records or Kids’ Records DHS dataset. To account for the differences in population sizes between clusters, sample weighting was applied during the analysis to ensure accurate representation. The DHS guideline sets four sampling weighting methods, and from that, we use the individual weight for women (v005). Individual sample weights are generated by dividing the individual weight for women (v005) by 1 000 000 before use to approximate the number of cases.¹⁶

Of particular interest, our analysis focused on women of reproductive age (15–49 years) who had given birth in the year preceding the survey. Of all reproductive-aged women from SSA countries during the survey period (n=14 253), those who had given birth within the year preceding the survey (n=34 304) were included in the analysis. Finally, a total of 31 940 (31 553 weighted) samples were included in the analysis after excluding those with missing data on maternal healthcare access barriers (n=2364) (online supplemental figure 1). Moreover, (online supplemental table 1) presents details on the regions, countries, survey year and sample sizes included from 14 SSA countries (online supplemental table 1).

Patient and public involvement

There is no patient or public involvement in this study.

Study variables and measurement

Outcome measures

Outcome variables include ANC uptake, health facility childbirth and PNC uptake within the last 12 months. The 2016 WHO ANC model recommends that a pregnant woman have at least eight ANC visits. The initial appointment is advised to occur during the first trimester (ie, within the first 12 weeks of gestation). Two subsequent appointments are recommended in the second trimester, scheduled for the 20th and 26th weeks of gestation. Additionally, five appointments are recommended in the third trimester, scheduled for the 30th, 34th, 36th, 38th and 40th weeks of pregnancy.¹⁷ However, due to the small number (less than 10%) of women who had eight or more ANC visits, which might introduce bias in the analysis, we considered ANC uptake for women who had at least one ANC visit. Therefore, an ANC visit was defined by whether women had at least one ANC visit (measured as yes/no) throughout their pregnancy within the last 12 months. Health facility childbirth (measured as yes/no) included delivery that occurred in public, private or non-governmental organisation affiliated health facilities for the last birth within the last 12 months. PNC visit (measured as yes/no) was defined as a woman who had received at least one health check-up by a health professional within 42 days after giving birth in the last 12 months (online supplemental table 2)

Treatment and covariate variables

The treatment variable is a barrier to healthcare access. It was measured by women’s affirmative response indicating big problems/concerns to one or more of the following four maternal healthcare access-related questions at the time of data collection: (1) getting permission to visit a health facility, (2) getting money to cover healthcare expenses or treatment, (3) travel distance to reach a health facility and (4) not wanting to visit facilities alone.¹⁶

Maternal education level, household wealth status, access to media, parity, the COVID-19 pandemic period, geographical residence and country of residence were covariates used in our matching analysis (online supplemental table 2).¹⁸

Data management and analysis

Matching variables (covariates) ascertainment

We assessed maternal health service uptake as separate outcomes, including ANC visits, health facility childbirth and PNC visits, in relation to the exposure variable, barriers to accessing healthcare. The analysis was conducted using PSM to control for potential confounding factors and estimate the independent role of healthcare access barriers on each outcome.

Before performing PSM, two consecutive analyses were conducted. First, unmatched mixed-effects logistic regression analyses were performed to examine the crude associations between healthcare access barriers (the exposure or treatment variable) and each maternal health service outcome (ANC visit, facility delivery and PNC visit). These unmatched logistic regressions provided baseline estimates of the associations between barriers and outcomes before adjusting for bias. Second, potential confounding variables (covariates) were identified by assessing variables thought to be associated with both treatment and the outcome variables using the χ² test. Literature suggests that covariate selection should be guided by trade-offs between the effects of variables on bias (the distance of the estimated treatment effect from the true effect) and efficiency (the precision of the estimated treatment effect).¹⁴ Based on the relationships with treatment and outcome variables, observed covariates can be categorised into three groups: covariates only related to the treatment variable, covariates related to both the treatment variable and outcome (ie, confounders) and covariates only related to outcome.¹⁴ However, in this analysis, only confounders were included in the propensity score model, and covariates related solely to the outcome, such as maternal age, were removed. These variables included maternal education level, household wealth status, access to media, parity, the COVID-19 pandemic period, geographical residence and country of residence (online supplemental table 2). We used a directed acyclic graph ¹⁹ to select the minimum set of matching covariates (online supplemental figure 2). We used similar sets of matching covariates to assess the role of healthcare access barriers for all three maternal healthcare service uptakes. This is important to compare the difference in the effect size of healthcare access barriers among ANC visits, health facility childbirths and PNC visits in SSA.

Choosing appropriate matching methods

PSM is one of the most commonly used approaches for managing confounding during the analysis stage and adjusting for selection bias in non-experimental study designs.^{20 21} The propensity score, representing the predicted probability of receiving the exposure, was estimated using a logistic regression model that included the specified covariates. The estimated mean and standard deviation (mean±SD) of the propensity score for women in the treatment group (having a healthcare access barrier) compared with the control group (not having a healthcare access barrier) was 0.609 (± 0.131), with a region of common support score ranging from 0.266 to 0.831 (online supplemental table 3 and figure 3). All the number of blocks (17) have common support, and there are no treated and untreated off-support blocks (online supplemental figure 4).

After generating a balanced propensity score, we chose the appropriate matching method. Based on the likelihood ratio test and the level of bias reduction, the nearest neighbour matching approach was selected as the optimal method.²² One-to-one nearest-neighbour matching without replacement was performed using the psmatch2 command in Stata, with the analysis restricted to individuals within the region of common support. To generate the valid standard errors, we used a bootstrap with 500 replications. We applied the logit regression model in the nearest-neighbour matching method to estimate and compare the Average Treatment Effect on the Treated (ATT) and the Average Treatment Effect on the Untreated for ANC visits, health facility childbirth and PNC visits. Nearest-neighbour matching allows each treated individual (those reporting healthcare access barriers) to be paired with a comparison individual (those without such barriers) who has the most similar propensity score.¹⁴ Although the average treatment effect on population was also reported, this study is focused on estimating ATT. Our primary interest was to compare and estimate the difference in the probability of maternal healthcare utilisation (ANC visit, health facility delivery and PNC visit) between women who reported healthcare access barriers (treatment group) and those who did not. The difference in the average treatment effect between the treatment group and the controls was reported in percentage-point differences, along with their standard errors. This represents the change in the likelihood of maternal healthcare service uptake attributable to experiencing healthcare access barriers. In addition to our ‘psmatch2’ analysis, we conducted a ‘teffects’ matching analysis and included the results in the supplementary file. This allows comparison of the treatment effect estimates obtained using different SE estimation methods (online supplemental tables 4 and 5).

Post-matching analysis

A kernel density plot was used to evaluate the balance between the treatment and control groups (the overlap of propensity scores). Balancing tests were also objectively assessed using standardised differences and variance ratio tests. A covariate is considered perfectly balanced if it has a standardised difference of zero and a variance ratio of one.^{21 23} We also assessed bias reduction graphically and statistically. The balance between the treatment and control groups was assessed before and after PSM, standardised percentage bias, two-sample t-tests and model diagnostics tests. For each covariate included in the propensity score estimation model, percentage bias and per cent reduction bias, along with a statistical test’s significance level (p value), were reported. Rosenbaum sensitivity analyses were used to examine how strongly unmeasured confounding would need to influence treatment assignments to alter the PSM results.²⁴

Reporting and missing data handling

Women with missing data for the exposure variable, one of the items used to assess barriers to accessing healthcare (n=2364), were excluded from the analysis. Although matching was performed, these missing observations appeared to be completely random. All matching covariates and outcome variables had no missing values in the final analytic sample. For reporting this research, we have followed the ‘Strengthening the Reporting of Observational Studies in Epidemiology (STROBE)’ checklist for reporting cross-sectional studies.²⁵ This ensures that all essential reporting items are addressed transparently (online supplemental table 6).

Results

Barriers to accessing healthcare among women in SSA

In our data, 60.2% (95% CI 53.7% to 66.7%) of reproductive-age women reported at least one barrier hindering their access to and uptake of healthcare. From the types of barriers included, obtaining money to cover healthcare expenses or treatment (51%) and the distance to reach a health facility (35.2%) were the major barriers hindering maternal healthcare use (figure 1). The proportion of women reporting barriers to accessing healthcare ranged from the lowest in Zambia (43%) to the highest in Gabon (83%). As the educational status of women improves, barriers to healthcare access decrease, leading to increased uptake of maternal healthcare services. Notably, rural residents (67%) and women who had no access to media (69%) had higher rates of access barriers compared with their urban counterparts (49%) and those with media access (56%), respectively. Moreover, women from the lowest household’s wealth (75%) have more barriers to healthcare than those from the highest wealth households (36%) (table 1).

Figure 1. Proportions of women reporting having problems accessing healthcare (hint: * women who have at least one healthcare access problem from lists in this graph).

Table 1. The association between covariates with exposure and outcome variables, 2017–2021 SSA DHS (n=31 940).

Covariates	Healthcare access barriers (%)	P value	ANC visit (%)	P value	Health facility childbirth (%)	P value	PNC visit (%)	P value
Maternal age
15–24	7297 (61.3)	0.05	10 380 (87.3)	<0.001	8084 (68.0)	0.07	7902 (66.6)	0.004
25–34	8499 (59.6)		12 348 (86.6)		9601 (67.3)		9283 (65.2)
35–49	3643 (62.9)		4897 (84.6)		3834 (66.2)		3711 (64.2)
Educational level
No education	8583 (67.6)	<0.001	9804 (77.2)	<0.001	6618 (52.1)	<0.001	6850 (54.0)	<0.001
Primary	5502 (63.8)		7816 (90.7)		6034 (70.0)		5693 (66.1)
Secondary+	5354 (50.4)		10 005 (94.2)		8867 (83.5)		8353 (78.9)
Wealth status
Lowest	6138 (75.1)	<0.001	6350 (77.6)	<0.001	4021 (49.2)	<0.001	4276 (52.3)	<0.001
Low	4851 (68.6)		5945 (84.1)		4221 (59.7)		4222 (59.8)
Middle	3905 (60.2)		5790 (89.3)		4500 (69.4)		4311 (66.6)
High	2925 (51.3)		5263 (92.4)		4665 (81.9)		4282 (75.4)
Highest	1620 (35.9)		4277 (94.8)		4112 (91.2)		3805 (84.7)
Media usage
No	8108 (69.3)	<0.001	9200 (78.7)	<0.001	6220 (53.2)	<0.001	6257 (53.6)	<0.001
Yes	11 331 (56.0)		18 425 (91.0)		15 299 (75.6)		14 639 (72.5)
Residence
Urban	5339 (49.4)	<0.001	10 955 (92.2)	<0.001	8986 (83.2)	<0.001	8261 (76.7)	<0.001
Rural	14 100 (66.7)		17 670 (83.6)		12 533 (59.3)		12 635 (59.8)
Parity
Primiparous	4132 (57.3)	<0.001	6528 (90.5)	<0.001	5589 (77.5)	<0.001	5292 (73.6)	<0.001
Multiparous	9136 (59.6)		13 416 (87.5)		10 578 (68.9)		10 246 (66.9)
Grand multiparous	6171 (65.8)		7681 (81.8)		5352 (57.0)		5358 (57.2)
COVID-19 pandemic
Before COVID-19	13 803 (59.5)	<0.001	19 776 (85.2)	<0.001	15 558 (67.0)	0.03	15 320 (66.1)	<0.001
During COVID-19	5636 (64.6)		7849 (89.9)		5961 (68.3)		5576 (64.0)
Countries
Benin	1839 (61.7)	<0.001	2576 (86.5)	<0.001	2525 (84.8)	<0.001	2184 (73.3)	<0.001
Cameroon	1510 (75.8)		1707 (85.7)		1368 (68.7)		1249 (62.8)
Gabon	724 (83.0)		789 (90.5)		766 (87.8)		696 (79.8)
Gambia	860 (46.0)		1844 (98.7)		1573 (84.2)		1723 (92.2)
Guinea	1181 (72.0)		1410 (86.0)		864 (52.7)		901 (55.1)
Kenya	1255 (59.3)		2034 (96.1)		1723 (81.4)		1609 (76.0)
Liberia	612 (51.4)		1145 (96.2)		983 (82.6)		1050 (88.2)
Madagascar	1945 (73.7)		2263 (85.7)		978 (37.1)		1615 (61.4)
Mali	994 (49.6)		1573 (78.5)		1348 (67.2)		1218 (60.9)
Mauritania	1590 (66.7)		2007 (84.2)		1757 (73.7)		1209 (50.8)
Nigeria	3657 (55.2)		4979 (75.2)		2670 (40.3)		2935 (44.5)
Rwanda	846 (53.3)		1545 (97.4)		1503 (94.7)		1143 (72.0)
Sierra Leone	1559 (76.9)		1774 (87.5)		1713 (84.5)		1824 (90.0)
Zambia	867 (42.9)		1979 (98.0)		1748 (86.6)		1540 (76.3)

ANC, antenatal care; DHS, demographic and health survey; PNC, postnatal care; SSA, sub-Saharan African.

Maternal healthcare utilisation among women in SSA

In our study, 89.5% (95% CI 85.6 to 93.4) of women reported at least one ANC visit during their pregnancies. The lowest and highest proportions of women having ANC visits were reported in Nigeria (75.5%) and The Gambia (99%), respectively (figure 2A). Considering facility-based childbirth, 74.9% (95% CI 64.4 to 85.6) of women gave birth in health facilities, ranging from the lowest in Madagascar (38.7%) to the highest in Gabon (95.0%) (figure 2B). Of the total women, 71.3% (95% CI 61.7 to 80.8) reported having at least one PNC visit, with the lowest uptake observed in Nigeria (43.8%) and the highest in Gambia (93.2%) (figure 2C).

Figure 2. The proportion of women who took at least one ANC visit (A), had health institution delivery (B) and who took at least one PNC visit (C) in SSA countries. ANC, antenatal care; PNC, postnatal care; SSA, sub-Saharan African.

Healthcare access barriers and maternal healthcare utilisation

In our unmatched logistic regression analysis, there was a greater disparity in the likelihood of all three maternal healthcare service uptakes between women who faced healthcare access barriers and those who did not. Maternal healthcare use is higher in women who have no access-related barriers. Over three-quarters (75.7%) of women who had no healthcare access barriers had facility-based childbirth, while only 63.1% of women facing healthcare barriers did so. The crude OR showed that women who had barriers to healthcare access had 50% less chance of childbirth in a health facility as compared with women without healthcare access barriers (AOR=0.50; 95% CI 0.47 to 0.53) (oonline supplemental table 7).

Using covariate-adjusted PSM, we assessed how healthcare access barriers affected maternal healthcare utilisation by examining differences in ANC visits, health facility deliveries and PNC visits between women reporting barriers and a matched control group without barriers. We found the ATT for having a health facility childbirth in the treatment group and the control group was (β=0.62) and (β=0.71), respectively, with an ATT difference or coefficient of −0.084 (SE: 0.016). This means that women who have barriers to accessing healthcare have an eight-percentage point decrease in having a childbirth at a health facility compared with women who do not. For ANC and PNC services, the ATT estimates indicated reductions of 1.3 and 3.6 percentage points, respectively, among women with healthcare access barriers compared with women without barriers. These differences did not reach statistical significance (table 2).

Table 2. Unmatched and matched estimates of barriers to healthcare access on maternal healthcare utilisation.

Variable	Sample	Treated	Control	Difference	Coefficient	SE	P value
ANC visit	Unmatched	0.840	0.904	−0.064	—	—	—
	ATT	0.840	0.853	−0.013	−0.013 (−0.042, 0.016)	0.015	0.37
	ATU	0.904	0.862	−0.042	—	—	—
	ATE			−0.024	—	—	—
Health facility childbirth	Unmatched	0.621	0.755	−0.134	—	—	—
	ATT	0.621	0.705	−0.084	−0.084 (−0.115, −0.053)	0.016	<0.001
	ATU	0.755	0.753	−0.002	—	—	—
	ATE			−0.052	—	—	—
PNC visit	Unmatched	0.622	0.708	−0.086	—	—	—
	ATT	0.622	0.658	−0.036	−0.036 (−0.070, 0.001)	0.181	0.051
	ATU	0.708	0.698	0.011	—	—	—
	ATE			−0.018	—	—	—

ANC, antenatal care; ATE, average treatment effect (population); ATT, average treatment on treated; ATU, average treatment on untreated; PNC, postnatal care.

Post matching tests

Balancing test and bias reduction

Covariate balance between women with and without barriers to accessing healthcare showed that before matching, there was a substantial imbalance between the treated and control groups. Standardised differences (% bias) ranged from −38.4% (household wealth) to 34.4% (residence), with all significant differences in observed covariates (p<0.001). After matching, the standardised differences were substantially reduced, ranging from −0.8% (wealth status) to 2.0% (country), representing bias reductions between 70% and 99.8% across covariates, and with the t-test no longer statistically significant (p>0.05). The overall mean bias decreased from 22.9% to 0.7%, the median bias from 27.5% to 0.6% and the pseudo-R² of the propensity score model fell to <0.001, confirming that the matching procedure effectively eliminated systematic differences between treated and control units (online supplemental table 8). Figure 3 also shows the standardised percentage of bias across covariates before and after matching. After we conduct PSM, the standardised percentage of bias across all covariates is almost zero (figure 3). The distributions of the propensity scores overlapped perfectly after matching, achieving balance in the propensity score distribution between exposed and non-exposed women (online supplemental figure 5).

Figure 3. Standardised percentage of bias before and after matching across covariates for women with and without barriers to accessing healthcare.

Sensitivity analysis

We used the Rosenbaum sensitivity analyses to assess the overestimation (Q_mh+) and underestimation (Q_mh-) of the treatment effect. For health facility childbirth, the effect was highly robust and remained significant up to gamma (Γ) = 2.0 in both directions, suggesting strong resistance to hidden bias. For ANC and PNC visits, the significance of underestimating the treatment effect was maintained up to Γ ≈ 1.6 and 1.4, respectively, indicating moderate robustness. This means that an unobserved factor would need to increase the odds of treatment assignment by about 60% and 40% to overturn the result, respectively (online supplemental table 9).

Discussion

In this multi-country analysis using large population-based data and PSM, we aimed to generate high-quality evidence comparable to that of randomised controlled trials.^{26 27} In estimating the role of barriers to healthcare access on maternal healthcare service utilisation, the study provides important insights for policy and programme evaluations related to maternal healthcare utilisation. Overall, nearly three-fifths of reproductive-age women in SSA reported at least one barrier to accessing healthcare. Women who faced such barriers were about eight percentage points less likely to give birth in a health facility.

In our analysis, approximately three-fifths of reproductive-age women in SSA countries experienced at least one healthcare access barrier that affected health service use. Evidence shows that women in SSA countries faced multiple barriers to accessing maternal healthcare during their pregnancy and childbirth.^{12 18} These barriers to accessing maternal health services include obtaining permission to go for treatment, money for treatment, dealing with long distances to health facilities or being unwilling to go alone. Our findings also align with a previous study,¹² which found that a significant proportion of women faced barriers to accessing healthcare, predominantly financial barriers and limited access to health services. In Africa, a systematic review also showed that transportation barriers to health facilities, economic factors, lack of family support and cultural beliefs are the main barriers to accessing institutional birth and other maternal healthcare services.²⁸ A study in Bhutan also revealed that distance, inadequate transportation services and financial constraints are the primary barriers to accessing childbirth services at health facilities in districts with the lowest health facility childbirth coverage.²⁹ Moreover, a study in Malawi showed that distance significantly reduces the probability of having a facility delivery.³⁰ Given that remoteness and physical distance are significant impediments to increasing institutional delivery, some African countries, for example, Ethiopia, developed strategic interventions (health sector transformation plan—HSTP II) to narrow the existing huge urban-rural disparity in health facility childbirth.³¹ A systematic review in SSA also showed that the affordability of maternal health services, including household resources, willingness to pay and societal cash flow, remains a barrier to accessing care.³² Therefore, in addition to exempting maternal health services, government policies should address hidden costs, such as transportation, informal fees and opportunity costs, to ensure that essential maternal health services are truly accessible and affordable.

In this study, women who reported barriers to accessing healthcare had an approximately eight percentage-point lower probability of having childbirth in a health facility compared with those without such barriers. This is supported by a systematic review in SSA, which reported that access to obstetric care is limited by demand-side barriers, including household resources and income, lack of transportation, cultural beliefs and practices and limited awareness of required services. Supply-side barriers, including service costs, physical distance to facilities, long waiting times and poor staff communication and interpersonal skills, also contributed to reduced access.³³ Another study found that women from low-wealth households, those living in rural areas and those reporting major problems with distance to health facilities had higher odds of giving birth with the assistance of traditional birth attendants.³⁴ Similarly, studies from other lower-income countries, such as Bhutan and Afghanistan, have also documented an inverse association between healthcare access barriers and childbirth in health facilities.^{29 35} For instance, a study in Afghanistan identified healthcare access barriers, including distance, transportation costs and limited transportation availability, as the primary factors limiting access to health facility deliveries.³⁵ This may be because childbirth in many low-income countries is treated as an emergency event, and women often do not have pre-arranged bookings at health facilities for childbirth. Therefore, interventions that reduce these barriers, such as improving transport infrastructure, subsidising transportation costs and enhancing access to health facilities, can increase facility-based childbirth and improve maternal health outcomes.

In this study, barriers to accessing healthcare were not found to be associated with ANC visits. However, other studies have reported a negative impact of healthcare access barriers on ANC utilisation. For example, a study in Kenya found that shorter distances to health facilities were significantly associated with early ANC initiation.³⁶ Similarly, research in Rwanda showed that women who perceived barriers to healthcare had higher odds of inadequate ANC visits.³⁷ This difference may be due to the measurement of the exposure variable (barriers to accessing healthcare) used in our study. In addition to distance to the health facility, we included factors such as permission to seek care and reluctance to go alone in defining barriers to healthcare access. These factors, however, were not examined in previous studies and may explain the lack of association with the number and timing of ANC visits.^{32 38} For example, a study in three SSA countries (Guinea, Zambia and Mali) found that factors such as obtaining permission to visit health facilities, treatment costs, distance to healthcare centres and reluctance to go unaccompanied were not significantly associated with completing eight or more ANC visits.³² Another possible explanation is that, unlike other studies that focused on four or more ANC visits or eight or more ANC visits, we used a single ANC visit as the outcome, which may not capture a significant difference in healthcare access barriers compared with non-users. Moreover, the above study employed multivariable analysis to control for confounding, whereas our study used PSM, which is more effective in addressing selection bias in cross-sectional studies.

Barriers to women’s healthcare access have no significant association with the uptake of PNC services. This is supported by a study in Nigeria that showed that barriers to the utilisation of healthcare, such as distance to facilities, lack of confidence and the preference for female health workers, are unlikely to play a major role in PNC utilisation.³⁹ Similarly, a study in rural Malawi showed that, despite approximately 96% of women having facilities providing PNC within 10 km of their residence, only about 3% received maternal PNC within 24 hours.⁴⁰ In contrast, a study among women in SSA countries showed that women with no significant problems accessing healthcare have a 4% increased odds of PNC service utilisation as compared with those facing significant problems.⁴¹ Another study in South Sudan demonstrated a significant association between limited access to health services at government health facilities and the use of early PNC.⁴² This difference could be attributed to the fact that PNC services are often provided after discharge or delivery at home by any health providers, including community health workers or traditional birth attendants.⁴³ During this time, barriers to accessing healthcare, such as obtaining permission, securing funds, dealing with distance and reluctance to go alone, may not be major concerns. It is worth noting that previous studies on this topic used conventional regression models to investigate the association between barriers to accessing healthcare and maternal healthcare utilisation. Additionally, they did not consistently use similar variables to categorise the barriers to accessing healthcare.

This study is the first to examine the role of barriers to accessing healthcare on maternal healthcare service utilisation in SSA countries using PSM analysis, which is effective in reducing bias and accounting for confounding effects. Moreover, to minimise recall bias, the study focused on women who had live births in the year before the surveys. However, it is important to note some limitations of our study. First, although the components of barriers to accessing healthcare, including getting permission to visit health facility, not wanting to visit facilities alone, travel distance to reach a health facility and securing money to cover healthcare expenses or treatment, are predominantly behavioural, geographical and economic constraints that are unlikely to change substantially over short periods (eg, within 1 year), the possibility of reverse causality cannot be excluded. This is because maternal health service uptake was measured for the preceding 12 months, whereas exposure to access barriers was assessed at the time of data collection. Second, PSM matches the treated with controls based on measured covariates, and as a result, bias due to unmeasured covariates is not considered. Since we used secondary data, it was difficult to obtain the predictor variables for maternal healthcare barriers. This limitation could lead to overestimation or underestimation of the effects of barriers to healthcare access on maternal healthcare services. Third, by including only women who had live births, there is a possibility of selection bias. However, a recent methodological review suggests that the bias resulting from restricting analysis to live births may be minimal.⁴⁴ Nonetheless, the consistency of patterns observed across multiple countries suggests that the identified barriers and their role in maternal healthcare service uptake are likely relevant to many SSA contexts with similar demographic and healthcare characteristics.

Conclusion

About three-fifths of women in the sampled SSA countries have barriers to accessing healthcare, of which getting money for treatment and distance to the health facility were the major challenges. Having barriers to accessing healthcare has a significant association with childbirth in health facilities in SSA countries. Consequently, there is a pressing need for public health programmes designed to overcome these barriers and improve childbirth outcomes in health facilities. To address this issue effectively, increased efforts should focus on improving women’s financial empowerment and accessibility to maternal health services, aiming to prevent barriers to healthcare access and subsequently boost the maternal healthcare utilisation rates.

Data availability statement

Data are available in a public, open access repository.