Spatial variation and inequities in antenatal care coverage in Kenya, Uganda and mainland Tanzania using model-based geostatistics: a socioeconomic and geographical accessibility lens

Despite a 38% reduction in maternal mortality ratio (MMR) between 2000 and 2017, about 810 women died each day due to complications of pregnancy and childbirth in 2017 globally [1]. Similarly, two million stillbirths occurred in 2019, despite a 35% reduction since 2000 [2]. The majority of the maternal deaths (66%) and stillbirths (40%) occurred in sub-Saharan Africa (SSA) [1, 2]. Across the globe, SSA still has one of the highest disease burdens, with an 89-fold higher MMR and a 36-fold higher stillbirth rate compared to Europe. Within SSA, MMR and stillbirths vary between [1, 2] and within countries [3, 4]. This variation has been attributed mainly to inequities in access to quality health services, varying levels of poverty, and differences in education attainment [3‐6].

Most maternal deaths and stillbirths are preventable through high-quality care in pregnancy and during and after childbirth [7]. Antenatal care (ANC) is a crucial element of the continuum of care and aims to prepare for birth, prevent, detect, alleviate, and manage pregnancy-related complications that may occur. ANC also presents an opportunity for health promotion among women, families, and communities [8‐10].

The World Health Organization (WHO) developed the “focused ANC model” in the 1990s to guide routine care at four critical times during pregnancy (ANC4+) [11]. This guideline was revised to eight contacts in the 2016 update to improve the experience of care and minimize the risk of poor pregnancy outcomes [8, 9]. However, in SSA, the proportion of women who meet even the pre-2016 requirement of four ANC visits remains suboptimal. While eight in ten (81.9%) pregnant women in SSA report at least one ANC visit, only 53.4% had at least four visits in 2020 [12]. In Latin America and the Caribbean, 91% of women had ANC4+ visits [12]. The ANC4+ coverage in Kenya (58.5%), Uganda (56.7%) and Tanzania (62.2%) is moderate relative to other SSA countries like Ghana (90.5%) and Liberia (87.3%) [12, 13]. ANC coverage is also heterogeneous within countries in SSA, with wide coverage gaps by residence (rural and urban), maternal education, and household wealth quintile [14‐17].

To reduce maternal and perinatal mortality through ensuring equitable access to ANC services, it is crucial to examine how ANC4+ coverage varies across sub-groups at high spatial resolution [15, 18]. This will inform where and who should be targeted the so-called hotspots requiring action. The WHO-led Ending Preventable Maternal Mortality (EPMM) working group outlined global targets and strategies for reducing maternal mortality within the Sustainable Development Goals (SDGs) framework [19, 20]. ANC4+ coverage is one of the core priority indicators within the global monitoring and reporting framework [18]. In this framework, at least 90% of all countries and 80% of all districts in a country are expected to have over 70% (target coverage) of pregnant women having ANC4+ visits by 2025 [19]. We apply this target coverage to guide our exceedance probability analysis. Countries also set local targets; Kenya’s targets ANC4+ coverage of 57% by 2020/21 [21], 50% in Uganda by 2021/22 [22] while Tanzania targeted 60% by 2020 [23]. Tanzania is also tracking early ANC coverage (< 12 weeks) aiming a 60% coverage by 2025 [24]. These countries track the targets monthly using routine data supplemented by survey data when available. However, routine data has poor reporting rates and lacks socioeconomic data for equity analysis [25].

Recognizing that relying on broad, aggregate, and national-level estimates masks inherent spatial pockets of sub-national inequities, countries need to evaluate ANC4+ coverage along sub-groups [18, 26] at high spatial resolution. Previous studies have examined ANC4+ coverage across sub-groups in Kenya, Uganda, and Tanzania [13, 14, 16, 27‐30]. However, none of the earlier studies mapped ANC4+ coverage inequities per sub-group at high spatial granularity. Further, previous studies have not assessed the extent to which EPMM’s ANC4+ target coverage has been achieved overall and across subgroups. Model-based geostatistics (MBG) [31] offers a principled likelihood-based approach to problems concerning the modeling of the spatial variation of a phenomenon of scientific interest such as ANC4+ and robustly assesses attainment of target coverage. It has been applied widely across public health problems where the goal is to make inferences using spatially discrete cross-sectional survey data, especially in low resource settings where disease registries are incomplete or non-existent [32‐34]. In this study, we aimed to model ANC4+ coverage, likelihood of achieving target coverage and number of women who need to be reached disaggregated by three equity stratifiers (household wealth, woman’s education, and travel time to nearest health facility) using data from household surveys in Kenya, Uganda, and mainland Tanzania. All analyses were at 3 × 3 km spatial resolution and aggregated by district.

Monitoring ANC4+ coverage and associated inequities requires quantifying and describing the coverage across population groups defined along socioeconomic and geographic equity lines within countries [19, 20]. This should be at a high resolution, the so-called precise public health [71], to highlight hotspots areas within a country. Our findings show that ANC4+ coverage was moderate, with six in every ten pregnant women reporting having received at least four ANC visits in the three East African countries. At the national level, this is short of the 70% coverage anticipated to be achieved by 2025 under the EPMM strategy. However, national targets set by the governments of each of the three countries were achieved. Compared to similar national estimates about decade ago (since 2021), there have been slight improvements. In the early 2010s, between four and five in ten pregnant women had ANC4+ visits - that is, 47.1% in Kenya (2009), 47.6% in Uganda (2011) and 42.8% in Tanzania (2010) [13]. These improvements may be explained by the concerted efforts of stakeholders which included healthcare investment focused on access, training health professionals, decentralized health care, maternal health education, user fees reduction or abolishment among other targeted initiatives [72‐80].

However, despite the moderate national improvements and associated efforts, the current ANC4+ coverage is inequitable, and falls short of recommended levels. Yet, the role of ANC in preventing, detecting, alleviating, and managing pregnancy-related complications that might lead to maternal deaths and perinatal mortality is well known. Our findings show the specific districts that have the least coverage and the linked inequities dragging the coverage. This will aid in targeted allocation of resources, subsequent monitoring and evaluation, and benchmarking. This aligns with the SDG mantra of leaving no one behind and starting with the farthest behind, first. The high-resolution maps in Fig. 2 aid in identifying hotspots within the districts with poor coverage, while the exceedance probabilities minimize the chance of misclassifying districts and pixels. This ensures persistent foci of low coverage are correctly identified such that resources are not wasted on interventions and populations who do not require them. We have provided all the district estimates in Additional file 2 for use by policymakers.

The most left behind (lower levels of ANC4+ coverage) districts bore a treble burden where the poorest, with the least education and geographically marginalized from healthcare reside. Women from these districts maybe at a higher risk of maternal mortality and perinatal deaths. There were also districts that had both lowest coverage of ANC4+ and at the same highest number of pregnant without ANC4+ visits. Certainly, resources, and infrastructure are concentrated in wealthier urban places and are scant in poorer and remote areas [81]. The hotspot districts and most in need, include West Pokot, Wajir, Mandera, Turkana, Baringo, Garissa, Elgeyo-Marakwet, Marsabit and Trans Nzoia mainly northern Kenya; Amudat, Moroto, Napak, Nabilatuk, Nakapiripirit, Kalangala, Buvuma, Namayingo, Napaka and Palissa majorly located in eastern Uganda and finally, Kakonko, Biharamulo, Kaliua, Kibondo, Bukombe, Chato, Bariadi TC, Urambo, Nzega, Igunga and Itilima mainly north-west Tanzania.

The hotspot counties in northern Kenya have been historically marginalized, are predominately arid and semi-arid and sparsely populated. The region has poor infrastructure, often stricken by conflict and insecurity which may lead to poor geographic access to healthcare. Further, women in this region have low education attainment, mainly come from poor households, and practice some cultural beliefs antagonist to western medical practices [14, 82, 83]. Likewise, eastern Uganda is among the poorest region in the country and has poor coverage of other maternal and child health indicators [28, 84, 85]. Long distances, poor roads and high transport costs, poor services at the health facilities and lack of access to health-related information also impede women to utilize maternal services in this region [86]. Similar situation exists in North-western Tanzania which is poor and has low conditional probability of transitioning from poor to non-poor status [87]. Further, socio-cultural beliefs, distance, lack of transport, perceived poor quality of ANC services have been reported as barriers to ANC use in this region [88]. Combined in the three countries, these factors provide insights on how to improve the poor coverage in the hotspots. However, our study was concerned with identification of these hotspot through predictive modelling [54], therefore, granular (detailed and context-specific) quantitative and qualitative studies should be conducted to better understand why the districts have been left behind.

Our results showed that the poor had lower ANC4+ coverage. It’s the poor who have the highest disease burden, reduced access to healthcare services and the majority do not utilize health services at all [89]. The pro-rich inequities have been observed before [30] and continue to be persist even among the poor pregnant women who are beneficiaries of government initiatives to improve ANC uptake [80, 81, 89]. Ensuring sufficient and timely reimbursements to prevent out-of-pocket payments and minimizing indirect costs of transport [75, 76, 90] will likely increase uptake among the poor ANC clients where initiatives already exist. It is the poor ANC beneficiaries of initiatives who are negatively affected by stock-outs, dysfunctional medical equipment, shortage of healthcare workers, strikes and discrimination [29, 89] since they cannot afford paying services in the private sector. These bottlenecks require addressing so that the woman who have been left behind can benefit from programs and initiatives put into place. The high ownership of mobile phones in East Africa can be leveraged to create mobile health program simultaneously with community health workers (CHWs) to facilitate follow-ups and minimize socioeconomic barriers [91] among the poor. Determining the degree of follow-up needed based on ANC user characteristics during the first ANC visit can also be used to increase return visits and ANC uptake.

Women without formal education had lower ANC4+ coverage. Maternal education and household wealth and are linked. Women from poor households often have lower educational attainment which negatively affects utilization [92] as observed in the hotspot districts. In the short run, health promotion and outreach campaigns among pregnant will be useful [91, 93] at the village-level [93] or through mass media [94] in the hotspots. This could neutralize harmful traditions and cultural beliefs, misinformation from family or traditional healers, or cases where pregnant women are misled to delay ANC visits [84, 95]. There is a need to raise awareness about new initiatives meant to increase uptake of ANC since lack of awareness has been a barrier in previous initiatives [38, 77, 96]. There is a necessity to integrate and bolster the need for maternity care seeking into educational curriculum. In the long term, higher education attainment will be vital in increasing women’s autonomy, improved access to healthcare information, and may lead to higher socioeconomic status [97] in the hotspot areas.

Long travel time remains a challenge among women in remote areas even where interventions have been implemented [90, 98] and has been linked with lack of public transport and roads in poor conditions [89, 99‐101]. Access to bicycles has shown to be a pro-poor option in increasing access to health centers and can be used as entry point to intervene on areas with poor geographical access [100], supplemented with contracted transporters [77]. Mobile services could also be implemented to meet the women in their communities [14]. Under the Beyond Zero campaign in Kenya, mobile clinics have provided healthcare to poor and marginalized communities [102]. CHWs are integral in promoting maternal care seeking [103] and might be effective in the hard-to-reach areas [104].

Beyond the demand side challenges, there is also a need to strengthen the supply side to guard against inadequate drugs, equipment, infrastructure, skilled human resources, overburdened health facilities, longer waiting times, reduced health worker motivation and quality of care [38, 72, 75‐77, 90, 96]. Further, coverage might have been affected by the COVID-19 pandemic, health workers strikes and absenteeism which were associated with a lower likelihood of attending ANC [105‐107]. The poor usually bear the burden since they rely mainly on the public sector and cannot afford care from the private sector [108, 109]. The pandemic strained the health system, disrupted essential health services due to inability to access healthcare, transport restrictions, curfew, and fear of contracting the virus when seeking care [110].

The key strengths of our study lie in deriving high resolution maps per each equity stratum, unlike previous studies and if they do, the resolution is course and unsuitable for granular targeting and prioritization. Notable effort is STATcompiler by the DHS program [13] that produces similar estimates as our study and make it publicly available, however, they disaggregate at broad administrative regions. We have also used exceedance probabilities to account for the uncertainty in the data and quantified the likelihood of meeting target ANC4+ coverage, an aspect that has not been considered in previous ANC4+ coverage studies. Another strength is the use of nationally representative surveys which makes our findings to be comparable and generalizable.

Despite the strengths of our study, there are some limitations. There might have been recall bias synonymous with any retrospective data. There was also selection bias since the surveys included women with a live birth 3 years preceding a survey. Women who might have died during pregnancy or with other birth outcomes were excluded. Related to this is the population data that represented all pregnancies; however, ANC visits were asked only when those pregnancies resulted in live births. The conceptual discrepancy might have biased the estimated number of women with ANC4+ visits. The surveys were conducted at different time points across the three countries - Kenya (2020), Uganda (2018/19) and Tanzania (2017)- limiting temporal comparisons between the countries.

The displacement of cluster coordinates due to confidentiality was not accounted for but was minimized by taking averages of estimates within a buffer. Factors that are associated with ANC beyond those collected during the MIS were not considered except for travel time and NTL. We assumed pregnant women used their nearest facility, yet some proportion bypass their nearest facility [111]. We also did not account for weather variation, traffic jams and other factors that affect transport when estimating travel time. Further, having geographical access is not equivalent to either use of care nor its high quality [112]. We used the number of ANC visits with a qualified professional but did not incorporate data on the content or quality of this care, which is critical to the effectiveness of ANC as a maternal and perinatal mortality reduction strategy. We focused on ANC4+ coverage, however, timing of first visit is also critical to achieving four visits. Women who start late, have very low likelihood of reporting ANC4+ visits, which merits examination in a similar way as we did for ANC4 + .

Household surveys provide an opportunity to monitor the coverage, however, they are conducted every 3 to 5 years, limiting tracking at a higher temporal granularity. In addition, sample size from surveys is often limited and inadequate for high spatial resolution risking a covariate driven ANC4+ coverage [113] especially when stratified as we did. On the other hand, routine health data offer an alternative source of information to monitor ANC4+ coverage. However, routine data are limited due to poor reporting rates, challenges in determining accurate catchment population [25] and does not collect socioeconomic datasets relevant to equity assessment. However, routine data can be linked on spatially smoothed equity stratifiers from household surveys and used for equity monitoring. Finally, despite the findings, we cannot infer causality with the cross-sectional survey data that we used.

ANC coverage rates have remained moderate, with about 60% of pregnant women having the recommended four or more visits provided by skilled health personnel in East Africa. The likelihood of attaining district-level target coverage by 2025 is very low. Further, the coverage is inequitable, with women from poor households, without formal education and geographically marginalized from formal healthcare having persistently lower coverage and lower likelihood of receiving at least four visits. The spatially disaggregated information will be valuable to policymakers for improved targeting of annual appropriations and leveraging initiatives aiming to improve coverage of recommended interventions and reducing maternal and perinatal mortality.