Mapping maternal and newborn healthcare access in West African Countries

Background: The Sustainable Development Goal (SDG) three emphasizes the need to improve maternal and newborn health, and reduce global maternal mortality ratio to less than 70 per 100 000 live births by 2030. Achieving the SDG goal 3.1 target will require evidence based data on concealed inequities in the distribution of maternal and child health outcomes and their linkage to healthcare access. The objectives of this study were to estimate the number of women of reproductive age, pregnancies and live births at subnational level using high resolution maps and to quantify the number of pregnancies within user-defined distances or travel times of a health facility in three poor resource West African countries: Mali, Guinea and Liberia. Methods: The maternal and newborn health outcomes were estimated and mapped for the purpose of visualization using geospatial analytic tools. Buffer analysis was then performed to assess the proximity of pregnancies to health facilities with the aim of identifying pregnancies with inadequate access (beyond 50km) to a health facility. Results: Results showed wide variations in the distribution of maternal and newborn health outcomes across the countries of interest and districts of each of the countries. There was also clustering of health outcomes and health facilities at the urban capital cities of Bamako, Conakry, and Greater Monrovia. Conclusion: To bridge the gap in inequity in healthcare access, and improve maternal and newborn health in the study countries, there is need for equitable distribution of human resources and infrastructure within and across the various districts. in in Pita and Lelouma in Mamou and Labe regions respectively. Our study also revealed that 46% and 89% of pregnancies in Mamou in central Guinea were located outside 50km and 25km distance of a health facility respectively. This finding corroborates the report of 2012 Demographic and Health Survey (DHS) and Multiple Indicators Cluster Survey (MICS) 37 which indicated that a higher proportion of women (84%) in Mamou did not receive maternal healthcare after child birth which is crucial for the survival of the mother and newborn 38 .This is an indication of lack of adequate access to maternal healthcare services, which could be due to the long distance (of 50 and 25 kilometres) that the women had to travel to access healthcare.


Introduction
Improvement in maternal and newborn health in developing countries has been a major priority in public health since the 1980s. This is reflected in the consensus reached at different international conferences, such as the Safe Motherhood conference in Nairobi in as well as specific targets in the Millennium and Sustainable Development Goals. In spite of these efforts to increase access to reproductive health services and reduce maternal mortality, maternal health is still poor in most developing countries. Globally, about 830 women die from pregnancy-or childbirth-related complications every day, and it was estimated that in 2015, roughly 303 000 women died during pregnancy and childbirth 1 .
Unfortunately, almost all of these deaths (99%) occurred in low-resource settings, and most could have been prevented with adequate access to healthcare. Although a number of countries in sub-Saharan Africa halved their levels of maternal mortality since 1990, mortality rates for newborn babies have been slow to decline compared with death rates for older infants. The Sustainable Development Goals (SDGs), target 3.1, is to reduce the global maternal mortality ratio to less than 70 per 100 000 live births by 2030 and improve maternal and child health.
For this target to be achievable and realised there has to be a concerted effort to improve the maternal and newborn health in low income countries, and in particular in the sub-Saharan African region. However, this requires accurate data for adequate planning of safer births and healthier newborns, which is difficult in these low-income countries due to poor vital registration information, and a lack of data on the geographical distribution of women of reproductive age.
Also, subnational estimates of the projected future numbers of pregnancies are needed for more effective strategies on the allocation of human resources and infrastructure; and to assess coverage of services, there is a need to link information on pregnancies to travel distances for better information on health facilities in districts and regions 2 .
Understanding the magnitude of inequities in health outcomes for both women and newborns is important in improving maternal and newborn health and services.
Furthermore, the application of geospatial analysis and mapping of maternal and neonatal outcomes in relation to how close they are to health facilities is useful in identifying high priority areas or districts where women have low access to healthcare services; and also important in the fair distribution of these services 3 . Moreover, data visualization with the use of maps and geospatial analysis has been found to play an important role in addressing the need for improved maternal and newborn health service provision and access to emergency obstetric care at sub-national scale. In fact, geospatial analysis is highly recommended for application to maternal health programs in poor resource settings 4 .
Several studies using geospatial applications and mapping have focused on child mortality and childhood co-morbidity 5,6 ; reviews on the importance of geospatial applications to maternal and newborn outcomes 7 ; and geographical factors within the vicinity of severe injury related to motor vehicle accidents 8 . Others have examined the distribution of births and pregnancies in Afghanistan, Bangladesh, Tanzania Ethiopia and Ghana , 2,9,10 ; health metrics and geography of maternal and newborn health 11 ; spatial accessibility to health professionals in France 12 ; geographic disparities in utilization of care in East African countries 13 ; and spatial distribution patterns of healthcare facilities in Nigeria 14 .
But to the best of our knowledge, studies on the estimation and mapping of maternal and newborn outcomes at subnational level and linkage to health facility access in poor resource settings in West Africa is sparse. In view of the fact that most births (60%) take place at home and the burdens of disease and maternal and neonatal deaths are unacceptably high in most West African countries, this study is timely and adds to the body of knowledge through identifying districts with low maternal and newborn healthcare and access. This identification will inform decisions on appropriate maternal health policy and interventions in particularly disadvantaged districts.
Against this backdrop, this study examines spatial variability in the distributions of women of reproductive age, pregnancies and births in three West African countries (Mali, Liberia and Guinea) with a high burden of maternal and neonatal deaths. These three countries share borders, and given the porosity of borders in Africa (with communities of the same language group often separated by country borders), we can examine whether there are any differences in accessibility of maternal and newborn health services across neighbouring countries.
The objectives of the study were to describe and visualize the distribution of women of reproductive age, pregnancies and live births using high resolution maps and also quantify the number of pregnancies within user-defined distances or travel times of a health facility. This work has a distinct usefulness in health systems and policy through showing where there is demand for maternal and newborn health services. The estimates of proximities of pregnancies to health facilities link estimates of population in need with locations of facilities designed to meet this need 15,16 . Furthermore, the use of maps to display the results is a clear way of showing the spatial heterogeneity that exists at a subnational level and highlights geographic inequities in service provision 17,18 .

Data And Methods
This study utilizes publicly available WorldPop 19 data derived from an integration of satellite, census and household survey data for three (3) West African countries, namely Mali, Liberia, and Guinea. The WorldPop dataset was constructed using the most recent and spatially detailed datasets available 2 . Detailed maps of settlement extents were derived from Landsat satellite imagery through either semi-automated classification approaches or expert opinion-based analyses 20,21 . These settlement maps were then used to refine land cover data. Additionally, local census data mapped at fine resolution enumeration area level from a random selection of countries across the continent were utilized to identify typical regional per-land cover population densities. These were subsequently applied to redistribute census counts to map human population distributions at 100m grid square spatial resolution. Additional country-specific datasets (where available), that provided data on population distributions not captured by censuses, were incorporated into the mapping process. Estimates of age and sex structures on subnational population compositions from the last 20 years were obtained from a variety of sources for the study countries. The datasets on numbers and proportions of individuals by age and sex were collated for as many subnational units as available within the last two decades, using sample weights where applicable to household surveys to provide aggregate estimates, and these were matched to corresponding geospatial datasets showing the boundaries of each unit. Africa-wide geospatial linked data on the number of individuals by age and sex within administrative unit were created. Furthermore, UN statistics and other sources on growth rates, age specific fertility rates, live births, stillbirths and abortions were then integrated to convert the population distribution datasets to gridded estimates of births and pregnancies.
In addition, information on health facilities was obtained from the Humanitarian Health Exchange (HDX) 22 and pre-processed for each country of interest. This data resource provides an open, accessible, and publicly available platform and repository to make health data easy to find and use for analysis. The Humanitarian Health Exchange collates information from various datasets from international and national organisations and business portals for the purpose of ensuring a coherent response to emergencies, and is managed by the UN's Centre for Humanitarian Data, under the auspices of the UN Office for the Coordination of Humanitarian Affairs (UN OCHA) 23,24 . The health facilities' information from the Humanitarian Health Exchange has to be geo-located and verified before being uploaded to the repository, through a quality assessment involving incountry experts 24 . This verification process provides an indication of the quality of the data assembled in the Humanitarian Health Exchange. In countries with good data infrastructure, not only is comprehensive geographic location information on health facilities available, but also available is information on their features and functions (e.g. number of doctors and nursing staff, services, care management and environment). This allows us to ascertain which health facilities can be determined to provide either basic or comprehensive emergency obstetric and newborn care (i.e. higher-level facilities with the capacity for caesarian section and blood transfusion). For our study countries, unfortunately, this is not readily available. We therefore, decided to restrict the listing of health facilities to include only those that met a stringent criteria of verifiability. This facilitated the compilation of an accurate though not comprehensive list of geocoded health centres that have the capacity to provide basic and emergency obstetric and newborn care, based on information from the Humanitarian Health Exchange for use in our analysis. After excluding those facilities that did not meet our criteria, the final analytic sample included 430 facilities (with 62 in Mali, 188 in Guinea, and 180 in Liberia). Finally, this list of health facilities was imported into ArcGIS software to use the latitude and longitude (geo-location) information to create a shapefile of health facilities throughout the study countries. The map situating the locations of the three study countries is provided in Figure 1, and in the next section, brief profiles of the countries are given.

Mali
Mali is a large landlocked country in the West African Sahel region, with a population of 18.5 million. It is one of the poorest countries in the world with a gross national product of approximately US$2,160 per capita 25 . Roughly three-out-of-five people (58%) live in rural areas where roads, schools and health facilities are scarce, and it was estimated that less than 30% of the population lived within 10km of a health facility in 1990 26 . Having children early in life exposes adolescent women to unnecessary risks: their chances of dying during childbirth is twice as high as that of women who wait until their twenties to begin childbearing 27 . In Mali, the adolescent birth rate per 1000 women aged 15-19 is 174 and this is second only to Niger 28 .

Guinea
Guinea has a population of 12.4 million. Its large deposits of mineral wealth (bauxite, gold, diamond) potentially makes Guinea one of the richest countries in Africa, but it still remains poor with a gross national income of US$ 2,270 per capita, with 64% of the population living in rural areas 29 . In December 2013 it was the first country affected by an outbreak of the Ebola virus. Weak surveillance systems and poor public health infrastructure contributed to the difficulty in containing the outbreak, which meant that it quickly spread to Guinea's bordering countries of Liberia and Sierra Leone 30 . While there did not seem to be biological sex differences regarding vulnerability to Ebola, many of the sociocultural and healthcare related factors increased the risks for women 31 . The adolescent birth rate per 1000 women aged 15-19 is 146 and ranks fourth worldwide 28 .

Liberia
Liberia went through 14 years of civil conflicts which ended in 2004, but had a devastating effect on its population and infrastructure. Liberia is currently in recovery and rebuilding, and has one of the fastest growing populations in the world with a current population of 4.7 million 32 . However, with a gross national income of US$710 per capita, the majority of its population remain poor: it has been estimated that 84% of Liberians live below the international poverty line 28 . This has been made worse by the Ebola epidemic which overwhelmed the already fragile healthcare system 33 . Besides, one in three adolescent girls aged 19 in Liberia is either currently pregnant or already experiencing motherhood.
Additionally, the adolescent birth rate is 149 per 1000 live births 28.

Maternal and Newborn Health in Mali, Guinea and Liberia
The maternal mortality ratio (MMR) is a measure of risk of mortality due to pregnancy and  Table 1 shows the maternal mortality ratio and neonatal mortality rate for the three countries included in the study. For comparative purposes we also present the rates for Sub-Saharan Africa (excluding high income countries) and the Euro-zone area (comprising of the 19 EU member states that share the common currency), over the period from 1990 to 2015. While these show the general pattern of improved health access leading to lower rates of mortality (with reductions of roughly 50% over the 15 year period), there are remarkable differences between the sub-Saharan region and the Eurozone area maternal and neonatal statistics. Although there have been reductions in the rates over time, the levels of neonatal and maternal mortality in the study countries remain unacceptably high. The rates in the Euro-zone area are roughly 100 times smaller than in sub-Saharan Africa. Further, the three study countries all have higher maternal and neonatal mortality than the sub-Saharan (average) rate, highlighting that women and children in these countries in fact experience worse outcomes than others within the sub-Saharan African region. Notes: Source -World Development Indicators 34 .
MMR is the maternal mortality ratio per 100,000 live births. NMR is the neonatal mortality rate per 1,000 live births.

Analytical methods
This study uses exploratory spatial data analysis techniques to estimate and visualize the spatial distribution of women of reproductive age , but with a focus on live births and pregnancies to women aged 15-19 and 40-44 years old. The choice of women in these two age groups in this study was due to the fact that they are women at most risk during pregnancy and childbirth complications due to young age or old age. Data were analyzed using ArcGIS version 10.6. We utilized the clip function tool to extract the age structure dataset of each of the three countries of interest from the high resolution age-structured population distribution map of Africa, while maintaining the clipping geometry and extent of the district administrative maps.
These high resolution maps of age structure, births and pregnancies were then uploaded into ArcGIS to obtain the number of women of 'at risk' reproductive age (i.e. aged [15][16][17][18][19] and 40-44 years old), live births and pregnancies per 100mx100m grid square. To examine the distribution of the outcome variables, we use descriptive chloropeth maps using the ArcGIS software to obtain the number of women of reproductive age, live births and pregnancies per grid square. These chloropeth maps use shading in proportion to the magnitude of the outcome variable to highlight areas with high prevalence (lighter shades represent low prevalence while darker shades represent high prevalence), and thereby can be used to visualize the spatial distribution and identify inequities.
Buffer analysis was then performed to quantify the proximity of pregnancies to health facilities. The buffer is a zone of specified radius or width around a selected map feature or raster of grid cells measured in distance. This geographic location or buffer zone allowed us to estimate the proximity of pregnancies to health facilities through calculation of distances to the nearest health facility. Following the methodology developed by WorldPop project 2 , we then created buffers of 50km radii around each facility since this is approximately equivalent to 2 hours travel time by motorized transport (based on the standard yardstick used for national disparity assessment of access to health services for international comparison 35 ). This is based on the well-established link between travel distance and health access, and the fact that travelling further is associated with health outcomes, particularly in terms of pregnancy and childbirth.
These buffer zones were then overlaid to each country's pregnancy dataset and the numbers within the 50km buffer calculated to obtain an estimate of pregnancies with 'access to (adequate) maternal healthcare'. In contrast, the number of pregnancies residing outside these buffers provided an indication of those 'without (adequate) access'.
The percentage of pregnancies that fell within and outside 50km of a health facility was calculated using the erase, clip and zonal statistics tools in ArcGIS; and the results were mapped to provide an understanding of the geographic and spatial variation of health accessibility and highlight inequities in maternal health service provision. In simple terms, this measure of proximity is used map the areas with women and babies at risk due to having to travel more than two hours in order to access life-saving interventions. Since the choice of a 50km buffer zone did not fully represent the difficulty of travel between two locations (for this we would require impedance information that would reflect the most efficient route to the nearest health facility 35  To provide useful interpretative results from the buffer analysis we grouped districts within regions. We first computed the proportion of pregnancies and births that were within 50km (and 25km) of health facilities, and summed these at a regional level to take account of any spatial heterogeneity effects due to small numbers. Through this analysis we can visualize the distribution of pregnancies, births and health accessibility, and identify locations with clusters of live births and 'at risk' women, at sub-regional levels of geography (these are districts in Mali and Liberia and prefectures in Guinea). We will now provide some information to familiarize the reader with the unique geographical and administrative context of each country. Mali is divided into ten regions, and one capital district (the Bamako district). Each of these regions are divided into 56 districts referred to as 'cercles'. Guinea is divided into 8 administrative regions which are further divided into 34 sub-regions known as 'prefectures'. Liberia is divided into 15 administrative regions, referred to as counties. These counties are divided into 108 districts. Figure 2 shows the administrative geographical divisions for each of the three study countries.

Spatial distribution of Live Births
For the spatial distribution of births, there were clusters of live births in the capital cities of Mali, Guinea and Liberia, as can be expected due to the large population resident in the urban capital areas. In Mali, the minimum number of live births per km 2 was as low as zero

Quantifying the proximity of pregnancies to health facilities
After mapping the spatial variation in the pregnancies, we undertook a buffer analysis to measure the proximity of pregnant women to health facilities. Results of the buffer analysis for Mali revealed disparities in health facility coverage and access in many districts. As shown in Table 2, all the pregnancies in Bamako were within 50km and 25km of a health facility indicating adequate access. In contrast, all the pregnancies (100%) in Kidal region were outside 50km (and consequently also outside 25km) in distance to a health facility, implying that pregnant women had to travel more than 2 hours to access healthcare services. There was some additional spatial variation in access based on the region the women were located in. In Timbuktu the majority of pregnancies (83%), over two-thirds in Mopti (72%) and Kayes (71%), and more than half in Koulikoro (62%) and Gao (54%) were outside 50km of a health facility. Similarly, more than three-quarter of pregnancies in the districts of Timbuktu (90%), Segou (79%), Mopti (79%), Kayes (85%) and Sikasso (77%) were outside 25km buffer zones, while over two-thirds of pregnancies in Gao (69%) were outside 25km distance of a health facility.
The visual maps (Figures 7a, b) and zonal statistics (Table 3) showed spatial variations in health facility access across the various districts of Guinea. Estimates from the zonal statistics indicated that 100 percent of the estimated pregnancies in Conakry were within 25km of a health facility. Out of all the pregnancies outside 50km and 25km buffer zone, the highest percentages were observed in Mamou region (46% and 89%) Labe (39% and 62%) respectively. Results also revealed that one out of every four pregnancies in Boke (25%) was not within 50km of a health facility and the lowest percentage (5%) was found in Kindia. For pregnancies outside 25km buffer zone, about 3 out of 5 occurred in Labe (62%) and Boke (59%), half in Kanakan (50%), one third in Farana (32%) and 1 out of 4 in Nzerekore (26%). Note: *Only regions with pregnancies outside 50km and/or 25km of a health facility are displayed The percentage of estimated pregnancies outside 50km distance to a health facility in Liberia was highest in Grand Gedeh (14%) lowest in Sinoe. Meanwhile in Bong, Lofa and Margibi regions, all the pregnancies were within 50km radius. More than half (52%) in Grand Kru, roughly 2 out of 5 pregnancies in Rivercess (42%), Sinoe (39%) and Grand Gedeh (40%), and one-quarter in Lofa (25%) and slightly over one-third in River Gee (35%) and were outside 25km buffer zone. The lowest percentage (0.3%) was found in the Bong region.

Discussion
The focus of this study was to describe and visualize the spatial distribution of women of proportion of women (84%) in Mamou did not receive maternal healthcare after child birth which is crucial for the survival of the mother and newborn 38 .This is an indication of lack of adequate access to maternal healthcare services, which could be due to the long distance (of 50 and 25 kilometres) that the women had to travel to access healthcare.
Furthermore, according to the WHO, 57% of health facilities were rated to be 'poor' in Guinea 39 . In addition, the same report found that there was a large concentration of health workers in the urban areas. For instance, 16% of the population live in the capital city of Conakry, but roughly half of all health professionals live there. There is therefore a major shortage of adequately trained health force, particularly in the rural areas. As one of the three countries largely affected by Ebola, and the first to have a recorded case, Guinea has been particularly affected by the disease. The first cases were recorded at Gueckedou, in Nzerekore region located in the south of the country. The region is near the borders of Liberia and Sierra Leone, and has been left impoverished by the civil unrest in Guinea, and the neighbouring countries. Its health infrastructure has also been severely damaged. Another unintended consequence of Ebola was that health workers have borne the brunt of infections: health workers were up to 32 times more likely to be infected 40 .
From the maps of Liberia (Figures 3e and f, 4c and 5c Consequently, the people living in rural areas particularly were under-served; and as shown in our analysis one such area is Lofa county, located on the border with Guinea, which happened to be where the first Liberian Ebola case was recorded.

Strengths and limitations of the study
Our study has admittedly some limitations, and highlights further areas of work. In the main, our analysis did not include an exhaustive listing of health facilities (this would have required an inordinate amount of time and resources to locate all primary, secondary and tertiary health facilities), due to unavailability of data on emergency obstetric and newborn care in all the countries. In addition, calculation of standard or actual travel time is beyond the scope of this study. However, though there are uncertainties in the health facility and travel times data used in this study due to unavailability of comprehensive health facility data, the linkage of pregnancy data to datasets with the location of health facilities clearly showed that there were localities where increasing numbers of pregnancies and births have not been matched by commensurate increases in the availability of appropriate health facilities. This finding was also highlighted by the WorldPop project team 2 . This therefore suggests that inadequate maternal healthcare services and distribution of healthcare providers is leading to poor maternal health in these countries.
Another limitation is the fact that the majority of births (60%) occur outside health facilities, and a significant proportion of these births are delivered without complications with the help of skilled and unskilled birth attendants 45 . However, the need for access to maternal and newborn care is important for emergency interventions when complications do arise, and geographical access becomes a barrier to safe motherhood 15 . Nonetheless, our study adds depth to the current knowledge from small area estimation of maternal and newborn outcomes by extending the work to subnational level (districts) in West Africa.
Data generated would be useful for effective planning to promote safer pregnancies, births and healthier newborns and equitable distribution of human resources and infrastructure, thus bridging the gap in health inequity within and across the countries.

Conclusions
This study has revealed spatial variations in the distribution of women of reproductive age, pregnancies and births, and access to healthcare services at district levels in Mali, Guinea and Liberia., and this can be related to the numerous findings11, 46,47,48 which have shown that proximity to health facilities is important in accessing maternal and newborn services. Our study found that in urban areas geographic distance does not appear to affect access to healthcare. However, in rural and remote areas, where transport infrastructure tend to be lacking and weak, there are geographic barriers to