Mapping for maternal and newborn health: the distributions of women of childbearing age, pregnancies and births

Improved understanding of geographic variation and inequity in health status, wealth, and access to resources within countries is increasingly being recognized as central to meeting development goals. Approaches based on local epidemiological and coverage data have been identified as vital to reducing childhood mortality for Millennium Development Goal (MDG) 4 [1], while the sub-national heterogeneity in HIV [2, 3] and malaria [4, 5] prevalences mean that effective targeting of interventions remains vital in achieving MDG 6 [6]. Indicators assessed at national level can often conceal important inequities, with the rural poor often least well represented [7, 8].

The fifth MDG, which targets maternal health, is widely recognised as being the most off-track of the goals endorsed by leaders in 2000 [9]. Moreover, with a little more than two years to go before the MDG deadline, maternal mortality is still the most dramatic indicator of global health inequity with 99% of these deaths occurring in developing countries, and large numbers likely going unrecorded. For every death, at least another 20 women suffer illness or injuries related to childbirth or pregnancy [10]. Mortality rates for newborn babies have also been slow to decline compared with death rates for older infants. Known solutions, such as the provision of a skilled health worker at births, are effective in tackling both maternal and newborn health problems, but scaling up services is still hampered by a lack of basic subnational data. The lack of spatial datasets to aid in identifying the magnitude of inequities for women and newborns both in outcomes and services are now a key constraint to progress [11, 12]. Moreover, as international funding for health and development comes under pressure, the ability to target limited human resources and health services to underserved groups becomes crucial.

For maternal and newborn health (MNH) and survival there are three key sets of information that are currently lacking or poorly provided – that could be better estimated at small geographies to inform planning and policy and to pinpoint communities in need: (i) Planning for safer births and healthier newborns would be improved firstly by providing more accurate estimations of population distributions – targeting particularly women of childbearing age. (ii) Better projections of the approximate numbers of actual pregnancies that are likely to occur in the short and long-term are needed for more effective strategies on human resources and infrastructure. (iii) There is a need to link information on pregnancies to better information on health facilities in districts and regions so that distance to services and their coverage can be assessed. The latter is now becoming feasible given the move towards geo-referencing hospitals and health centres, and while there are many other factors that influence service utilisation, it should be a key indicator of progress given the importance of geographical access to services for the survival of women and newborns with complications [13, 14]. The GIS techniques that can be applied to derive these three needed sets of estimations have been applied only sporadically in the area of maternal and newborn health [15]. Despite some progress in very recent years (e.g. [16]), it is recognised that there is insufficient global attention paid to disaggregating national data by geographical units [17] and the MNH community has yet to fully capitalize on the emerging capacity of GIS.

Here we present methods to estimate firstly women of childbearing age and secondly pregnancies and live births in relation to current health infrastructure in four countries (Afghanistan, Bangladesh, Ethiopia and Tanzania) with high maternal and neonatal mortality, and details on the MNH situation in each country are provided in the next section. The methods presented here are based on the integration of satellite, census, household and Emergency Obstetric and Newborn Care (EmONC) survey data for the construction of high resolution datasets depicting estimates and distribution of the number of people, women of childbearing age, pregnancies and live births per 100×100 m grid cell, plus locations of EmONC facilities. The application of these methods for constructing such datasets is presented, followed by a simple demonstration analysis in linking the likely distribution of need (pregnancies) compared with services to highlight geographic inequities in service provision.

The methods and calculations originate from health systems and policy research in the four countries, undertaken as part of the High Burden Countries Initiative (HBCI) supported by UNFPA and its UN partners in the H4+ [18]. As part of their broader responses to the UN Secretary-General’s Global Strategy for Women’s and Children’s Health, the UN health agencies - “H4+” (UNAIDS, UNFPA, UNICEF, World Bank, WHO), alongside the government and development partners, have initiated national assessments of the midwifery workforce in 8 countries (Afghanistan, Bangladesh, Democratic Republic of Congo, Ethiopia, India, Mozambique, Nigeria and the United Republic of Tanzania), representing nearly 60% of the global maternal and newborn deaths. All four countries are listed amongst the 10 countries with the highest burden of maternal and neonatal deaths per annum [19–21]. The work here describes the development of the methods needed in preparation for a larger midwifery workforce assessment that seeks to generate new evidence on ‘what is the appropriate midwifery workforce, and how is it best deployed, to equitably deliver essential MNH interventions at scale and quality, and what (including costs) needs to be put into place to achieve universal access?’ Each of the country assessments follows an Operational Guidance and Assessment Framework [19] and includes a module on the ‘Geography of MNH’ to estimate and map where women of child-bearing age are living and how many pregnancies are likely to be occurring. This provides a proxy for population need for MNH services that can ultimately be linked with georeferenced facility data to compare the current supply-side capacity of health facilities and workforce deployment to identify inequities in access to midwifery care and services.

In order to complete the Geography of MNH module we developed new methods using a geographical information system (GIS) based approach, enabling the construction and combination of spatial data layers. Firstly we construct population distributions at subnational level and provide estimates of women of reproductive age per grid cell. We then estimate numbers of pregnancies and live births for each grid cell using national household survey data on age specific fertility rates, published estimates of live births from the UN Population Division [22] and new estimates of stillbirths, miscarriages and abortions from the Guttmacher Institute [23]. A third layer of the geographical location of health facilities providing basic and comprehensive emergency obstetric and newborn care is then added. Finally we demonstrate the utility of these layered datasets in estimating proximities from pregnancies to health facilities – linking estimates of population need with the locations of facilities designed to meet this need.

Mapping and the application of GIS in many health related fields is becoming sophisticated and demonstrated as a valuable tool for providing evidence to guide strategies, but its application in MNH is lagging behind. The midwifery workforce assessments sought to integrate the capacity and added value of GIS technology and the approaches outlined here for population, age group, pregnancy, birth and facility mapping to produce gridded surfaces, with each 100 m by 100 m grid cell providing an estimate of the number of people, live births, and pregnancies within it. Such datasets provide exceptional flexibility in terms of enabling summarization to any level required. This may include, for example: (i) summing all grid square estimates within district boundaries to provide district-level estimates of number of pregnancies, (ii) usage of a map of urban areas overlaid onto the gridded data to provide per-city estimates of number of pregnancies, or urban rural proportions, and (iii) integration of the gridded datasets with health facility locations and road network data within travel models to quantify numbers of pregnant women residing at more than two hours from a comprehensive EmONC facility. In each of these example cases, simple national averages that do not account for subnational geography cannot reveal the spatial heterogeneities that exist in this way.

While it has been shown that accounting for sub-national heterogeneity in population attributes likely results in significant improvements in the accuracy of health metrics [33], it is clear that many sources of uncertainty and error remain. Primarily, it is clear that uncertainties in the output maps increase and accumulate at each stage of the process. While the location and number of total people and women of childbearing age subgroup are known with a relatively good degree of precision, the subnational variations in birth rates are less precisely known, while subnational information on pregnancies, abortions and stillbirths do not currently exist. This means for pregnancies, a reliance on the assumption of no spatial variation within a country on rates of stillbirths and abortions through the use of national-level estimates [23]. Further, all of the census and survey-based data used here are subject to various sources of error and bias, many of which have been well-documented [33, 38, 39], while the underlying WorldPop population datasets also contain uncertainties [24]. It is also clear that the sampling frames used in undertaking the DHS surveys mean that uncertainties exist in defining sub-national urban/rural ASFRs, with sample sizes becoming relatively small when summarizing at such levels. Nevertheless, the resultant ASFRs do still match closely with what is known about regional and urban–rural differences in each of the countries, with lower rates in urban areas and the highest rates in the most isolated and deprived rural areas of each country (Figure 3). Like most other population parameters reported for administrative polygons, the population, age, sex and fertility rate data used here are also subject to the modifiable areal unit problem, i.e. that summary measures are influenced by the administrative boundaries that they are reported at [40]. Finally, as highlighted in the methods section, assessments of geographical access to care could be improved through the modelling of travel times, rather than simple straight-line distances.

There is clearly a need to more rigorously quantify the uncertainties inherent in spatial demographic datasets [39], such as those presented here to better communicate the spatial variations in reliability of input datasets and guide prioritization of additional data collection, and future work will aim to tackle this. The advancement of theory, increasing availability of computation, and growing recognition of the importance of robust handling of uncertainty have all contributed to the emergence in recent years of new, sophisticated approaches to the large-scale modeling and mapping of disease based on geostatistics (e.g. [4]) and small area estimation [41], but such methods have yet to cross over to the spatial demographic databases with which such maps are used. The regular availability of new national household surveys means that more contemporary data is continually becoming available to aid in updating and improving the accuracy of the datasets presented here, and future steps will involve the development of semi-automated systems that can rapidly adapt to new incoming data and integrate them into the output spatial datasets, alongside robust methods to account for temporal differences [42]. Further, the linkage of these pregnancy datasets with the location of health facilities and spatial models of travel time will enable improved estimates of the spatial coverage of healthcare to be made.

Despite the limitations and caveats above the results in all four countries provide new intelligence on disaggregated population needs for MNH services and are informing the policy discourses on the distribution of health facilities and health personnel who provide MNH/midwifery services. However, not all countries maintain an accurate, current list of EmONC facilities with georeferenced codes and even fewer maintain a live database on the number, type, competency and skill levels of health personnel per facility. The absence of this basic information therefore diminishes the ability to conduct detailed analysis of supply-side constraints to respond to population need. Improvements in human resource information systems are critically needed and if linked to facility GIS codes would lead to new insights on accessibility to a skilled and competent health worker. Additionally we recognise that comparing need to supply means that we are missing an important step: women’s demand for and utilisation of MNH/midwifery services. In all four countries, coverage of antenatal care, skilled attendance at birth and postnatal care is variable, with significant difference between urban/rural areas and socio-economic quintiles. Barriers to access care are therefore beyond the geographical location of services and require triangulation with other sources of data to appreciate the realities that women experience in seeking, accessing and affording quality care during pregnancy, birth and the post-natal periods.

Thus far, the measurement of health outcomes and service needs has focussed on indicators at country level for member states that signed up to the original MDG goals. But increasingly there are concerns that even where progress has been made, this is subject to very wide inequalities. This means that countries that have progressed well include subgroups in their populations where survival rates and access to services have not changed – or have even worsened. Investigation of subnational situations is therefore needed and a geographical analysis is now increasingly required. An improved understanding of geographic variation and inequity in health status, wealth, and access to resources within countries is increasingly being recognized as central to meeting development goals.

Substantial demographic variations exist across countries and between urban and rural areas [39]. With MDG health indicators tied to specific vulnerable groups, there is a need to know who and where these vulnerable groups are and the numbers of individuals at risk that exist in order to accurately characterize denominators. High fertility is still a feature of many high burden countries and in rural outlying areas where resources are stretched, stalled fertility declines and sheer population momentum mean that an increasing number of births need to be attended. Though facility workloads are affected by many factors, clearly there are localities where increasing numbers of pregnancies and births have not been matched by commensurate increases in the availability of the appropriate MNH workforce, leading to a deterioration of the MNH situation even while the national picture may have been improving. In recent years tracking results and resources in sub national areas are not only more possible but also more necessary.

The methods presented in this paper, notwithstanding their uncertainties, represent a step forward in terms of estimation that can help to quantify progress and problems at very small geographies. The initial work in support of the national midwifery workforce assessments in Afghanistan, Bangladesh, Ethiopia and Tanzania has established the potential value to inform policy-dialogue on geographical gaps in service provision in response to population needs [43]. We envisage that such methods will increasingly inform international and national level evidence-based decision-making where denominators are uncertain, i.e. where civil registration and vital statistics of births and deaths are poor or patchy, and GIS-based techniques such as these are an important part of the data revolution to assist policymakers to improve services and survival. The live birth and pregnancy datasets outlined in this paper are freely available through the WorldPop project website (http://www.worldpop.org.uk).