Despite the prominence of international targets for maternal mortality reduction, around a third of a million women continue to die annually from complications of pregnancy or childbirth and many more suffer prolonged or permanent post-partum ill health or disability [1–3]. The greatest share of this burden, around 90%, is borne by developing nations of sub-Saharan Africa and South Asia [2, 4]. Limited access to health facilities staffed by appropriately trained personnel and offering midwifery competencies and life-saving obstetric interventions at birth is known to be a key driver of maternal mortality [5, 6]. Unlike the vertical programmes of intervention that can be effective in combating other global health challenges [7–9], improving access to effective care at birth relies on strengthening health systems and this partly explains the slow pace of improvement in access and resulting mortality reductions relative to some other global health targets [10, 11].
Access to care at birth is determined by a diverse set of factors related to both the services offered (such as their availability, quality and cost) and the population being served (such as their wealth, education, and culturally-mediated perceptions) [10, 12, 13]. Linking population and health system factors is geography: the physical accessibility of facilities to women in labour. These factors interact in potentially complex ways and in some settings, such as urban areas with robust health infrastructures , geographical distance may play only a minor role in determining levels of access and subsequent health outcomes. However, where service provision is sparse, transport infrastructures weak, and populations predominately poor, geography often presents a fundamental and insurmountable barrier to accessing adequate care at birth and therefore plays a central role in sustaining high maternal mortality [15–22]. This is manifest both in those women who die having not sought facility-based care as well as in the significant number that die en route to health facilities, en route from one hospital to another with more appropriate resources, or after arriving too late [20, 23, 24]. The delays in (i) making the decision to seek care, (ii) reaching an adequate health facility and (iii) receiving the needed care within a facility limits uptake of emergency obstetric care [11, 20, 25]. While delays (i) and (iii) have received much research attention, delay (ii) has not been studied systematically [26–28].
Given the importance of geographical access, its measurement should arguably form a central component of maternal health system assessment and strategic planning, as well as providing a key development target indicator. Measuring geographical access robustly is, however, fraught with data and methodological challenges and the result is that policy-makers typically revert to crude alternatives such as regional facility-population ratios [29, 30]. These ratios are potentially biased and inadequate proxies for auditing the true number of women able to access care at birth and therefore hamper progress towards solutions.
Assessing geographical accessibility over large regions is problematic for several reasons. First, the data requirements are considerable and rarely met in developing countries. Comprehensive data are required on the geographical distribution of both the population and the health facilities to which mothers must travel to access care. Such data sets at the required level of spatial detail, contemporariness, and completeness are the exception rather than rule in SSA . Second, it is well established that straight-line distances act as a poor proxy for the actual cost (distance, time, expense) of journeys [32–35]. This precludes straightforward Euclidean analysis of access and necessitates that the actual landscape across which journeys are made, and the availability of different means of transport, must be known with some degree of detail [36, 37]. These factors combine to mean that, whilst the importance of distance to maternal and newborn health services has been demonstrated by many small-scale studies [15–22], the detailed measurement of geographical access to maternity care at birth across national populations that might support country-level strategies for scaling up care is rare. Of only two published studies at a national level that are known to the authors, one focuses exclusively on emergency referrals rather than population access to care  and the other uses a simple distance measure that is not necessarily representative of actual journeys faced by women in labour .
Ghana is representative of most countries in sub-Saharan Africa (SSA) in that rates of maternal mortality remain unacceptably high, with estimates for 2008 of 350 (range of uncertainty 210–630) maternal deaths per 100,000 live births . It is also typical in that (i) a significant proportion of births continue to occur at home without a professionally trained health worker, (ii) the fraction that give birth at home increases dramatically with distance from main urban centres, and (iii) progress in reducing maternal deaths remains slow [6, 40]. Geographical access plays a characteristically large role in limiting uptake of maternity care services in Ghana, especially at the time of birth, and was the most commonly cited reason for non-attendance in a recent national survey . Further, a review of 322 maternal deaths occurring in Ghanaian health facilities in 2011 found delay in arrival at a health facility to be a contributing factor in nearly half (46%) of cases. Distance, rather than decision making at home, is strongly implicated in many of these delays .
In this study, we use the example of Ghana to present the first detailed national-scale assessment of geographical access to maternity care at birth in a high burden country based on a calibrated journey-time model. This represents one outcome of a four-year study that has included the assembly of comprehensive population, health facility, and landscape data, augmented by multiple existing national and sub-national population sample surveys, and the development of a geospatial framework for modelling realistic journey-times. We use these components to demonstrate the current geographical accessibility of three levels of care at birth in Ghana and identify populations where this is dangerously inadequate. We also compare these results to existing metrics of access, based on regional facilities-per-capita ratios, currently in use by decision-makers.