Distance is “a big problem”: a geographic analysis of reported and modelled proximity to maternal health services in Ghana

Maternal mortality has declined globally, but progress in reducing pregnancy and childbirth-related deaths in Sub Saharan Africa is slow [1]. Maternal deaths have reduced by 38% within approximately two decades from the year 2000, but the gradual annual rate of change (2.9%) is not sufficiently rapid to achieve the Sustainable Development Goal for reducing maternal mortality by 2030 [2]. Most maternal deaths (94%) happen in poorly resourced settings [3], and Sub Saharan Africa has the highest estimated mortality, which is declining only gradually. Like its neighbours, Ghana has a high maternal death rate, declining by only 2.7% every year. The distribution of maternal deaths by location and socio-demographic group can be uneven [4].

The leading causes of maternal mortality are severe bleeding after childbirth, pregnancy-induced high blood pressure and infections after childbirth [3, 5]. These causes are responsible for three-quarters of maternal deaths. Therefore, the primary interventions to prevent maternal deaths are quality emergency obstetric care provided by skilled health professionals during pregnancy, childbirth and the first 2 days after birth [6].

Pregnancy and childbirth-related complications can be identified and managed when women receive timely quality health services, but they have to overcome geographic and transport barriers to access care [7]. Even among populations with higher socio-economic status, health services might not be available or accessible [8], leading to complications and death. In several locations across Africa, maternal health services are unavailable, inaccessible, and often below the recommended geographic accessibility requirements [9].

Proximity to quality health services is a crucial determinant of maternal mortality, stillbirths, and child mortality because children and women living closer to health facilities have better health outcomes [10‐12]. Women living farther from health facilities might either not use them [13] or reach the health facility late with a life-threatening complication. Cost of transportation, lack of public transport, and poor road networks are other journey-related factors that could hinder women’s ability to reach a health facility [14].

Geographic accessibility is one of the main barriers that inhibit women from using maternal health services in Africa [15]. Many Sub-Saharan Africa countries have at least one sub-national area lacking at least one hospital per half a million population [9]. About half of Ghanaian women live two or more hours’ drive from health facilities providing comprehensive obstetric care [16]. Thus, it is unsurprising that some women in Ghana describe the distance they travel to access health services as a “big problem” [17].

Previous studies have compared self-reported geographic healthcare access measures and modelled travel times to nearest health facility [18‐21]. Different approaches were used to measure self-reported distance and modelled travel times. Self-reported healthcare proximity was deduced from journey start/end times [18, 20] or distance reported by patients or their relatives [19‐21]. The modelled travel times were based on road networks or impedance grid cells estimating the cost of travel. Generally, these studies have found that modelled travel times are lower than self-reported travel times [19‐21]. These comparisons have often been based on self-reported measures of geographic access available only in one country or sub-region via a research study, rather than on a self-reported geographic access measure incorporated into an international household survey series.

As part of the Demographic and Health Surveys (DHS) implemented in several countries, women are asked whether the distance to the health facility is a big problem when visiting a health facility with illness [22]. Some studies have relied on this DHS question to assess the uptake of maternal health services. Their results found self-reported challenges with distance were associated with lower use of antenatal, skilled birth and postnatal services, but none compared self-reported and modelled distance [23‐25]. Furthermore, studies comparing self-reported versus modelled travel times [18‐21] have not evaluated the standard DHS question by comparing self-reported distance challenges to modelled travel times. Therefore, there is currently no evidence for the robustness of the DHS question concerning distance-related challenges accessing healthcare.

This paper firstly aims to assess the relative effects of modelled travel times and self-reported problems with distance to healthcare on skilled attendance at birth. Secondly, it aims to assess socio-demographic factors associated with women reporting problems with distance to healthcare.

This is the first study comparing modelled travel times with distance challenges reported by women via the DHS survey questions. Our comparison found that women reporting distance as a challenge in accessing healthcare had a significantly longer modelled travel time to the nearest health facility. Generally, women with poor socio-demographic backgrounds had longer travel times to their nearest health facility. The travel times varied by region and urban areas had shorter journey times. Women with better maternal health outcomes (skilled birth attendant, four or more skilled ANC appointments and timely skilled PNC) had shorter travel times to the nearest health facility. There is thus consistency between modelled travel times and women’s responses to the DHS question on distance challenges accessing healthcare, suggesting the former is a valid construct reflecting geographic healthcare access.

Our findings support the experience of women who said distance to the health facility is challenging when they are sick. These women are likely to be seeking ANC, SBA, PNC and other maternal health services at more distant health facilities than other women. This study and several others found associations between healthcare proximity and maternal health service uptake [25, 37]. For instance, a woman attributed her stillbirth to her community’s lack of a health facility, leading to her travelling further to give birth in a nearby region [38]. Lack of a nearby health facility at reasonable travel times can lead to home births and adverse birth outcomes.

We found modelled travel times underestimate geographic access problems for poor, rural and less educated women. The underestimation of geographic access problems among disadvantaged women is consistent with similar studies [18]. However, our travel times could be lower than the real journey times for these groups of women due to our modelling approach [21]. Poor, rural women may be more likely to experience delays waiting for infrequent public transport such as buses, and public transport may be slower than other forms of mechanised transport. Such delays are not incorporated into modelled travel times. Finally, our modelled travel time assumes a constant travel speed on each road class and does not account for road conditions that might affect the journey time, particularly in rural areas. Incorporating road conditions in travel time models could improve modelled travel time estimates.

Geographical barriers limit access to healthcare and slow universal healthcare progress [39]. Geographic barriers can include the quality of roads, cost of transportation, and the travel time to a nearby health facility where services are available, ready and fit for purpose. To reach health facilities, women must overcome contextually varied geographic barriers [14]. For example, women in rural areas might travel at least twice as far as their urban counterparts because there are likely fewer health facilities, poorer roads, and less equipped health facilities in rural areas. Although health facilities can be closer to urban dwellers, the urban poor might be disadvantaged than the rural poor because of expensive services in nearby private facilities [40]. Comparatively, urban women might live closer to health facilities but have longer travel times due to traffic congestion [41]. In some settings, the needed intervention might be better transportation services in the form of links to major roads, asphalted surfaces, availability of vehicles and affordable cost of travel.

Both geographic access measures are related to SBA in unadjusted models but not following adjustment for other characteristic such as rurality, poverty and education. Maternal and household characteristics can moderate the relationship between distance and the utilisation of services [42]. Health care access transcends geographic access to include other factors such as the cost of services, social access, the quality of care, acceptability amongst other dimensions of access [43, 44]. In Ghana, the cost of services and decision-making autonomy were key determinants of access to health services [45]. Also, maternal literacy, health insurance and household wealth were linked with women’s health seeking autonomy and SBA utilisation [45, 46]. Rural women are predisposed to unfavourable maternal health outcomes than urban women because they are mostly poor, less educated and live farther from health facilities. For instance, poorer women might not afford services; less educated women might be poorer and have lower obstetric risk perceptions. Therefore, addressing the social determinants of health could improve socioeconomic status and might promote better maternal health outcomes [47].

The self-reported distance problems in the DHS correlated with the estimated travel time to the nearest health facility. The results underscore the relevance of perceived challenges or subjective ratings in the DHS. Responses to the DHS survey question on proximity to healthcare services are broadly consistent with modelled travel times, suggesting this construct is valid. Also, women from a cluster location receiving care at the same health facility might rate the distance differently. For instance, Fig. 3A shows that wealthier households reporting distance as a problem lived closer to health facilities than poor households with no distance challenges. A plausible explanation could be self-reported distance reflecting mode of transportation and other difficulties in accessing care than modelled travel time. For example, urban women might reach longer distances quicker in a mechanised vehicle than urban women travelling shorter distances on foot. Therefore, the question should be maintained and possibly expanded to include perceived quality of maternal health care. Were a service quality question to be included in the DHS, it could be evaluated through spatial linkage to service provision assessments. If patient-perceived quality of maternal health care correlates with service provision assessment quality indices, they can inform spatially targeted healthcare quality improvements.

This study combines different data sources to evaluate proximity to maternal health services. We joined routine health data, DHS, and travel time models to assess proximity to health services, similar to other studies integrating spatial data [48]. Each data source has its relevance. First, routine data shows where services exist by examining the counts of service users captured in health management information systems. Then, service provision assessments measure the quality of the services.

Furthermore, household or neighbourhood wealth characteristics can be derived from census, surveys or remote sensing [49]. Finally, proximity measures can be estimated using spatial models implemented in this study. These data sources present opportunities to explore and unpack spatial inequalities among rural-urban, urban poor and other sub-groups. Through such integrated analysis, areas with poor access to care will be apparent for locating new facilities or upgrading poor ones.

The current results are specific to Ghana and could change when new health facilities are built, or road conditions improve before subsequent surveys. However, the analysis can be expanded to include other countries in Sub-Saharan Africa because the DHS survey is widely available. Also, the analysis could be conducted using multiple DHS waves to assess changes in reported distance problems. Global travel time to health facility data [50] is essential for large scale sub-regional analysis comparing many countries. However, travel time to all health facilities can only be used as a proxy for access to specific health services because health facilities can be distinct in the services they provide. Therefore, for contextual analysis investigating access to maternal health services, studies can rely on routine health data to identify locations where services are available to estimate journey times.

We used the most recent survey data available and relied on known health facility locations to measure travel times. Also, we relied on location information from the DHS survey cluster locations to estimate the respondents’ proximity to the nearest health facilities. Following accepted practices, the travel time estimates described the modes of transport (hybrid walking and motorised), origin (DHS cluster location) and destination (nearest health facilities) [51]. We used travel time estimates because they account for changes in elevation and the mode of transportation.

Travel times and distance were measured to the nearest health facility because the DHS does not name health facilities. Hence, modelling proximity to the nearest health facility can be problematic as some women might bypass the nearest facility and travel longer [52]. Furthermore, the random displacement of cluster locations to preserve confidentiality could affect the travel times, but recommended distance buffers were used to moderate their effect [31]. Additionally, we did not account for the possible variations in travel times due to travel disruptions in the rainy season, road conditions, traffic congestion in urban areas and the type of transportation.

The outcome variable, “distance as a problem”, depends on the ability of the respondent to recall events rendering the response susceptible to bias. Furthermore, there could be question order bias with the importance of proximity reduced because distance to the health facility is the third question among the four assessing barriers to care. The four questions cover getting permission to see the doctor, getting money for advice or treatment, distance to the health facility and not wanting to go, in that order.

This study combines different data sources to evaluate proximity to maternal health services. We joined routine health data, DHS, and travel time models to assess proximity to health services similar to other studies integrating data. The study finds a correlation between perceived and modelled proximity to health facilities. The findings unpack the inequalities in access to maternal health services and their impact. The impact of proximity is lessened when socio-demographic characteristics are improved. Hence, efforts to improve maternal health outcomes and coverage should target health facility location and the livelihood of households.