Menu
Menu
Menu
Menu
Background: Access to skilled attendance at childbirth is crucial to reduce maternal and newborn mortality. Several different measures of geographic access are used concurrently in public health research, with the assumption that sophisticated methods are generally better. Most of the evidence for this assumption comes from methodological comparisons in high-income countries. We compare different measures of travel impedance in a case study in Ghana’s Brong Ahafo region to determine if straight-line distance can be an adequate proxy for access to delivery care in certain low- and middle-income country (LMIC) settings. Methods: We created a geospatial database, mapping population location in both compounds and village centroids, service locations for all health facilities offering delivery care, land-cover and a detailed road network. Six different measures were used to calculate travel impedance to health facilities (straight-line distance, network distance, network travel time and raster travel time, the latter two both mechanized and non-mechanized). The measures were compared using Spearman rank correlation coefficients, absolute differences, and the percentage of the same facilities identified as closest. We used logistic regression with robust standard errors to model the association of the different measures with health facility use for delivery in 9,306 births. Results: Non-mechanized measures were highly correlated with each other, and identified the same facilities as closest for approximately 80% of villages. Measures calculated from compounds identified the same closest facility as measures from village centroids for over 85% of births. For 90% of births, the aggregation error from using village centroids instead of compound locations was less than 35 minutes and less than 1.12 km. All non-mechanized measures showed an inverse association with facility use of similar magnitude, an approximately 67% reduction in odds of facility delivery per standard deviation increase in each measure (OR = 0.33). Conclusion: Different data models and population locations produced comparable results in our case study, thus demonstrating that straight-line distance can be reasonably used as a proxy for potential spatial access in certain LMIC settings. The cost of obtaining individually geocoded population location and sophisticated measures of travel impedance should be weighed against the gain in accuracy.
Ghana is a West African country with a high maternal mortality ratio estimated at 328 per 100,000 in 2011 [26]. The study area consists of 7 contiguous districts with a population of more than 100,000 women of reproductive age (14-45 yrs), where demographic surveillance was established for several field trials [27-29]. Travel occurs on roads, and mainly on foot to the closest health facility, as reported by approximately 58% of households in a 2003 national survey [30]. A geospatial database of the study area was created, mapping population location in compounds and village centroids, service locations for all health facilities offering delivery care (including higher level facilities with capacity for surgery), and a detailed road network (Figure 1). We included administrative boundaries and topography (land-cover, including water bodies) [24,25]. We combined data sources in a workflow (Figure 2), and describe the fieldwork in more detail here: Study area showing topographic cover in Brong Ahafo region, Ghana. First inset shows study area in Ghana with administrative divisions. Second inset shows detail of example village with centroid, compounds, road network and a delivery facility. Workflow for geospatial analysis. 1. Road network A detailed road network of all roads in the study area was created using GPS trackers. The study area covers approximately 15,302 km2 and our road network includes over 1,900 km of roads. Extensive deskwork was done in order to transform these road tracks into a network dataset appropriate for analysis in ArcGIS, ensuring functional connectivity between roads. A tool was developed in a PostGIS geodatabase to validate the connectivity of the network roads, and the road network was subsequently cleaned in GRASS GIS [31]. The road network was then integrated into the land-cover raster layer for analysis using a 200 m resolution. Additional information on road condition, surface type, and usability in the rainy season was collected for all roads. Travel times by vehicle between village centroids were collected for one study district. A total of 88 journey segments were used in order to calibrate road speeds, which were assigned with reference to speeds used in the literature [32,33]. Road speeds ranged from 30 km/h on dirt roads, to 90 km/h on good tarmac roads. Very few roads (four in the study area) were reported as impassable during the rainy season, so we model the dry season scenario only. 2. Health facility census We conducted a health facility assessment of all 86 geocoded health facilities in the study area to categorize facilities according to the availability and quality of maternal and newborn services: 64 facilities offered delivery services and 8 offered comprehensive emergency obstetric care (CEmOC), i.e. higher-level facilities with the capacity for cesarean section and blood transfusion [34-36]. The majority of the hospitals, health centers, and clinics with delivery care are publically owned, and all maternity homes are operated privately by the Ghana Registered Midwives Association. 3. Surveillance Surveillance of all women of reproductive age in the study area through monthly visits was undertaken as part of health and demographic surveillance for several field studies [27,28]. The surveillance included taking GPS coordinates of 433 village centroids and, in 173 larger villages, coordinates of 47,537 individual compounds (with a median of 450 compounds per village (IQR 258–844, max 3,204)). For the analysis of facility use (objective 3), we included villages and compounds where deliveries occurred in 2009 with known birthplace and compound coordinates, resulting in 169 villages, 8,120 compounds and 9,306 births. There was a median of 96 births per village (range 1–634), and a median of 1 birth per compound (range 1–8). All six impedance measures were calculated to two levels of care, distance to closest facility with delivery care, and distance to closest facility with CEmOC (Table 2). In ArcMap version 10.0, we used the Spatial Analyst tool “Near” to calculate Euclidean distances and the Network Analyst tool “Closest Facility Analyst” to calculate network distance and time (ESRI software, California). For the raster-based analyses, we used the cost surface algorithm in GRASS GIS to determine the fastest route (least-cost path) from starting points to given destinations [31,37]. To address objective 1, we used Spearman rank correlation coefficients to compare the six impedance measures within each origin destination pair (i.e. village centroid to closest delivery facility and village centroid to closest CEmOC in a dataset of all villages; compound to closest delivery facility and compound to closest CEmOC in a dataset of all compounds; Table 2). We assessed potential spatial aggregation error (objective 2) in three ways using the surveillance dataset. First, we compared the correlation of the measures calculated from the two origins, and then whether the different measures identified the same facility as closest from both origins for each birth. Finally, we calculated distance deviance, the absolute difference in distance or time between measures starting from compounds compared to measures starting from village centroid for each birth. These absolute differences represent the potential error in access estimates that result from using average village centroids as opposed to individual compound coordinates, and are dependent on the dispersion of villages. Spatial access to health care is known to be a facilitator of delivery in a health facility [3]. The impedance measure that is the best proxy of spatial access to delivery care, i.e. has the least measurement error, should then show the strongest association with facility delivery in a regression model (objective 3). We modeled the association of each impedance measure with whether or not a woman delivered in a facility as a binary outcome variable, in a logistic regression model for all births in the study area. For ease of comparison between measures with units in distance and time, we standardized our impedance measures to have a mean approximately equal to zero and standard deviation (SD) of one. In order to account for clustering of women by village, we used logistic regression models with robust standard errors. All analyses were done in Stata version 12.0. This study uses data collected for the Newhints trial, which was approved by the ethics committees of the Ghana Health Service, Kintampo Health Research Center and the London School of Hygiene and Tropical Medicine (LSHTM) [28]. The additional analyses were approved in an amendment by the LSHTM ethics committee.
N/A
