Estimating the spatial accessibility to blood group and rhesus type point-of-care testing for maternal healthcare in Ghana

Background: In Ghana, a blood group and rhesus type test is one of the essential recommended screening tests for women during antenatal care since blood transfusion is a key intervention for haemorrhage. We estimated the spatial accessibility to health facilities for blood group and type point-of-care (POC) testing in the Upper East Region (UER), Ghana. Methods: We assembled the attributes and spatial data of hospitals, clinics, and medical laboratories providing blood group and rhesus type POC testing in the UER. We also obtained the spatial data of all the 131 towns, and 94 health centres and community-based health planning and services (CHPS) compounds providing maternal healthcare in the region. We further obtained the topographical data of the region, and travel time estimated using an assumed tricycle speed of 20 km/h. We employed ArcGIS 10.5 to estimate the distance and travel time and locations with poor spatial access identified for priority improvement. Findings: In all, blood group and rhesus type POC testing was available in 18 health facilities comprising eight public hospitals and six health centres, one private hospital, and three medical laboratories used as referral points by neighbouring health centres and CHPS compounds without the service. Of the 94 health centres and CHPS compounds, 51.1% (48/94) and 66.4% (87/131) of the towns were within a 10 km range to a facility providing blood group and rhesus type testing service. The estimated mean distance to a health facility for blood group and rhesus POC testing was 8.9 ± 4.1 km, whilst the mean travel time was 17.8 ± 8.3 min. Builsa South district recorded the longest mean distance (25.6 ± 7.4 km), whilst Bongo district recorded the shortest (3.1 ± 1.9 km). The spatial autocorrelation results showed the health facilities providing blood group and rhesus type POC testing were randomly distributed in the region (Moran Index = 0.29; z-score = 1.37; p = 0.17). Conclusion: This study enabled the identification of district variations in spatial accessibility to blood group and rhesus type POC testing in the region for policy decisions. We urge the health authorities in Ghana to evaluate and implement recommended POC tests such as slide agglutination tests for blood group and rhesus type testing in resource-limited settings.

This is a follow-up on our previously published cross-sectional study which investigated the availability and use of pregnancy-related POC tests in the UER utilising a hundred PHC clinics (health centres and CHPS compounds) randomly selected from a total of 356 clinics from all the districts in the region [20]. The sampling strategy used to select the 100 PHC clinics as well as the study area (UER) has been adequately described in the previously published survey [20]. The survey revealed that of the 100 PHC clinics, blood type screening test was available only in six clinics, and all six were health centres with some form of laboratory services [20]. To measure the spatial accessibility to health facilities where blood type screening testing is available in the regions, the cost-distance algorithm was applied using the ArcGIS desktop software. The flow diagram (Figure 1) illustrates the data, methods, and models used. All the attributes and spatial data used for this study were obtained in 2018. A flow diagram illustrating the data, methods, and models used. PHC—primary healthcare. We extracted the attribute data on all the 100 PHC clinics and health facilities where expectant mothers (who require blood group and rhesus status testing) were referred for blood type screening tests during ANC. These data which were obtained in text data format were loaded into the ArcGIS 10.5 software and transformed into a shapefile to allow for performing spatial analysis using ArcGIS. The topographic data gathered for this current study included roads, rivers, and the slope obtained as the digital elevation model (DEM). The topographical data were obtained for the whole of West Africa and the UER processed as a subset of Ghana. Road data were obtained to inform travel routes and, in view of this, types of roads were appended as attribute data to the spatial data to inform the potential speed users are likely to experience on each road type which may as well greatly inform travel distance and time. The DEM for the West Africa region was obtained to inform the slope of every location in the UER. This was necessary because identifying valleys and hills was key to determining which areas serve as barriers and are inaccessible to users. The DEM dataset was obtained from Adu Manu Kwame (AMK) Consultancy and compared with the data obtained from the University of Ghana Remote Sensing and Geographic Information Systems Laboratory for accuracy. We then reclassified the slope data into highlands (more than 200 m high) and flatlands (between 119 and 200 m high) as informed by the DEM data which showed that the highest point in the UER was about 470 m, and 119 m was the lowest point. To realise the objectives of this study, the geo-location data of all the 100 PHC clinics and health facilities such as public hospitals or clinics offering blood type testing services as well as private medical laboratories or hospital/clinics being utilised as referral points for blood type testing services were obtained from the Regional Health Directorate of the Ghana Health Service, UER. We also obtained the spatial data of all the 131 towns in the region using the global positioning system. To accurately map all the latitude and longitude to their relative location on the earth, the World Geodetic System 30 North coordinate system was applied to the entire dataset because Ghana falls within this zone. As a key aspect of this study, the model for determining the travel time was carefully developed taking into consideration all the datasets. Using the PyScripter integrated development environment and relying on the Python capabilities of ArcGIS 5.0, a model was developed to calculate the travel time and for data transformation. The cost distance model which calculates the shortest time to a source based on a cost dataset was used to determine distance. A motorised tricycle was identified as the commonly used public transport for travel within UER, hence, travel time was estimated via road and via paths and tracks using an assumed motorised tricycle speed of 20 km per hour. We recalibrated travel time per pixel (10 m × 10 m grid) for both roads and paths to enable estimation of travel time from PHC clinics where blood type testing services are not available to the nearest hospital, clinic, or medical laboratory for all the districts in the region. Although travel time can be influenced by many factors, we chose to estimate travel time via roads and paths because they are the commonly used routes in the region. Likewise, the motorised transport system was chosen because we found it was the most used public mode of transport for journeys within the region. The model and procedure used to approximate the travel time for this current study have been published in our previous studies focusing on geographical access to tuberculosis diagnostic services and comprehensive ANC POC diagnostic services [12,13]. Supplementary file 1 presents a detailed description of the procedure. We employed the geospatial analysis proximity tool (Buffer) to identify towns and PHC clinics without a blood group and rhesus type testing service located within a 10 km radius, and those located beyond 10 km to a health facility providing the service. We additionally estimated the proximity of all the 131 towns to the nearest health facility providing blood group and rhesus type testing service. Evidence shows that access to healthcare elsewhere more than 10 km away is associated with higher risks of poor health outcomes [22]. ‘Buffer’ in geographic information systems (GIS) refers to a boundary defined by specific units that surround a source or feature. For the purposes of this study, the point buffer was employed because the origin feature in this study which is health facilities providing blood type testing service is a point feature (vector dataset). To assess the accuracy, we created a set of random points from the ground truth data and compared that to the classified data in a confusion matrix using three geoprocessing tools: create accuracy assessment points, update accuracy assessment points, and compute confusion matrix. We utilised the spatial autocorrelation tool in ArcMap 10.5 to determine the spatial distribution of the health facilities providing blood type testing services in the region. Spatial autocorrelation mirrors the first law of geography which states that “everything is related to everything else, but near things are more related than distant things” [23]. In spatial autocorrelation, the null hypothesis of the Moran’s Index statistic states that the feature being measured is distributed randomly, however, when the p-value obtained from running the spatial analyst tool proves to be statistically significant, then the null hypothesis can be rejected [24]. Guided by this, we considered a positive correlation to mean similar values clustered together while negative correlation is representative of different values clustered in a location, and zero means no correlation. This study was approved by the Navrongo Health Research Centre Institutional Review Board/Ghana Health Service (approval number: NHRCIRB291) on 8th January 2018 and the University of KwaZulu-Natal Biomedical Research Ethics Committee (approval number: BE565/17) on 12th February 2018

• kuupiel, desmond
• adu, kwame m.
• bawontuo, vitalis
• adogboba, duncan a.
• mashamba-thompson, tivani p.