Introduction Access to emergency neonatal health services has not been explored widely in the Ethiopian context. Accessibility to health services is a function of the distribution and location of services, including distance, travel time, cost and convenience. Measuring the physical accessibility of health services contributes to understanding the performance of health systems, thereby enabling evidence-based health planning and policies. The physical accessibility of Ethiopian health services, particularly emergency neonatal care (EmNeC) services, is unknown. Objective To analyse the physical accessibility of EmNeC services at the national and subnational levels in Ethiopia. Methods We analysed the physical accessibility of EmNeC services within 30, 60 and 120 min of travel time in Ethiopia at a national and subnational level. We used the 2016 Ethiopian Emergency Obstetric and Neonatal Care survey in addition to several geospatial data sources. Results We estimated that 21.4%, 35.9% and 46.4% of live births in 2016 were within 30, 60 and 120 min of travel time of fully EmNeC services, but there was considerable variation across regions. Addis Ababa and the Hareri regional state had full access (100% coverage) to EmNeC services within 2 hours travel time, while the Afar (15.3%) and Somali (16.3%) regional states had the lowest access. Conclusions The physical access to EmNeC services in Ethiopia is well below the universal health coverage expectations stated by the United Nations. Increasing the availability of EmNeC to health facilities where routine delivery services currently are taking place would significantly increase physical access. Our results reinforce the need to revise service allocations across administrative regions and consider improving disadvantaged areas in future health service planning.
The Ethiopian health system is based on the primary healthcare approach with three levels of care. The primary level includes primary hospitals, health centres and health posts (the lowest-level facility at a village level). The secondary level includes specialty centres (eg, maternal and child health (MCH) specialty centres), specialty clinics and general hospitals that serve as referral centres for primary hospitals, and the tertiary level includes specialised referral hospitals.26 Most of the EmNeC signal functions are performed in hospitals and MCH specialty centres.18 A specialty centre differs from a specialty clinic as specialty centres have inpatient admissions and offer 24 hours emergency services. A specialty centre differs from a hospital in that they do not offer the full spectrum of specialties required for a general hospital.27 The target population for this analysis were all live births in Ethiopia taken from the 2016 UN estimates of numbers of live births per 1 km grid square.28 The primary outcome variable in this study was accessibility to fully functioning EmNeC within 30 min, 1 hour and 2 hours travel time at a national and regional level. Secondary outcomes include accessibility of fully functioning health services through walking travel and access to health facilities with partial EmNeC signal functions. Accessibility was defined as the access from a residence to a health facility within 2 hours of travel time based on the WHO optimal access definition.29 The facilities were considered to be fully functioning if all the seven signal functions for EmNeC were available in the past 3 months before the survey. No patient involved. We used three data groups for this analysis: statistical, geospatial and national norms data. The statistical data included national and regional population sizes and the number of functional EmNeC health facilities. The geospatial data were regional administrative boundary data, geographical location of all health facilities providing delivery services, road network, hydrographical network, land cover, digital elevation model (DEM) data and spatial distribution of live births in Ethiopia in 2015. The national norms data include the maximum travel speed expected for a motor vehicle on the different road types and the average capacity of health facilities that they could serve. We used the Ethiopian 2016 EmONC survey, which is a national census of health facilities providing maternal and neonatal services.18 A total of 3804 health facilities providing delivery services were included in the EmONC survey. Ethiopia’s 2016 produced boundary shapefile, a geospatial vector data format matching the level of disaggregation of the subnational statistical data, was accessed through OpenAFRICA30; land cover and DEM data of 2015, a representation of the bare ground topographical surface of the Earth were accessed from the DIVA-GIS webpage.31 The 2015 raster data for live births per 1 km grid square was accessed via the WorldPop webpage.28 We used OpenStreetMaps via the World Food Programme data repository of 201732 to estimate travel speeds for motorised vehicles based on the primary, secondary, and tertiary and unclassified road surfaces. We used DEM data to estimate the effects of slopes on travel time.33 The barrier to travel (hydrographical data) was obtained from RCMRD GeoPortal34 and DIVA-GIS31 web pages produced in 2015. We created a travel scenario for walking and motorised transportation based on the land cover structure and road types. We assumed that walking speed ranged from zero km/hour for water bodies to 2.5 km/hour for established residential areas, assuming a pregnant woman in her last month of pregnancy would be able to walk at half the average walking speed as used elsewhere.35 We assumed driving speeds of 100, 70, 50 and 30 km per hour for primary, secondary, tertiary and unclassified road types based on the country’s speed limit.36 We used these speeds as ambulances would be able to travel at the maximum speed in emergency situations. We also conducted a sensitivity analysis for fully EmNeC services coverage, lowering the maximum vehicle speed by 25%. The road network data were classified based on Ethiopian speed limit norms. The main road classifications in Ethiopia are primary, secondary, tertiary and unclassified.32 However, the road network data available online includes several classifications, including the linking roads between primary roads-so we need to reclassify into the above four main classifications. Primary, primary link, motorway and trunk roads were classified as primary, secondary and secondary links as secondary, tertiary and tertiary links as tertiary and track and unclassified were merged as unclassified road classes. We assigned the maximum capacity for each health facility category based on the standard WHO assumption of births per skilled birth attendant per year.37 We assigned 175, 100, 75 and 50 births per skilled birth attendant per year for hospitals, MCH specialty centres, health centres and clinics. We then multiplied the number of births per year by the number of skilled birth attendants at each facility to estimate the total number of possible births per year at each health facility. We integrated multiple geospatial datasets in AccessMod, which is a free, open-source software package developed by the WHO. AccessMod uses geographical information systems (GIS),38 which are computer-based systems to gather, store, retrieve, analyse and display spatial data, to assess health facilities physical accessibility and geographical coverage.39 AccessMod models the coverage of catchment areas linked to an existing health facility network integrating population distribution, travel time and the population coverage capacity specific to each health facility in the network.40 AccessMod computes catchment areas using the least-cost path algorithm.41 The least-cost path approach calculates the distance between a focal location and all cells in the surroundings, dividing surface areas into grid cells. It identifies the best path from one point to another over a cost surface, identifying the cost of travelling through each grid cell, which has been given to cost how expensive, it is to pass through that cell.42 The cost given to each cell is the travelling time to cross the grid cell, which is determined through the travelling speed attributed to the elevation and land cover of the cell. Finally, it produces a point estimate of cumulative access coverage of health services to catchment areas population. The vector and raster geospatial data files were projected based on Ethiopia’s geographical coordinate system at Adindan UTM zone 37N43 to make it suitable for analysis. A projection is the means by which we display the coordinate system and data on a flat surface, such as a piece of paper or a digital screen. A projected coordinate system is a two-dimensional flat surface, and locations, in this case, are identified by x, y coordinates.44