Articles

Access to emergency hospital care provided by the public

sector in sub-Saharan Africa in 2015: a geocoded inventory
and spatial analysis
Paul O Ouma, Joseph Maina, Pamela N Thuranira, Peter M Macharia, Victor A Alegana, Mike English, Emelda A Okiro, Robert W Snow

Summary
Background Timely access to emergency care can substantially reduce mortality. International benchmarks for access Lancet Glob Health 2018;
to emergency hospital care have been established to guide ambitions for universal health care by 2030. However, no 6: e342–50

Pan-African database of where hospitals are located exists; therefore, we aimed to complete a geocoded inventory of Published Online
January 24, 2018
hospital services in Africa in relation to how populations might access these services in 2015, with focus on women of
http://dx.doi.org/10.1016/
child bearing age. S2214-109X(17)30488-6
See Comment page e240
Methods We assembled a geocoded inventory of public hospitals across 48 countries and islands of sub-Saharan Africa, Population Health Unit
including Zanzibar, using data from various sources. We only included public hospitals with emergency services that (P O Ouma MSc, J Maina BSc,
were managed by governments at national or local levels and faith-based or non-governmental organisations. For P N Thuranira BSc,
hospital listings without geographical coordinates, we geocoded each facility using Microsoft Encarta (version 2009), P M Macharia MSc,
E A Okiro PhD,
Google Earth (version 7.3), Geonames, Fallingrain, OpenStreetMap, and other national digital gazetteers. We obtained Prof R W Snow FMedSci) and
estimates for total population and women of child bearing age (15–49 years) at a 1 km² spatial resolution from the Health Services Unit
WorldPop database for 2015. Additionally, we assembled road network data from Google Map Maker Project and (Prof M English FMedSci), Kenya
OpenStreetMap using ArcMap (version 10.5). We then combined the road network and the population locations to Medical Research Institute
(KEMRI)-Wellcome Trust
form a travel impedance surface. Subsequently, we formulated a cost distance algorithm based on the location of Research Programme, Nairobi,
public hospitals and the travel impedance surface in AccessMod (version 5) to compute the proportion of populations Kenya; Department of
living within a combined walking and motorised travel time of 2 h to emergency hospital services. Geography and Environment,
University of Southampton,
Southampton, UK
Findings We consulted 100 databases from 48 sub-Saharan countries and islands, including Zanzibar, and identified (V A Alegana PhD); and Centre
4908 public hospitals. 2701 hospitals had either full or partial information about their geographical coordinates. We for Tropical Medicine and
estimated that 287 282 013 (29·0%) people and 64 495 526 (28·2%) women of child bearing age are located more than Global Health, Nuffield
Department of Clinical
2-h travel time from the nearest hospital. Marked differences were observed within and between countries, ranging
Medicine, University of Oxford,
from less than 25% of the population within 2-h travel time of a public hospital in South Sudan to more than 90% in Oxford, UK (Prof M English,
Nigeria, Kenya, Cape Verde, Swaziland, South Africa, Burundi, Comoros, São Tomé and Príncipe, and Zanzibar. Only Prof R W Snow)
16 countries reached the international benchmark of more than 80% of their populations living within a 2-h travel Correspondence to:
time of the nearest hospital. Mr Paul O Ouma, Population
Health Unit, Kenya Medical
Research Institute
Interpretation Physical access to emergency hospital care provided by the public sector in Africa remains poor and (KEMRI)-Wellcome Trust
varies substantially within and between countries. Innovative targeting of emergency care services is necessary to Research Programme,
reduce these inequities. This study provides the first spatial census of public hospital services in Africa. PO BOX 43640–00100, Nairobi,
Kenya
[email protected]
Funding Wellcome Trust and the UK Department for International Development.

Copyright © The Author(s). Published by Elsevier Ltd. This is an Open Access article under the CC BY 4.0 license.

Introduction among low-income and middle-income countries.4,5

In the past 50 years, health system investments in Without emergency care, many conditions would result in
sub-Saharan Africa have focused on primary care and high mortality rates—notably, haemorrhage and hyper­
have substantially increased access to basic curative tensive disorders associated with maternal death.6,7
and preventive services for communicable diseases.1 Improvement of quality, access, efficiency, and admini­
Attention is now increasing for emergency care in Africa, stration of timely emergency services has been suggested
which continues to have high mortality rates from acute to lead to a 45% reduction in mortality and a 36% reduction
maternal and non-communicable diseases.2 Emergency in disability in low-income and middle-income countries.8
care refers to the health system capacity required to ensure The African Federation for Emergency Medicine,
effective provision of curative services for life-threatening through a series of consensus conferences, highlighted
events.3 These events include a diverse set of conditions the need to increase access to emergency care via
spanning injuries, obstetrics, and surgical interventions, identification of gaps in service provision for targeted
with the needs of patients being disproportionately highest interventions.9,10 A maximum of 2-h travel times are

www.thelancet.com/lancetgh Vol 6 March 2018 e342

 Articles

Research in context
Evidence before this study this list to determine geographical access to hospital care in
Emergency or acute conditions are becoming major 48 sub-Saharan African countries and islands, including
contributors of mortality, contributing to 45% of mortality and Zanzibar. Given recommendations for countries to have at
35% of disability in low-income and middle-income countries. least 80% of their population to be living within 2-h travel time
They include trauma, surgical, and obstetric conditions, and of a hospital, this threshold was used to determine populations
should normally be handled at higher level facilities called likely to be geographically marginalised. Our analysis of
hospitals. Defining access to hospital is therefore crucial for geographical access revealed substantial gaps in both the total
driving the emergency care agenda, particularly with the population and women of child bearing age being
expected epidemiological shift, where non communicable marginalised from access to public hospital care.
diseases are expected to be major contributors of mortality in
Implications of all the available evidence
sub-Saharan Africa. However, many countries of sub-Saharan
Physical access to public hospital care is poor in Africa.
Africa have a paucity of information about the coverage of
Sub-Saharan African countries need to develop strategies for
hospitals, which makes it difficult to measure access gaps.
bridging these gaps. We show that at least 30 countries need
Such a resource can be used in combination with population
accelerated investments in hospital care to have more than
distribution in a spatial domain to identify inequities in access
80% of their respective populations within 2-h travel time of a
to care. We searched PubMed and Google Scholar for available
hospital. Our assembled database provides an important
literature associated with measuring of geographical access to
resource that can be easily updated and is made available with
hospital care in sub-Saharan Africa using the search terms
this publication. Additionally, the database provides a baseline
“geographic access”, “access”, “spatial access” and “hospitals”.
for future service availability assessments and should be used to
Our literature searches revealed 30 studies that have defined
undertake more detailed censuses of services provided across
spatial access to emergency services, such as trauma and
hospitals, assess how access might affect trauma, surgical, and
surgery in individual countries, but only eight studies in
obstetric mortality outcomes on the continent, and prioritise
seven countries where analysis was undertaken at national
investments in emergency care service provision to achieve
levels were identified, highlighting a paucity of information
universal health access by 2030.
about geographical access to hospital care for most countries.
Added value of this study
We have undertaken, to the best of our knowledge, the first
geolocated Pan-African database of public hospitals. We used

proposed for emergency hospital care for obstetrics11 specifically for women of child bearing age who would
and emergency surgical interventions.12 A 2015 be at the highest risk of maternal mortality if located far
proposal12 suggests that at least 80% of any country’s from obstetric emergency care.
population should have access to emergency surgical
and anaesthesia services by 2030 to reach international Methods
targets of universal health care access to essential Assembling of a geocoded inventory
medical services. We assembled a geocoded inventory of public hospitals
To understand within and between country variability with emergency services in sub-Saharan Africa. In this
in the provision of emergency hospital care, an audit, we focused on the inclusion of public hospitals that
inventory of hospital service providers and a detailed are managed by governments at national levels or locally
knowledge of a population’s access to these services are at municipality (eg, in Zimbabwe and South Africa) or
required. Hospitals are health facilities expected to local authority (eg, in Kenya and Tanzania), by faith-based
provide care for a range of emergency conditions.13 organisations, and by non-governmental organisations.
Currently, there is no single and spatially defined health In sub-Saharan Africa, these public sector hospitals are
facility list for Africa. Although countries have been often the main service providers of emergency care,
encouraged by WHO to develop master health facility especially for rural populations, and are governed by
lists,14 most facilities have not been defined with national health policy guidelines and regulations. We
longitudes and latitudes and there is no single Pan- excluded private hospitals, which were difficult to audit,
African database of where hospitals are located. although we accept the fact that these hospitals do provide
Therefore, for the first time in Africa, we aimed to important critical care services for those who can afford
present a composite geocoded assembly of public sector them—notably, in urban areas. Additionally, we excluded
hospital services across the continent south of the public hospitals that offer only specialised services (eg,
Sahara. We use this spatial platform to assess general specific psychiatric, leprosy, ophthalmic, spinal, re­
population access to hospital care metrics in 2015, and habilitative, or tuberculosis facilities). Finally, we excluded

e343 www.thelancet.com/lancetgh Vol 6 March 2018

 Articles

services provided to special population groups—notably, rechecked all coordinates using Google Earth to ensure
military and police service hospitals—and institutional that the facilities were within the respective country and
hospitals. Our focus, there­ fore, is hospital services administrative bound­aries of their original lists and were
targeted at a broad range of emergency or referral care to located on major settlements and not on water or offshore.
the general population. To cross reference the completeness of our national
We used data from different sources to construct inventories of public hospitals, we compared our results
country-specific public hospital lists, including those from to the numbers reported in health sector strategic plans
ministries of health, health management infor­ mation (HSSPs), service availability and readiness assessment
systems, and government statistical agencies. We also reports or service provision assessment reports, and the
included data sources from the Humanitarian Data WHO global audit of medical services undertaken in
Exchange portal of the UN’s Office for the Coordination of 2014 (appendix pp 2, 3).15
Humanitarian Affairs (OCHA) and international organi­
sations such as UNICEF and WHO that assemble facility Population data
lists for various purposes. Additionally, we contacted We obtained estimates of total population and women of
individuals working in various departments of the health child bearing age (15–49 years) at 1 km² spatial resolutions
ministries for additional sources of master health facility from the WorldPop database for 2015. This population data For the WorldPop database see
lists that are used for health commodity planning and surface is described elsewhere.16 In brief, the most recent http://www.worldpop.org.uk/
resource allocation. To ensure there were no duplications census data at the highest administrative unit resolutions
of hospitals from the data sources, because countries used by the national census were disaggregated to land use
could list more than one single facility as available on land cover grids. The different land use land cover classes
various databases, we cross-checked and reconciled the were assigned weights on the basis of probability of being
assembled data (appendix pp 2, 3). We selected hospitals populated and a random forest technique was used in the See Online for appendix
from our data sources on the basis of either the use of the disaggregation while adjusting for rural–urban differences
word hospital in the facility name or a level of service and UN urban–rural population growth estimates within
provision indicated in the originator lists. each country.16 The 1 km² population grids were further
In some databases, the hospital listings were accom­ disaggregated using additional age and sex sample census
panied by longitude and latitude for each facility. However, and household demographic data to provide estimates of
for those without geographical coordinates, geocoding was women of child bearing age.17
required with use of Microsoft Encarta (version 2009),
Google Earth (version 7.3), Geonames, Fallingrain, Road network data For Geonames see http://www.
OpenStreetMap, and other national digital gazetteers from Two sources of road network data for Africa are geonames.org/

national education, census, or statistics departments. We publicly available: Google Map Maker Project and For Fallingrain see http://www.
fallingrain.com/world/index.
html
Country Number of Total population Percentage of population Total number of women Percentage of women of
For the OpenStreetMap see
code public in 2015 outside 2-h travel time (UI) of child bearing age child bearing age outside
https://www.openstreetmap.org/
(ISO 3) hospitals in 2015 2-h travel time (UI)
For the Google Map Maker
Angola AGO 150 21 811 453 36·9% (35·5–39·0) 5 489 490 37·3% (35·8–40·0) Project see https://services.
Benin BEN 48 10 880 616 23·3% (20·9–27·2) 2 136 501 23·7% (21·3–27·6) google.com/fb/forms/
Botswana BWA 29 2 258 625 23·3% (20·9–27·5) 622 953 23·3% (20·9–27·5) mapmakerdatadownload/

Burkina Faso BFA 62 18 069 713 46·9% (44·8–50·0) 4 091 918 45·2% (43·1–48·2)

Burundi BDI 49 11 162 902 4·3% (3·9–5·0) 2 515 577 4·2% (3·8–4·9)
Cameroon CMR 184 23 342 359 17·4% (16·4–18·8) 5 230 085 17·1% (16·3–18·7)
Cape Verde CPV 9 488 986 6·6% (6·6–6·6) 132 377 7·0% (6·9–7·0)
Central African Republic CAF 20 4 898 576 51·5% (47·3–55·6) 1 192 470 51·6% (47·4–55·7)
Chad TCD 78 14 022 236 53·1% (51·4–55·7) 2 973 342 53·2% (51·4–55·8)
Comoros COM 3 697 609 3·4% (2·4–5·8) 173 474 3·5% (2·5–6·0)
Congo (Brazzaville) COG 25 4 584 395 27·7% (26·6–29·1) 1 186 661 27·8% (26·7–29·1)
Côte d’Ivoire CIV 100 22 699 552 34·4% (32·9–36·8) 5 659 963 34·4% (32·9–36·8)
Democratic Republic of COD 435 72 001 218 46·3% (44·0–49·2) 14 330 432 46·2% (43·9–49·2)
the Congo
Djibouti DJI 13 870 713 16·7% (16·1–17·6) 211 489 17·0% (16·4–17·9)
Equatorial Guinea GNQ 18 824 276 24·2% (22·2–27·4) 188 338 24·2% (22·3–27·3)
Eritrea ERI 22 5 210 651 57·4% (55·5–59·8) 1 345 007 55·7% (53·9–58·0)
Ethiopia ETH 161 99 337 653 49·3% (45·5–54·3) 22 721 668 49·3% (45·4–54·2)
Gabon GAB 59 1 628 849 16·4% (15·9–17·0) 397 063 16·4% (16·0–17·1)
(Table continues on next page)

www.thelancet.com/lancetgh Vol 6 March 2018 e344

 Articles

Country Number of Total population Percentage of population Total number of women Percentage of women of
code public in 2015 outside 2-h travel time (UI) of child bearing age child bearing age outside
(ISO 3) hospitals in 2015 2-h travel time (UI)
(Continued from previous page)
The Gambia GMB 6 1 950 904 28·5% (28·1–29·3) 411 783 29·3% (28·9–30·0)
Ghana GHA 178 27 098 194 13·8% (12·8–15·5) 7 126 334 13·9% (12·8–15·6)
Guinea GIN 35 12 546 646 37·3% (35·0–40·3) 2 520 496 37·2% (34·9–40·2)
Guinea-Bissau GNB 8 1 745 803 38·5% (34·6–44·4) 429 740 38·4% (34·5–44·3)
Kenya KEN 399 45 737 778 7·1% (6·5–8·1) 11 243 809 7·1% (6·4–8·0)
Lesotho LSO 20 2 136 640 43·3% (40·8–46·4) 570 583 43·3% (40·9–46·5)
Liberia LBR 38 4 451 499 38·5% (36·9–40·7) 1 090 502 38·5% (36·9–40·6)
Madagascar MDG 125 24 120 532 53·4% (52·0–55·8) 5 262 812 53·3% (51·9–55·7)
Malawi MWI 56 17 207 197 7·2% (5·5–10·2) 3 938 576 7·2% (5·5–10·2)
Mali MLI 76 17 619 152 36·2% (33·3–40·6) 3 258 813 35·3% (32·4–39·7)
Mauritania MRT 18 4 026 075 61·4% (59·1–64·1) 880 617 61·4% (59·1–64·2)
Mozambique MOZ 61 27 673 736 49·9% (44·8–52·7) 6 431 717 49·8% (44·6–52·5)
Namibia NAM 35 2 461 440 23·2% (20·2–27·4) 644 284 23·2% (20·3–27·5)
Niger NER 41 19 805 985 57·2% (53·6–61·7) 3 376 284 57·1% (53·5–61·6)
Nigeria NGA 879 182 178 061 7·7% (6·9–9·0) 43 659 033 8·5% (7·8–10·1)
Rwanda RWA 47 11 585 862 11·2% (10·6–12·1) 2 793 168 11·3% (10·7–12·2)
São Tomé and Príncipe STP 2 186 623 2·7% (2·4–3·1) 43 894 2·7% (2·5–3·2)
Senegal SEN 29 14 967 332 39·7% (36·2–41·2) 3 583 623 39·6% (36·1–41·1)
Sierra Leone SLE 32 6 418 015 39·6% (36·9–39·1) 1 537 021 39·6% (37·0–39·2)
Somalia SOM 79 10 688 048 44·0% (42·5–47·4) 2 179 717 42·3% (40·9–45·6)
South Africa ZAF 327 54 345 833 5·2% (4·4–6·3) 15 021 723 5·2% (4·5–6·4)
South Sudan SDS 40 12 347 507 77·2% (76·5–79·0) 2 786 192 77·3% (76·6–79·0)
Sudan SDN 272 40 249 394 53·8% (53·2–56·5) 9 346 646 53·7% (53·1–56·3)
Swaziland SWZ 7 1 285 392 6·1% (3·7–8·9) 289 274 6·1% (3·8–9·0)
Tanzania (mainland) TZA 210 53 265 074 24·9% (22·1–29·0) 12 585 409 24·8% (22·1–29·0)
Togo TGO 38 7 304 010 14·7% (12·9–17·4) 1 588 063 14·8% (13·3–17·7)
Uganda UGA 121 39 032 494 17·5% (15·5–20·7) 7 979 425 17·5% (15·5–20·8)
Zambia ZMB 91 16 218 094 40·1% (37·1–43·7) 3 699 046 40·0% (37·0–43·7)
Zanzibar ·· 4 1 579 927 2·7% (2·0–5·6) 376 652 2·6% (1·9–5·5)
Zimbabwe ZWE 169 15 604 001 20·7% (18·6–23·9) 3 453 496 21·5% (19·4–24·6)
Total ·· 4908 990 627 630 29·0% (27·1–31·5) 228 707 540 28·2% (26·4–30·8)

UI=uncertainty intervals.

Table: Number of public hospitals and access quotients with UI across 48 sub-Saharan African countries and islands, including Zanzibar for 2015

OpenStreetMap. We combined these data using travel speeds and this information was stored as
ArcMap (version 10.5). We included information about attributes.
major roads that allowed year-round motorised travel,
including trunk or highway, primary, secondary, and Geographical access
tertiary or arterial roads. We reclassified these major To define geographical access to public hospitals, we
roads as primary, secondary, and tertiary roads. Those used a cost distance algorithm that modelled a
classified as primary were high volume roads that composite of walking and motorised travel time to the
mainly connect international borders. Secondary roads nearest public hospital. We generated a travel impedance
were those that fed into primary roads or connected surface by assigning travel speeds of 100 km/h to
major towns in different regions of a country. Tertiary primary roads, 50 km/h to secondary roads, and
roads were those that connected secondary roads while 30 km/h to tertiary roads on the basis of previous
connecting smaller towns or market centres. We studies.18,19 The model does not give any measures of
corrected the data to connect segments of roads uncertainty; and for sensitivity analysis, we varied the
omitted through digitisation and deleted those that motorised speeds by more or less than 20% of the
extended into water bodies. Each road segment was original speeds19,20 to define an upper and lower bound
accompanied by information about average motorised of uncertainty around travel speeds. We assigned the

e345 www.thelancet.com/lancetgh Vol 6 March 2018

 Articles

A B
Primary roads
Secondary roads
Tertiary roads

Santo Antão
Cape Verde
Santo Antão Boa Vista
Saõ Nicolau
Pemba
Boa Vista Santiago Pemba
Saõ Nicolau Maio Bioko
Zanzibar Fogo Zanzibar
Maio
Cape Verde Príncipe
Fogo Santiago Bioko

Príncipe São Tomé

Annobón
São Tomé
Annobón Comoros
Comoros Grande Comore
Population per 1 km2 grid Grande Comore Anjouan
Mohéli
Anjouan
0 134 583 Mohéli

C
Public hospital*

Santo Antão Cape Verde

Pemba
Saõ Nicolau Boa Vista
Bioko
Zanzibar
Maio Príncipe
Fogo Santiago

São Tomé
Annobón

Comoros
Grande Comore
Anjouan
0 500 1000 2000 km
Mohéli

Figure 1: Population density, road network coverage, and locations of public hospitals in sub-Saharan Africa in 2015
Regions shaded in grey were not included, as they are not part of sub-Saharan Africa. (A) Population density per 1 km². (B) Coverage of road network where motorised travel to hospitals is possible.
(C) Distribution of 4893 public hospitals. *15 hospitals could not be geocoded.

other non-road raster cells with speeds of 5 km/h, We used the impedance surface and location of public
assuming patients could walk, were carried, or were hospitals to estimate time in hours needed to travel to the
transported with use of other means to the nearest road nearest public hospital using AccessMod (version 5). The For AccessMod 5 see
before obtaining motorised travel. algorithm we used for estimating travel times in https://www.accessmod.org/

www.thelancet.com/lancetgh Vol 6 March 2018 e346

 Articles

<1 person per 1 km2 who would take >2 h to travel to the nearest hospital DVN/JTL9VY (Ouma P, Okiro EA, Snow RW, 2017), under
≥1 person per 1 km2 who would take >2 h to travel to the nearest hospital a CC BY 4.0 license.
Within 2 h travel time of the nearest hospital

Role of the funding source

The funder of the study had no role in study design, data
collection, data analysis, data interpretation, or writing of
the report. The corresponding author had full access to
all the data in the study and had final responsibility for
the decision to submit for publication.

Results
We consulted 100 databases from 48 sub-Saharan
countries and islands (appendix pp 6–22). Zanzibar, which
has an independent health system and is governed by a
Cape Verde
Santo Antão separate ministry of health to mainland Tanzania, was
Saõ Nicolau
Boa Vista
Pemba analysed separately. International organi­sations such as
Santiago Bioko OCHA, WHO, and UNICEF contributed to 33 databases.
Maio Zanzibar
Fogo
Príncipe 31 databases were sourced from ministries of health, nine
from the national malaria control programmes, nine from
health management information systems, and eight from
São Tomé other ministries and government agencies such as
Annobón statistical offices. Other sources included journal articles
(n=3), web information pages (n=2), unpublished
Comoros individual documents (n=2), databases from Christian
Grande Comore
Anjouan health associations (n=2), and Google Earth (n=1). Of
0 500 1000 2000 km Mohéli the 100 sources of information, 24 were obtained
from personal contacts in ministries of health or non-
Figure 2: Geographical access of the general population to public hospitals governmental organisations.
Regions shaded in grey were not included. 4908 public hospitals were identified across 48 countries
and islands of sub-Saharan Africa, including Zanzibar.
AccessMod uses the Manhattan distance, based on travel The numbers of public hospitals likely to be offering
along road infra­structure, compared with the Euclidean emergency or referral care ranged from two in São Tomé
method, which assumes travel in a straight line. Because and Príncipe to 879 in Nigeria (table). 2701 hospital data
this algorithm relies on converting roads to a raster sources had either full or partial information about the
surface, the accuracy of the model is in part affected by longitude and latitude, the remaining 2207 hospitals
the spatial resolution. Therefore, we used 100 m spatial required geocoding. We were unable to geocode 15 public
grids to capture finer heterogeneity in travel times. Thus, hospitals in Somalia (n=5) and Sudan (n=10). These
the time needed to get to a hospital is determined by hospitals were expected to offer services such as surgi­
cumulatively adding the time needed to cross the pixels cal care, obstetrics, paediatrics, radiological, and lab­
in the so-called least cost path from the location of interest oratory services (appendix pp 4, 5).
to a hospital. Further­more, we confined the analysis of It was possible to compare the summaries of hospital
travel and accessibility to hospitals to within the national services in 30 countries, where these were specified in
borders, assuming populations do not cross borders to both the HSSPs and the 2014 WHO audit.15 Overall, the
use hospitals in neighbouring countries. We defined number of hospitals audited during the present exercise
national means for island groups (Zanzibar, São Tomé (n=3203) and HSSPs (n=3469) were broadly comparable,
and Príncipe, Cape Verde, and Comoros) by averaging the but greater than those specified in the 2014 WHO audit
accessibility quotients from each island’s prediction. We (n=2518). There were substantial differences in the
computed the proportion of the total population and numbers of hospitals reportedly available in Ethiopia
women of child bearing age within 2 h-travel time to a (212 hospitals audited from HSSPs, 187 from WHO, and
hospital in AccessMod and mapped results using ArcGIS 161 from present audit), Sudan (428 from HSSP,
(version 10.5). 255 from WHO, and 262 from present audit), and
Uganda (160 from HSSP, 64 from WHO, and 121 from
For the full database that Data sharing present audit).
supports the findings of this The full database that supports the findings of this study As expected, where population density was high,
study see https://dataverse.
are available from the Kenya Medical Research Institute hospital density was also high (figures 1A, 1C).
harvard.edu/dataset.
xhtml?persistentId=doi:10.7910/ Wellcome Trust Research Programme’s Population 703 345 617 (71%) people and 164 212 014 (71·8%) women
DVN/JTL9VY Health page in the Harvard Dataverse, http://DOI:10.7910/ of child bearing age were living within 2-h travel time of

e347 www.thelancet.com/lancetgh Vol 6 March 2018

 Articles

the nearest public hospital across the 48 sub-Saharan

South Sudan
African countries or offshore islands (table). Conversely, Mauritania
the most geographically marginalised in 2015 comprised Eritrea
Niger
287 282 013 (29%) people and 64 495 526 (28·2%) women Sudan
of child bearing age who had travel times more than 2 h Madagascar
from emergency public hospital care. Populated areas Chad
Central African Republic
with more than one person per km² but outside the Mozambique
2-h motorised travel time to public hospital care are Ethiopia
Burkina Faso
shown in figure 2 to highlight areas in need of greatest Democratic Republic of the Congo
service accessibility. Somalia
Geographical accessibility to emergency hospital care Lesotho
Zambia
varied between countries (table, figure 3), ranging from Senegal
less than 25% of the population within 2-h travel time Sierra Leone
Guinea-Bissau
of a public hospital in South Sudan to more than 90% in Liberia
Nigeria, Kenya, Cape Verde, Swaziland, South Africa, Guinea
Angola
Burundi, Comoros, São Tomé and Príncipe, and Mali
Zanzibar. Seven countries had less than 50% of the Côte d’Ivoire
population within 2-h travel time of a public emergency The Gambia
Congo (Brazzaville)
care hospital: South Sudan, Mauritania, Eritrea, Niger, Tanzania (mainland)
Sudan, Madagascar, and Chad. Several large countries Equatorial Guinea
Benin
such as South Sudan, Mauritania, Democratic Republic Botswana
of the Congo, Mozambique, and Zambia have poor Namibia
access to hospital care compared with smaller countries Zimbabwe
Uganda
or islands such as Cape Verde, Zanzibar, and São Tomé Cameroon
and Príncipe. However, notable exceptions were Djibouti
Gabon
observed with large countries such as Nigeria, Kenya, Togo
and South Africa having more than 90% of their Ghana
respective population living within 2-h travel time of a Rwanda
Nigeria
hospital, whereas smaller countries such as Eritrea and Malawi
Lesotho have poor access. Only 16 countries met the Kenya
Cape Verde
international recommendation12 of more than 80% of Swaziland
the population within 2-h travel time of a hospital South Africa
(figure 3). Burundi
Comoros
São Tomé and Príncipe
Discussion Zanzibar
We have assembled the first Pan-African, geocoded 0 20 30 40 50 60 70 80 90 100
database of public hospitals in 48 countries and islands. Proportion of population living within 2-h travel time of a hospital (%)
We estimated that in Africa, about 29% of the population
Figure 3: Proportion of population living within 2-h travel time of a hospital in 2015 in sub-Saharan Africa
and about 28% of women of child bearing age are
Error bars are uncertainty intervals. The dotted line distinguishes between countries that have 80% of their
geographically marginalised from emergency medical, populations within 2-h travel time of a hospital and those yet to achieve this proportion.
obstetric, and surgical care, and live more than 2-h travel
time to the nearest public hospital; and only 16 of ambulatory services, or the numbers of hospitals in
48 countries have more than 80% of their population specific geographical locations.
living within 2-h travel time of emergency hospital care. Health facility lists in Africa are fragmented, only
We note that smaller countries and islands have 31 (31%) of 100 original sources used in this study were
proportionately better access to hospital services than in from ministries of health, and a diverse list of sources,
larger countries, where more than 40% of the population especially from other governmental and international
live more than 2-h travel time to a public hospital. agencies, were required to provide a more compre­hensive
However, we noted exceptions where large countries understanding of hospital care. The accuracy and
have better access quotients than in smaller countries. completeness of this resource now requires further
One important function of routine hospital services is country and regional level efforts, although our initial
the ability to provide emergency obstetric care, and we database serves as a useful entry point to future hospital
have defined access to hospital care for both women of censuses in Africa. There was no universal definition of
child bearing age and total populations. To reach hospital or emergency care, and definitions provided in
geographically marginalised populations with hospital national health policies varied between countries
care, innovative targeting of emergency care is required, (appendix pp 4, 5). This absence of a definition has been
including improvement of transportation modes, noted previously and demands a more standard

www.thelancet.com/lancetgh Vol 6 March 2018 e348

 Articles

classification of emergency hospital care provision in emergency services for trauma, surgical, and obstetric
Africa.21 We were not able to ascertain what services are patients. Most countries were well below the benchmark
provided at each of the mapped public hospitals. set for 2030, where less than 80% of the population lived
Additionally, the number of hospitals audited during the within 2-h travel time of emergency hospital care. The
present exercise differed to those specified in the HSSPs importance of hospital services goes beyond emergency
and 2014 WHO audit of specific countries, possibly as a care, hospitals additionally provide the core backbone to
result of differences in years for which an HSSP was surveillance of emerging or escalating antimicrobial
published or inclusion or exclusion of private sector resistance, the detection of new pathogen epidemics, and
hospitals where this number was not made clear in the provide the means to define the operational effectiveness
HSSP. A review of 48 HSSPs, the cornerstone policy of the introduction of new vaccines or other community-
documents that describe the national health system, based interventions. Hospitals provide an essential part
highlights the inadequacy of national definitions of of pathogen and intervention effect across Africa, and
hospital services, minimum clinical and laboratory knowing where they are located is essential. Definition
equipment, or staffing needs that should constitute this of the scope, service provision capacities, laboratory
higher level of the health sector (appendix pp 4, 5). The capacities, and optimal catchment populations for
actual level of service provision should in theory be emergency hospital care should be a priority. Ultimately,
available from national surveys of service availability and we provide a resource to begin these urgent censuses of
readiness assessment reports and service provision hospital services in Africa and where applications extend
assessment reports;22 however, these surveys are often beyond just defining physical access.
only sampled facilities and have only been undertaken in Contributors
18 sub-Saharan countries.23 POO did the data checking, data analysis, and writing of the first draft of
The role of the private sector in achieving universal the manuscript. JM, PMM, and PNT provided support to geocoding and
assembly of geographical data. VAA, ME, and EAO provided scientific
health coverage is essential,24 but it remains poorly advice and support throughout the project. RWS conceived the project,
enumerated in sub-Saharan Africa, including our audit, provided overall management, and contributed to second drafts of the
despite extensive searches from multiple sources. Future manuscript. All authors reviewed the final analysis, have full access to
hospital service censuses and audits must include the the data, and approved the final manuscript for publication.
private sector because these services compete with public Declaration of interests
sector services notably in urban areas. We declare no competing interests.
Our accessibility analysis had a number of limitations. Acknowledgments
We were unable to evaluate the efficiency, timeliness, Funding was provided to RWS as part of his Wellcome Trust Principal
Fellowship (number 103602) that in part supported POO, JM, PTN, and
and abilities of referral transport systems in each country PMM; RWS is also grateful to the UK’s Department for International
with adequate precision at a continental scale. We have Development (DFID) for their continued support to a project on
not been able to account for frequency of transport Strengthening the Use of Data for Malaria Decision Making in Africa
services on secondary to main roads connected to (DFID programme code number 203155) that provided support to
POO, JM, PNT, and PMM. ME is supported through a Wellcome Trust
hospital locations, the precise transport speeds, how Senior Fellowship (number 097170) and EAO is supported through a
prostrated emergency care patients are transported from Wellcome Trust Intermediary Fellowship (number 201866). POO and
households to arterial road networks, or the multitude of PMM acknowledge support under the IDeAL Project (number 107769).
All authors acknowledge the support from the Wellcome Trust to the
other physical and financial barriers to referral from
Kenya Major Overseas Programme (number 203077). We thank
home to hospital in each country or subregions. everyone across the African ministries of health and non-governmental
Additionally, we were unable to account for dynamic organisations who provided assistance in locating national health
population changes at very fine spatial resolutions facility lists acknowledged in the appendix (pp 6–22). Additionally, we
acknowledge the ongoing support and assistance provided by Ernest
because these data are unavailable at continental scales.
Dabire and Socé Fall (WHO Africa Regional Office) and Abdisalan Noor
These additional analyses are beyond the scope of (Global Malaria Programme, WHO, Geneva, Switzerland). We also
the present paper, but should be highlighted to build thank Dejan Zurovac for comments on an earlier version of this
knowledge across the extensive domains of hospital manuscript.
service access through increased spatial resolution References
studies25,26 and understanding of referral care for hospital 1 Razzak J, Kellermann A. Emergency medical care in developing
countries: is it worthwhile? Bull World Health Organ 2002;
services in Africa. 80: 900–05.
In summary, consensus on the need to integrate 2 GBD 2015 Healthcare Access and Quality Collaborators. Healthcare
emergency care into health systems is increasing.27,28 Key Access and Quality Index based on mortality from causes amenable
to personal health care in 195 countries and territories, 1990–2015:
towards addressing challenges in emergency care is a novel analysis from the Global Burden of Disease Study 2015.
defining access to hospitals and highlighting populations Lancet 2017; 390: 231–66.
most distal from these services. We have assembled 3 Calvello EJ, Broccoli M, Risko N, et al. Emergency care and health
systems: consensus-based recommendations and future research
the first geocoded database of public hospitals in priorities. Acad Emerg Med 2013; 20: 1278–88.
sub-Saharan Africa, and have used this audit or inventory 4 Norton R, Kobusingye OC. Injuries. N Engl J Med 2013; 18: 1723–30.
to provide a ranking of the worst and best hospital-served 5 Say L, Chou D, Gemmill A, et al. Global causes of maternal death:
countries that in theory should be able to provide vital a WHO systematic analysis. Lancet Glob Health 2014; 2: e323–33.

e349 www.thelancet.com/lancetgh Vol 6 March 2018

 Articles

6 Bhutta ZA, Das JK, Bahl R, et al. Can available interventions end 18 Esquivel MM, Uribe-Leitz T, Makasa E, et al. Mapping disparities in
preventable deaths in mothers, newborn babies, and stillbirths, access to safe, timely, and essential surgical care in Zambia.
and at what cost? Lancet 2014; 384: 347–70. JAMA Surg 2016; 151: 1064–69.
7 Graham W, Woodd S, Byass P, et al. Diversity and divergence: 19 Stewart BT, Tansley G, Gyedu A, et al. Mapping access to essential
the dynamic burden of poor maternal health. Lancet 2016; 388: 2164–75. surgical care in Ghana using the availability of bellwether
8 Kobusingye OC, Hyder AA, Bishai D, Joshipura M, Hicks ER, procedures. JAMA Surg 2016; 151: e161239.
Mock C. Chapter 68: emergency medical services. In: Jamison DT, 20 Tansley G, Schuurman N, Amram O, Yanchar N. Spatial access to
Breman JG, Measham AR, et al, eds. Disease control priorities in emergency services in low- and middle-income countries:
developing countries, 2nd edn. New York: Oxford University Press, a GIS-based analysis. PLoS One 2015; 10: e0141113.
2006: 1261–79. 21 English M, Lanata C, Ngugi I, Smith PC. Chapter 65: the district
9 Calvello EJ, Reynolds T, Hirshon JM, et al. Emergency care in hospital. In: Jamison DT, Breman JG, Measham AR, et al, eds.
sub-Saharan Africa: results of a consensus conference. Disease control priorities in developing countries, 2nd edn.
African J Emerg Med 2013; 3: 42–48. New York: Oxford University Press, 2006: 1211–28.
10 Reynolds TA, Calvello EJB, Broccoli MC, et al. AFEM consensus 22 WHO. Service availability and readiness assessment (SARA).
conference 2013 summary: emergency care in Africa—where are we Geneva: World Health Organization, 2017. http://www.who.int/
now? African J Emerg Med 2014; 4: 158–63. healthinfo/systems/sara_reports/en/index1.html
11 WHO, UNFPA, UNICEF, Averting Maternal Death and Disability. (accessed July 17, 2017).
Monitoring emergency obstetric care: a handbook. Geneva: World 23 ICF Macro. Service provision assessment surveys. Calverton:
Health Organization, 2009. http://apps.who.int/iris/bitstream/10665/ ICF Macro, 2017. https://dhsprogram.com/What-We-Do/Survey-
44121/1/9789241547734_eng.pdf (accessed May 15, 2017). Types/SPA.cfm (accessed Nov 1, 2017).
12 Meara JG, Leather AJ, Hagander L, et al. Global surgery 2030: 24 Morgan R, Ensor T, Waters H. Performance of private sector health
evidence and solutions for achieving health, welfare, and economic care: implications for universal health coverage. Lancet 2016;
development. Lancet 2015; 386: 569–624. 388: 606–12.
13 WHO. Hospitals. Geneva: World Health Organization, 2017. 25 Vanderschuren M, McKune D. Emergency care facility access in
http://www.who.int/hospitals/en/ (accessed Oct 16, 2017). rural areas within the golden hour? Western Cape case study.
14 WHO. Creating a master health facility list. Geneva: World Health Int J Health Geogr 2015; 14: 5.
Organization, 2013. http://www.who.int/healthinfo/systems/ 26 Manongi R, Mtei F, Mtove G, et al. Inpatient child mortality by travel
WHO_CreatingMFL_draft.pdf?ua=1 (accessed May 3, 2017). time to hospital in a rural area of Tanzania. Trop Med Int Heal 2014;
15 WHO. Global atlas of medical devices. Geneva: World Health 19: 555–62.
Organization, 2014. http://www.who.int/medical_devices/ 27 McCord C, Kruk ME, Mock CN, et al. Chapter 12: organization of
publications/global_atlas_meddev2017/en/ (accessed April 21, 2017). essential services and the role of first-level hospitals. In: Debas HT,
16 Stevens FR, Gaughan AE, Linard C, Tatem AJ. Disaggregating Donkor P, Gawande A, et al, eds. Essential surgery: disease control
census data for population mapping using random forests with priorities, 3rd edn. Washington, DC: The International Bank for
remotely-sensed and ancillary data. PLoS One 2015; 10: e0107042. Reconstruction and Development and the World Bank, 2015: 213–30.
17 Tatem AJ, Campbell J, Guerra-Arias M, de Bernis L, Moran A, 28 World Health Assembly. Health systems: emergency-care systems
Matthews Z. Mapping for maternal and newborn health: (WHA resolution 60.22). Geneva: World Health Organization, 2007.
the distributions of women of childbearing age, pregnancies and http://www.who.int/emergencycare/gaci/gaci_tor.pdf?ua=1
births. Int J Health Geogr 2014; 13: 2. (accessed Oct 13, 2017).

www.thelancet.com/lancetgh Vol 6 March 2018 e350