Putting health facilities on the map: a renewed call to create geolocated, comprehensive, updated, openly licensed dataset of health facilities in sub-Saharan African countries

There is a set of minimum essential attributes for HFDBs which should be prioritized to maximize their usefulness on local, national, and regional levels. These attributes include facility name, unique ID, type, ownership, operational status, and, importantly, geographical coordinates with the relevant subnational unit (Table 1). Furthermore, countries developing their HFDBs should consider additional attributes through technical consensus with stakeholders to meet their specific needs. Several resources can be utilized, such as the GHFD initiative, which includes four unique identifiers, allowing countries the flexibility to add other data elements (5). The Health Facility Assessment Technical Working Group led by United States Agency for International Development (USAID) proposed eight domains for core indicators to identify a health facility (28) which was adopted by Haitian HFDB (7). Further, the International Hospital Federation, in collaboration with healthsites.io, defined 15 core attributes/identifiers (30). With input from Médecins Sans Frontières (MSF) International, CartONG, and the Humanitarian OpenStreetMap Team (HOT), this foundational dataset was then mapped to OpenStreetMap(31), establishing healthsites.io as a Digital Public Good (32). For a country without an official HFDB, initial efforts should be made to compile an openly licensed HFDB with unique IDs used across all in-country programs. This can be achieved by triangulating all existing independent lists within the country to derive a single list, as was recently done by a group of researchers and staff from different governmental agencies in Senegal (27). Efforts such as the healthsites.io open campaign provide complementary approaches through on-ground validation exercises (30, 33). The minimum essential attributes should be updated frequently and expanded to include private facilities. In later sections, we list possible sources to inform the initial effort of compiling a HFDB.

Establishing a geocoded HFDB is merely an initial step. Equally important, a robust health system should be able to build on these minimum essential variables and include attributes of service availability and capacity linked with unique facility IDs. Therefore, all other program lists held by various in country programs, such as the immunization program, should reference the HFDB for essential attributes using the unique ID while at the same time feeding back to HFDB for updates of service and capacity. The ideal approach to compile a HFDB is a comprehensive census that records all essential attributes, services, and capacities of existing health facilities. For instance, in 2023, Kenya conducted a census incorporating both public and private facilities, collecting geographic coordinates, and adding data on health services and capacities while making the data open accessible (34). In the next section, we outline the historical developments in forming geocoded facility lists and further explore the proposed minimum attributes.

Geographical coordinates of facilities

Importance of geolocation in facility list

John Snow’s map of cholera cases and water sources showing that the distribution of the disease had a geospatial pattern related to drinking water sources in 1854 is a classic example of the utility of applying spatial dimensions to health data (35). The value of geocoded facility lists to health service provision is vast. They facilitate the linkage between the geographical location of health facilities with disease events within the catchment populations they serve (36), across the urbanicity continuum (core urban areas to rural areas) (37,38,39), for example, computing malaria incidence using routine data (40) or estimating the number of children likely to use a health facility after an episode of fever or different health outcomes for programmatic planning (41). This leads to a granular understanding of gaps in geographical accessibility to healthcare and deriving spatial patterns that can feed into national policymaking in the era of UHC (38). Consequently, the disease burden, social dynamics, and environmental factors that influence the health needs of particular catchment populations can be integrated to better adapt health services to meet the specific needs of those populations. This could facilitate more effective, equitable, and disease-specific healthcare delivery.

There are multiple examples of how geographical coordinates in a HFDB have been used in health programs. They have informed micro plan maps and facility catchments for improved targeting of populations needing health interventions, such as polio vaccination campaigns, reaching zero dose children, and monitoring the efficacy of COVID-19 routine vaccinations (36, 42). This might include using locations of facilities as reference for coordination, orientation, and knowing where the vaccines are stored. Other examples include determining subsets of the population that are geographically marginalized to different types of care, improving outreach activities by determining the optimal areas for CHWs linked to each facility. Further, databases of facilities have been used as gazetteers when geocoding villages and settlements and finally improving health outcomes as the location of facilities facilitates a well-coordinated and sustainable delivery of interventions in the areas that they are needed most (36).

Likewise, there are many examples of how geographical coordinates in a HFDB have been used in research. A geocoded database of health facilities in 50 countries in SSA (25) was used to estimate travel time to facilities within areas at risk of viral hemorrhagic fevers (43). Vulnerable populations that would benefit from new health facilities and reduced travel time were identified. Facilities in the vicinity of at-risk populations were also recognized for prioritization in their readiness capacity to detect, treat, and respond to emerging pathogens (43). During COVID-19, the same SSA facility list (25) was used to map geographical access to health facilities to inform where additional resources such as makeshift hospitals or transport programs might be needed for adults aged ≥ 60 years (15). A recent body of work targeting the optimization of geographical accessibility to emergency obstetric and newborn care (EmONC) facilities (44, 45) has led to the adoption of the first maternal health indicator using travel time to the nearest EmONC facility within the framework of the Ending Preventable Maternal Mortality (EPMM) initiative (i.e., EPMM target 4 indicator) (46). Perhaps, the most urgent and timely use will be the spatial precision needed for planning and response to Mpox (47, 48) in the same manner as COVID-19, Ebola (16), and cholera (7).

Considerations for accurately geolocating health facilities

It is vital to assign precise and accurate location attributes (geographical coordinates) in defining the location of a health facility (geolocation or geocoding). The gold standard for geocoding is using the location of a global positioning system (GPS) through a standard operating procedure. A review of sources of coordinates in national HFDBs across SSA showed that GPS locations are only available for a subset of facilities, predominantly public hospitals (25, 49). Locations for most other facility types are derived indirectly by digitizing paper maps or using proxy locations from other infrastructures such as schools, digital gazetteers (a list of geographic place names and their coordinates), and base maps such as Geonames, Google Maps, Bing Maps, and OpenStreetMap (OSM) (25).

As geospatial technologies evolved and developed in sophistication (50), geocoding sources have improved, leading to a shift from on-screen digitization (most rudimentary) to reasonably complete digital gazetteers and base maps. For example, the number and type of contributors to OSM have grown, including voluntary mapping communities, governmental, non-governmental, and humanitarian organizations, for example, emergency health mapping campaigns, national OSM chapters, and “The Missing Maps initiative” (33, 51). This has contributed significantly to accurately geocoding health facilities, for example, in Kenya and Senegal (8, 27). Further, opportunities exist to exploit recent advances in geocoding and value addition from HFDBs. For example, the Starlink satellite system offers high-speed internet, especially in isolated and remote areas, and can be leveraged in health services mapping or the use of artificial intelligence (AI) in predicting health service access, especially in urban areas affected by traffic congestion and extreme climatic events (52, 53).

The growth in geospatial technologies has been tremendous to the point that we can attribute disease events (cases, outbreaks) and health services (availability and provision) to a specific location (50). However, the application of these technologies has been heterogeneous across world regions. In high-income countries, locations of health facilities and other essential services (e.g., pharmacies, opticians) are part of well-defined registers with location parameters such as post-codes and street addresses. However, many low- and middle-income countries (including those in SSA) do not have such well-defined addresses, including defined location addresses of facilities (54). There are efforts to develop systems that overcome the limitations of postal addresses. For example, through the what3words project (55), which divides the world into a grid of 3-m squares, assigning each square a unique three-word combination. This method is useful in regions with limited address infrastructure, such as remote areas and informal settlements.

Maps of health facilities in sub-Saharan Africa

The use of geographic coordinates to make maps of health facilities is not new in SSA. During the colonial and immediate post-independence periods, SSA countries saw the value of displaying health facilities on maps. At the time, these were often hand-drawn maps of hospitals and health centers or incorporated as part of a country’s atlas (56). For example, in the Democratic Republic of the Congo (DRC), the oldest available health service provider map archived at the WHO is dated 1953. In Kenya, mapping of facilities began before independence in the 1950s (56). At the subnational level, district health offices displayed a map showing their health facilities, often hand-drawn or painted on a wall, a practice that has continued to this day (Fig. 1).

Over time, the practice of having a complete geocoded facility list represented on a map weakened. The lack of regular updates of facility lists that existed from the 1950s, coupled with an exponential, and in some places (predominantly urban areas), unregulated growth of the private sector, may have contributed to this situation. The lack of a single authoritative government-led HFDB has resulted in a situation where unofficial efforts/initiatives have had to be made to compile such data. While it does not necessarily follow that if a list is not official, it is of inferior quality; it is nonetheless likely to suffer from lack of data completeness, accuracy, precision, reliability, and not be openly licensed. Therefore, countries might be in positions with several conflicting and fragmented lists (with and without geographical coordinates) that were not officially endorsed by governments and often in portable document format, which made interoperability difficult (11, 57, 58). These factors and the advances in geospatial science led to the gradual resurrection of geocoded lists of health facilities. Although some improvements have occurred to geocode locations of health facilities, the situation has not improved in the last decade (14). Given the advancement of geospatial science the lack of progress is no longer due to technical issues, but due to a lack of financial resources needed to create and maintain HFDBs, political will, and coordination, which motivates our call for renewed attention. It presents the business and use case for an accurate and openly licensed baseline of health facility data.

Typologies of facilities

Each country has a unique typology (a way of classifying the levels of facilities ranging from community health promotion, mobile clinics to tertiary referral hospitals) within the structure of the healthcare system, in both the public and private sector. Ideally, the categorization or levels of facilities should contain meaning relevant to the health system, such as ability to provide various services, ability to receive referrals, responsibilities for lower-level facilities, and numbers and cadres of health workers needed (59). Therefore, two variables will be needed across HFDBs: a local (country-specific) typology of facilities reflecting the availability of services or levels of specialization and a harmonized typology (adaptable to accommodate these variations) across countries that allows comparison and cross-country analysis and planning. This inclusivity will enhance the comprehensiveness and relevance of the database across different contexts.

Comprehensiveness of private and public sector

HFDBs, when originating from national authorities, have, by default, included public health facilities. However, completeness has been biased toward higher-level facilities such as hospitals. In creating a HFDB, the first step should be to adequately capture higher-level facilities in terms of location and services offered in both the public and private sectors. On the other hand, the care providers at the lower level, including the primary level, CHWs, pharmacies, and outreach centers, both formal and informal, should not be forgotten. An improved understanding of the number and distribution of such providers will have major implications for service delivery and optimization. This is important as countries balance service coverage through a mixture of fixed and outreach sites (60). A recent call and guideline to implement a HFDB for CHWs have been made available (61) as a step in the right direction. For example, through the use of geospatial modelling, optimization studies in the deployment of CHWs in Sierra Leone (3), Mali (2), and Madagascar (4) have been conducted.

Despite their contribution to essential health service provision (62), private sector facilities have often been excluded from or underrepresented in HFDBs. Private facilities are a significant part of the healthcare system in SSA. In Benin (15%), Cameroon (24%), Congo (15%), DRC (16%), Eswatini (30%), Kenya (16%), and Uganda (17%), at least 15% of women gave birth in private sector facilities based on the most recent Demographic and Health Surveys (DHS) (63). These proportions were even higher (> 30%) for care or treatment-seeking for children with fever in the private sector in Benin, DRC, Gabon, The Gambia, Kenya, Liberia, Nigeria, Tanzania, and Uganda (63). However, the private sector is often not integrated, regulated, or accountable to the health system, with poor reporting rates in the routine health information system (64). Most private facilities are in urban areas, and their operation location often changes, with a high turnover of closing and opening of new facilities. Private facilities are heterogeneous in size, services they offer, profit motives, and quality of care (65, 66). This may explain why regulating and including the private sector in HFDBs is challenging.

While it is the role and mandate of the government to regulate private facilities, there are additional incentives in having private facilities as part of the HFDB. A substantial proportion of the population seek care from private facilities (63), the government contracts private facilities to provide healthcare and decongest the public sector(64), reimburses using health insurance-based finances (67), and distributes medicines, supplies and equipment such as bed nets to and through private facilities (64, 68). For such applications, governments need to know where private facilities are located and the services they provide. Therefore, we call for facilities in the private sector to be included in national HFDBs. This would also enable future policies to improve their integration in health systems (65, 66, 69). The attribute of ownership would specify the different sub-categories of private facilities, such as those operated by individuals, faith-based organizations (FBOs), non-governmental organizations (NGOs), for-profit companies, and insurance-based schemes.

Services and capacity offered

South and colleagues posit that facility lists are of limited use if the services provided by facilities are not captured within a HFDB (14). We view the expansion of HFDB attributes regarding services and the capacity of facilities as the second stage in developing a comprehensive HFDB (Table 1). However, a cross-sectional review conducted during the COVID-19 pandemic in SSA showed that service attributes were only available in the Kenyan HFDB (14). Few countries have accurate, up-to-date information on health capacity or readiness to provide quality services despite decades of investments in health information systems (24). Incorporating service availability and capacity at of each facility will support health service delivery and planning across different health domains, for example, assessing health accessibility and marginalization with respect to different types of services such as routine care (8), specialized and emergency care (49, 70,71,72), diagnostics, and mapping vulnerability, and risk preparedness for emerging pathogens (43).

Significant benefits for healthcare planning could have been realized if data on facility services and capacities had been available across SSA countries during the COVID-19 pandemic. For instance, these data could have enabled precise assessments of surge capacity for hospital and intensive care unit (ICU) beds and geographic access to critical care, providing actionable insights for policymakers and stakeholders (73). Data on oxygen availability or the number of hospital beds or doctors could have informed the planning and response (14, 73, 74). There are examples of studies assessing capacities and access to surgery and EmONC (49, 70,71,72) but less on other critical elements of service planning, including hospital care and diagnostics, due to a lack of data on services and capacity.

Catchment areas and population

Assuming that an accurate HFDB is in place, a major challenge in resource planning and allocation is determining an accurate catchment population. This creates two challenges. First, a high-resolution and accurate population denominator disaggregated by age and sex is needed. Second, accurate and robust health facility catchment areas are needed. We address both of these elements in turn.

There is a need to maintain population data alongside a HFDB in order to ensure that the systems and service delivery are meeting the needs of its populations. In most SSA countries, population censuses are conducted every decade. In some cases, the most recent censuses were conducted over 30 years ago, such as in the Democratic Republic of the Congo and Somalia. The census data are not only coarse in the temporal domain but are usually also aggregated into coarse subnational units such as districts, counties, wards, or local government areas in the spatial domain. The long repeat period and low spatial resolution mean census data are inadequate to determine the catchment population at the health facility level. To this end, there a number of modelling initiatives that take in census data and range of covariates to disaggregate administrative level data to fine-scale (raster cell level or gridded level) estimates using spatial statistical approaches (machine learning, areal weighting, or dasymmetric approaches) (75, 76). These approaches further disaggregate the estimates by age and sex while also projecting the estimates based on population growth rates. Such initiatives include Worldpop, Gridded Population of the World (GPW), High-Resolution Settlement Layer (HRSL), Landscan Global Population Database, Global Human Settlement Layer–Population (GHS-POP), and History Database of the Global Environment (HYDE) (75,76,77).

With all these high-resolution population datasets, perhaps the most challenging aspect for the end user is the lack of understanding which data product to use and when. The decision is likely linked to the quality of input population census, ancillary data, and the approach used for redistribution. A number of studies have been undertaken to compare these datasets systematically and can be used to inform choices made when HFDBs are being used together with population datasets to determine catchment populations (75,76,77). Finally, it is worth noting that none of these population disaggregation initiatives are based in SSA, where the datasets are required most. This is a good example of what happens when governments do not own, collate, update, or publish their data. Consequently, modelling groups will step up and find a solution. However, this leads to a situation where SSA governments, institutions, planners, and researchers rely on that modelled data instead of real ground information. While the modelling fills in for the gap in data, it also absolves governments of the responsibility to collect and provide its own data.

Even when geolocated health facilities and population data are available in the required format, determining the catchment area—a geographical area delineated around a health facility describing the population that uses its services—has remained a challenge (37, 38). Due to inaccurate catchments, facility-level estimates such as immunization coverage sometimes exceed 100%. A range of simple to complex approaches can be employed to define catchment areas, including buffers, and Thiessen polygons or based on modelled travel /distance or use of advanced spatial statistical models (37, 38). The choice of either of these approaches is based on the availability of data especially geocoded data on the residential addresses of those seeking care and related care-seeking behavior. We urge countries to create, maintain and updated health catchment areas in the same breath as population and HFDBs. These will provide the foundation for the computation of population denominators for applications such as surveillance and for hard-to-reach populations.

Updating HFDBs

Many existing country-level HFDBs lack a temporal dimension which allows changes in the capability, functionality, facility type (such as upgrades or facility designation), and attributes of facilities to be captured (25, 27). However, we argue that a HFDB should be a living database and should, therefore, be updated and validated continuously. The mechanism for tracking changes in health facilities, such as closures, openings, relocations, allocation to administrative and health zones, and changes in capacity and facility type, is critical. This is particularly relevant for facilities outside the public sector, where some of these changes may be more frequent and less regulated. The HFDB resource package by WHO guides countries on various aspects that should be considered when updating a HFDB and how often it should be updated depending on a country’s local context (1).

Countries can align periodic updates of the HFDB with other regular activities, such as the delivery of medical supplies. Other possible avenues to take advantage to update facilities might include an annual re-accreditation system, where facilities are only re-accredited upon updating any annual changes (23) or through a DHIS2 module which prompts for updating any changes on an annual basis. The updates could also be mandatory through an act of parliament. For example, in Kenya, The Independent Electoral and Boundaries Commission is tasked to review the names and boundaries of constituencies at a predefined time interval in the Kenyan constitution.

Understaffing due to resources in relation to updating HFDBs is often a limitation in the regular update of HFDBs, which are usually highly dynamic and politically sensitive. Staffing is not the only issue. The rate at which attributes of facilities—particularly operational status, services, and capacity—change can be high, especially in urban areas. For example, within a month, a private clinic might be closed by health authorities, another facility with the same name opens two streets down, and a pharmacy with a different name opens in the original location. This makes it resource intensive to maintain an up-to-date national HFDB. Overall, the value of essential minimum attributes of a HFDB area is illustrated in Fig. 2.