A spatial national health facility database for public health sector planning in Kenya in 2008

Background Efforts to tackle the enormous burden of ill-health in low-income countries are hampered by weak health information infrastructures that do not support appropriate planning and resource allocation. For health information systems to function well, a reliable inventory of health service providers is critical. The spatial referencing of service providers to allow their representation in a geographic information system is vital if the full planning potential of such data is to be realized. Methods A disparate series of contemporary lists of health service providers were used to update a public health facility database of Kenya last compiled in 2003. These new lists were derived primarily through the national distribution of antimalarial and antiretroviral commodities since 2006. A combination of methods, including global positioning systems, was used to map service providers. These spatially-referenced data were combined with high-resolution population maps to analyze disparity in geographic access to public health care. Findings The updated 2008 database contained 5,334 public health facilities (67% ministry of health; 28% mission and nongovernmental organizations; 2% local authorities; and 3% employers and other ministries). This represented an overall increase of 1,862 facilities compared to 2003. Most of the additional facilities belonged to the ministry of health (79%) and the majority were dispensaries (91%). 93% of the health facilities were spatially referenced, 38% using global positioning systems compared to 21% in 2003. 89% of the population was within 5 km Euclidean distance to a public health facility in 2008 compared to 71% in 2003. Over 80% of the population outside 5 km of public health service providers was in the sparsely settled pastoralist areas of the country. Conclusion We have shown that, with concerted effort, a relatively complete inventory of mapped health services is possible with enormous potential for improving planning. Expansion in public health care in Kenya has resulted in significant increases in geographic access although several areas of the country need further improvements. This information is key to future planning and with this paper we have released the digital spatial database in the public domain to assist the Kenyan Government and its partners in the health sector.


Background
Accurate health information is the cornerstone of effective decision-making and reliable assessment of disease burden and resource needs [1][2][3]. Efforts to tackle the enormous burden of ill-health in low-income countries are hampered by the lack of functioning health information structures to provide reliable health statistics [4][5][6][7][8][9][10]. Central to a fully operational Health Information Systems (HIS) is a basic inventory of all functioning health facilities and the services they provide. Such an inventory requires a spatial dimension, allowing facilities to be linked to the populations they serve and other proximate determinants of health such as environment, poverty and education. This spatial linkage can be provided by geographic information systems (GIS). The use of GIS for health services planning is widespread in developed countries [11][12][13] but there are few examples of their development and operational use in resource poor settings in Africa [14][15][16][17].
In Kenya, a map of health facilities was produced in 1959 [18] but not up-dated until 2003 [15]. Since then, there has been an expansion in funding and resources in the health sector following the election of a new government in 2002 that promoted the establishment of a constituency development fund (CDF) to fund local development projects [19] including the building of new health centres and dispensaries where need was defined by the constituency. To reflect these changes and map the equity of expanded service provision since 2003, we have updated the spatial audit of public health facilities in Kenya against high resolution population density maps projected to 2008.

Developing a National Health Facility Database
In 2003 a national database of government, mission, nongovernmental organization, local authority and private sector health service providers was completed and its assembly is presented in detail elsewhere [15]. In brief, all available health facility listings were reconciled to identify a single, comprehensive list and each facility provided with a unique code based on its location. Facilities were further coded by level of service provision (from hospital at the highest level through to dispensary at the lowest level) and by sector (Ministry of Health (MoH), Mission, Non-Government Organization (NGO) or private). Each facility was then geo-referenced using a variety of available national and district-level mapping sources including 1:50,000 scale topographical maps, on-screen digitized hand drawn maps from district-level reports, digital place names gazetteers and specialized surveys undertaken by research groups [15]. All information was exported to ArcView 3.2 (ESRI Inc., USA) and health facility maps and lists were sent to each District Health Management Team responsible for the over-sight of service provision in 69 second-level administrative areas to finalize checking and confirmation.
By 2007 there had been no concerted effort to update a single list of health service providers across Kenya with a unique spatial identifier. However, since the first facility audit and mapping exercise there have been several notable improvements in information available on both the existence of service providers and their location that could be used to update and improve the database constructed in 2003. These included improved commodity distribution lists used to audit the delivery of new anti-malarial drugs, anti-retroviral drugs and insecticide-treated nets by the MoH and its partners; a nationwide mapping project undertaken using global positioning systems (GPS) of all road and major public institutions close to these roads by the Ministry of Roads and Public Works (MRPW) [20]; and an increasing use of GPS to map district level institutions as part of research projects, development assistance programmes and a nationwide schools mapping project funded by the United States Agency for International Aid (USAID) managed by the Ministry of Education [21]. Finally, the launch of Google Earth in June 2005 [22] has meant that iteratively the imagery of Kenya has improved, allowing visualization of buildings and roads when streamed at high resolutions. This added capacity to locate structures in space has been used to triangulate crude coordinates provided from hand-drawn maps and other less reliable sources to visualize building structures likely to be health facilities and re-positioned more accurately. Over a period of three months in 2008 these combined sources were used to re-configure an updated inventory of health service providers and improve the resolution of geo-referencing. No new information was available on the private sector, a prolific and hard to enumerate source of health care in Kenya. The analysis presented here therefore considers only the public sector providers.

Defining geographic access to public health services in Kenya
According to the national health sector strategy for 1994 -2010 a key benchmark for progress is that of geographic access to health services [23]. The strategy requires that all households in the country are located within 5 km, or a 1 hour travel time equivalent, to a public health facility. Here we have regarded health services as general out-patient care offering clinical services to ambulatory patients and have thus elected to exclude from this analysis the specialist health clinics that may only provide maternity services or specialized care for tuberculosis or mental illness. To compute access we have assessed the number of people within distance bands of each of each health service location in 2003 and 2008 using a 100 m × 100 m interpolated population density map ( Figure 1A) described in detail else-  where [24] and available from the Malaria Atlas Project (MAP) [25]. In brief, a combination of satellite imagery and land cover maps was used to develop models that identified the location of settlements [24,26]. The settlements map was used to redistribute census population counts within the small area polygons resulting in a population distribution map at 100 m × 100 m resolution. The accuracy of this population map was subsequently validated using actual data and was found to be significantly better than other gridded population products [24]. This raster 1999 population surface was projected to 2003 and 2008 using provincial inter-censal growth rates [27]. Euclidean (straight-line) distances were computed from each health facility to each population pixel at 100 m resolution for both 2003 and 2008 using ArcGIS 9.2 (ESRI Inc., USA). The continuous distance map was then classified into two distance bands: population within 5 km; and > 5 km from a public health facility and population counts in 2003 and 2008 extracted within each distance band.

Discussion
The Kenya health sector has experienced dramatic increases in the number of service providers over the last five years but their documentation, in terms of the number of health facilities and their location has been fragmented, programme specific, and difficult to integrate. It is difficult to imagine any effective service planning without an inventory of current providers spatially configure in relation to the distribution of the human population. Despite attempts in 2003 to reconcile all available information on the location, level and management of health services in Kenya [15] there remains no centralized authoritative spatially-defined inventory of health service providers in 2008. The 2003 database was provided to the MoH in 2004 with the intention that the resource would serve as the platform for future planning.
In the absence of an updated, centralized database we have repeated the audit of service providers in 2008. We estimate that there has been a 34% expansion in public health facilities providing out-patient care since 2003, the majority supported by the MoH.  (Table 2). Most areas in this province, however, also registered the lowest proportional increase in population within 5 km of a public health facility in the same period. ***Several of the large, sparsely populated areas of northern part of the country had 100,000 or more people of the population outside of a 5 km of a public health facility accounting for 80% of the population in some of these areas. Areas within provinces in 2B and 2C represent the districts as at December 2007 (n = 69). Since then the number of districts have increased to 149 but the digital boundaries these new districts were not available at the time of the study.   Population estimates were derived using inter-censal growth rates from the 1999 national census [citation [27]]

Province maps of Kenya showing
The construction of new health facilities has led to significant improvement in geographic access by patients to service providers from 71% of population within the national target of 5 km of public health service providers in 2003 to 89% in 2008. In spite of these improvements, sparsely populated, predominantly pastoralist, areas in the north of the country contained between 100,000 and 500,000 people outside of the required geographic access to public health care ( Figure 2C) and accounted for over 80% of the approximately 4 million people living beyond 5 km from a public health service provider in 2008. It is unlikely that any construction of new health facilities will yield significant improvements in geographic access in Central, Nairobi, Nyanza and Western provinces, all of which have 97% to 100% of their population within 5 km of a public health facility. The Coast, Eastern and Rift Valley provinces already have more than 80% of their populations with appropriate geographic access to public health care and only a few strategically located additional health service providers are required to increase access, particularly in their remote sparsely populated reaches. North Eastern province, however, lags considerably behind other provinces in terms of geographic access and significant investment in new health service providers is required to bridge this gap.
In addition to monitoring geographic access, the applications of a spatially referenced inventory of service providers are manifold because they can be linked to other georeferenced datasets. For example we have shown elsewhere that where only a minority of health facilities report their health statistics, these can be used, together with the full dataset, and with appropriate geo-statistical methods, to estimate complete disease burden coverage with acceptable accuracies at national and provincial levels [28,29].Such approaches have been adapted to examine temporal changes in out-patient malaria burdens [30] and the estimation of national anti-malarial drug requirements [31]. At management levels a facility inventory should provide the basis for financial planning and monitoring, stock management, personnel deployment and regular quality assurance sampling. Whether modeling disease risks or planning commodity distribution it is critical that sources of data or recipients of drugs are positioned spatially.
The development of a database of health service providers is a continuous process and its long-term success depends on a number of technical, financial and management issues. In this study, it became clear that the quality and completeness of existing health facility lists are driven by programme-specific interests to the extent that key departments within the Ministry of Health rely on different facility lists, each believed to be the 'universe' of health services in the country. In compiling the lists into a single database a major challenge has been the lack of a consistent naming of health facilities and the absence of a unique and common health facility code making data integration difficult. In addition, although 93% of the public health facilities in the 2008 database are now spatially-referenced, only 38% were mapped using 'gold-standard' GPS method, albeit an 18% increase from those in 2003.
Given that most of the facilities are located in the densely populated middle belt of the country ( Figure 1) and are located close to each other, it is important that the proportion of facilities mapped using GPS is increased to ensure they can be linked accurately to the populations they serve using newly available high resolution maps [24].
With publication of this paper we have released all the spatially configured data we have been able to assemble into the public domain. It is available as a database with longitude and latitude and along with metadata describing how these were established [25] along with KMZ files suitable for export to Google Earth [22]. The MoH is currently up-grading its facility listings [32], however this inventory does not have facility coordinates and cannot be used in any mapping or spatial analysis. We hope that the release of our geo-referenced data will assist those concerned with the measurement of equity and service provision. This database should serve as a template for future work and we still recommend a nationwide census of public health service providers, similar to the national school mapping exercise sponsored by the USAID [21], to serve as a confirmatory exercise and improve the qualityof-care and human resource attribute data necessary to transform a population-to-provider platform into a truly valuable planning tool. An efficient system for regular updating of the health service database also needs to be put in place by the Ministry of Health. Possibilities of expanding such an exercise to the important but poorly regulated and prolific private health sector in Kenya must be explored. production of final manuscript. PWG contributed to data, analysis, interpretation and production of the final manuscript. RWS was responsible for overall scientific management, analysis, interpretation and preparation of the final manuscript.

Funding source
This study received financial support from The Wellcome