HIV/AIDS is the leading cause of mortality and morbidity in Africa 18. Since its appearance more than two decades ago the virus has spread to almost every country in the world affecting an estimated 34 million people 19. Nearly 24 million people in Africa currently live with HIV/AIDS and the epidemic continues to ravage the development prospects for millions of Africans throughout the continent. In 1999, about 3.8 million Africans were infected with HIV during that year, and a total of 10.7 million children were estimated to be orphaned by it15. The 21 countries with the highest HIV prevalence are in Africa. In South Africa, Botswana and Zimbabwe, one in four adults is infected. A child born in Zambia or Zimbabwe today is more likely than not to die of AIDS. In many other African countries, the lifetime risk of dying of AIDS is greater than one in three 15. While prevalence in many west and central African countries has remained relatively low and stable, eastern and southern Africa have experienced explosive epidemics with HIV prevalence exceeding 40% among pregnant women in some regions. Around 5 million new infections are currently occurring annually worldwide, over 90% in developing countries19.

One of the reasons for the severity of Africa's HIV/AIDS epidemic is the high prevalence of other sexually transmitted infections (STIs) and the inadequacy of STI services. Another reason for the recent rise in HIV in Africa is the gradual adaptation to new environments, for example, as people migrate from rural to urban areas in search of work. However, the spread of sexually transmitted diseases can also be sharply intensified by crises such as natural disasters, social disintegration, armed conflict and mass population movements20. HIV is especially burdensome as the infection and resultant disease primarily affects young and mature adults in their most productive years (15–25) when older and younger family members are dependent on them. The global HIV pandemic is composed of a series of several smaller epidemics. Even within Africa, where levels of infection are the highest in the world, there is substantial heterogeneity of levels of infection.

Tuberculosis is the leading infectious cause of death worldwide, killing more people aged over 5 years of age than AIDS, malaria, diarrhoea and all other tropical diseases combined. The World Bank estimate that the disease accounts for 26% of all avoidable adult deaths in less-developed countries21. So serious is the threat of tuberculosis that in 1993, the World Health Organisation took the unprecedented step of declaring this disease a global emergency22. HIV infection renders a person infected by Mycobacterium tuberculosis much more likely to develop overt tuberculosis, and the evolution of the disease is considerably accelerated. About 20% of tuberculosis cases in Africa are believed to be related to HIV infection23. WHO has calculated that, unless urgent action is taken the annual global number of deaths could rise from 3 million to 4 million by the year 2004. The need for effective intervention is compelling because tuberculosis treatment is one of the most cost-effective of all health interventions. In response to this re-emerging epidemic, the World Health Organisation is promoting the DOTS control strategy (directly observed therapy, short course) with community based treatment at its core24.

In the last decade, in Africa, the incidence of malaria has been escalating at an alarming rate. Cases in Africa account for 90% of malaria cases in the world 25. Until recently, malaria was ranked as the leading disease in terms of disease burden21. It is now estimated that only HIV has a larger impact on the health of the African population than that of malaria18. Malaria is estimated to cause disease in 400 million individuals in Africa and is responsible for 20–50% of all hospital admissions. Mortality associated with cerebral malaria has not improved in the past 30 years 26and severe malaria anaemia is on the increase27. One study has estimated (using empirical methods) that during 1995, 0.75 to 1.3 million deaths resulted from malaria in Africa and that approximately 80% of these occurred in children < 5 years of age16.

The development of drug-resistant strains of the malaria parasite Plasmodium falciparum has been one of the greatest obstacles to controlling the disease 28. Drugs such as chloroquine, which were once highly effective, are now almost useless for treating malaria in many parts of the world 29. Frequent armed conflicts, migration of non-immune populations, changing climatic patterns, adverse socioeconomic patterns (e.g. gross inadequacies of funds for drugs), high birth rates and changes in the behaviour of the vectors are also responsible for the upsurge30. The upsurge has also been attributed in part to the declining nutritional status of individuals in both urban and rural areas2. Malaria and underdevelopment are closely intertwined. The disease causes widespread premature death and suffering, imposes financial hardship on poor households, and holds back economic growth and improvements in living standards. Malaria flourishes in situations of social and environmental crisis, weak health systems and disadvantaged communities18.

Health systems in Africa face increasingly diverse and complex health problems, rapidly growing populations, and severe resource constraints. Improving the performance of health systems has been identified as a major global health priority18. Health systems' performance makes a profound difference to the quality, as well as the length of the lives of the billions of people they serve. If health systems are poorly constituted and managed, life-enhancing interventions cannot be delivered effectively to those in need. Malaria and tuberculosis are examples of diseases that thrive in the absence of well constituted, effective health systems. This is particularly pertinent for Africa where health systems often perform poorly and are unreliable.

Much remains to be understood about the relationship between space and disease. The spatial dynamics of tuberculosis, HIV and malaria are different because of the different modes of transmission and differing relationships to the environment. For example, tuberculosis (transmitted by respiratory droplets) and HIV (transmitted largely through sexual contact) rely on close human contact for transmission. Malaria however is transmitted by mosquito and is constrained only by the flight distance of mosquitoes. This has been measured in one vector species at a maximum distance of 1.8 km 31. Climatic factors play a large part in determining the distribution of malaria, whereas HIV and tuberculosis are affected more by the social environment. These differences will necessarily affect the types of GIS methodologies used to understand the various spatial components of these diseases.

GIS has been widely applied to the understanding and management of malaria in Africa. For example GIS has been used to generate models of malaria occurrence 3233, seasonality 3435 and transmission intensity 363738394041 using climatic and remotely sensed data. The outputs of such models have been combined with population data 1642 to estimate population exposure, mortality and morbidity 1642 and to analyse 43 and project 4434 the effects of climate change on malaria. GIS has been used to map malaria vectors 454647, vector habitats 48 and infection 49. It has also been used in the management and control of malaria 505152, to measure the effects of access to malaria treatment 53 and to evaluate the effects of intervention strategies 39. The above studies were undertaken at scales ranging from micro to continental.

We were only able to locate a handful of published studies using GIS to study tuberculosis in Africa, all of which were undertaken in South Africa. GIS has been used to map tuberculosis cases in an Urban area in the Western Cape of South Africa 54 and analyse childhood tuberculosis in two urban communities of Cape Town 55. In a rural area of KwaZulu-Natal, GIS has been used to analyse the distribution of treatment points and the effect of community-based (as opposed to facility-based) treatment on increased access to nearest treatment supervision point 5657.

GIS could have an important role to play in tuberculosis control programme management, service development, and research. In terms of planning and managing the service, GIS can assist in the planning of the number and distribution of the supervision points as proximity of treatment is one important factor in promoting adherence to treatment. Much remains to be understood about tuberculosis transmission dynamics in developing countries58 and GIS will be a useful addition to molecular techniques and conventional epidemiology, in elucidating transmission pathways, and clusters of multi-drug resistant cases for example.

Although several studies analysing geographic variations in HIV in Africa have been conducted 5960616263, and the importance of place in targeting areas for priority intervention64 has been emphasised, only one published study could be located that applied GIS to the analysis of HIV 65. The study provided some evidence for an ecological relationship between transport accessibility (distance to roads) and HIV prevalence. This was believed to be related to the amount of sex work taking place along the major routes as well as the higher mobility of persons living near transport routes. However, this relationship needs to be tested at an individual as well as at an ecological level. Though studies have documented heterogeneity in the geographical distribution of the HI virus 666768, much remains to be learnt about the causes and nature of this heterogeneity. Most research has focussed on temporal analysis ignoring the spatial dimensions of the HIV/AIDS epidemic. Yet, spatial analysis may be an important tool to monitor the epidemic, predict future treatment demands and to target areas for public health interventions. Work in Europe and North America [e.g. 6970] has focused largely on the distribution and diffusion of the disease 71. In addition, the delimitation of high-risk areas (based on the distribution of co-factors) using standard risk analysis techniques could prove invaluable in Africa. Furthermore, the technology could also assist in the optimal spatial organisation of health care delivery including home-based care. The difficulty of obtaining HIV data and the stigma associated with disclosure however, is a major obstacle to the use of GIS in HIV research in Africa.

Surprisingly, there were few published examples of the use of GIS in health systems research in Africa but there is an encouraging amount of work in progress. One group used GIS to study inequalities in population per bed ratios and the implications of open access to the private and the formerly white hospital services in the province of KwaZulu-Natal, South Africa 72. One study used GIS to equitably distribute fieldworker workload in a large health survey. The methodology predicted average inter-homestead walking time and divided the heterogeneous study area into units of equal workload 73. The author suggests that an extension of the same methodology can be used to optimally distribute community health workers and tuberculosis DOT supervisors, for example. Another study analysed modal patterns of fixed and mobile clinic attendance across an integrated rural health district 74 and developed indices to analyse the relative attraction and repulsion by the various clinics in the district. The most important outcome of the research was the development of a composite measure of clinic usage and inter-clinic interaction based on the ratio of total actual versus predicted distance travelled to attend clinic. The same data set has been used to validate a model of travel time to the various clinics based on a network analysis of a road network. Relative clinic attraction and inter-clinic interaction were again studied using travel time as the denominator in the index (F.C. Tanser & K. Herbst, In prep., 2002). A study using a similar methodology is underway in the Rufiji district of Tanzania to investigate the relationship of wealth quintiles and health outcomes to travel time to nearest health facility (D. de Savigny et al., In prep., 2002). Researchers are using GIS to investigate the relationship between clinic access and maternal and child health indicators in rural Kwa-Zulu Natal South Africa (J. Tsoka et al., In prep., 2002) In four districts of Kenya, GIS is being used to capture and model both population's access and utilisation of health services with a view to increasing the effectiveness of malaria treatment coverage (A. Noor et al., In prep., 2002).

Limited physical access to primary health care is a major factor contributing to the poor health of populations in developing countries75. The world health report of 2000 18 was dedicated to improving the performance of health systems. Health systems performance make a profound difference to the quality, as well as the length of the lives of the billions of people they serve. However, an important omission from the report was the spatial aspect of health systems research. GIS can be used to effectively spatially analyse health systems coverage and identify deficiencies. The potential exists for GIS to play a key role in rational and more cost-effective health service planning and resource allocation in Africa.

GIS is largely technologically (as opposed to research) driven. Some of these global technological trends are irrelevant to health research in Africa at the present time. However, some global trends (both technological and non-technological) are of significant relevance to Africa's health crisis.

It is becoming clear that although GIS started out as a technological tool, it is rapidly evolving into a science in its own right76, albeit in embryonic form. At present it lies somewhere along the continuum between the two. As software becomes increasingly powerful and new datasets become available and GIS is increasingly used to understand and forecast the dynamics of (particularly environmental) disease, this evolution is likely to continue. A parallel exists between GIS and epidemiology. In the same way that epidemiology evolved into a science in its own right in the 1970s 77, GIS is beginning to be recognised as a science. Like epidemiology its tenets have been established piecemeal 77 with contributions coming from a number of different disciplines, in particular the earth sciences. It is now time to draw the different facets of GIS together under the umbrella of geographic information science.

Computer hardware is becoming increasingly cheaper and more powerful, so that even complex analyses of GIS and image data can be carried out on a desktop computer. At the same time, commercial software has been developed into stand-alone solutions capable of performing increasingly complex tasks through increasingly user-friendly interfaces. Whilst there is an increasing amount of free software, the commercially available comprehensive packages remain expensive 11.

Since the 1^st May 2000 the accuracy of off-the-shelf global positioning systems (GPS) has improved by an order of magnitude. Low cost units can now perform tasks that they previously weren't suitable for. This development is likely to result in a sharp increase in the number of geo-referenced health projects making use of GPS technology in the near future.

The paucity of qualified staff, which has prevented many GIS projects from surviving the donor involvement phase, is a major problem in Africa 78. GIS applications in Africa are often found to be initiatives funded or supported by international aid agencies and many are pilot or research projects as opposed to operational systems. They also tend to be controlled by outsiders, not by African scientists79. If GIS are to be useful and effective, then they must be introduced by local scientists who understand both the technological and the socio-economic context in which the systems are to operate. Training creates capacity and leads to an increase in terms of data needs. It however also provides the capacity to fulfil these needs and the new products that result are often of value to many other sectors. Capacity development of African staff should therefore be prioritised.

In addition to lack of capacity, a lack of suitable GIS data sets is a major impediment to the growth of GIS in Africa. The access to spatial data (which are fundamental to any GIS application) continues to be difficult and expensive10. This is not specific to health but to all sectors that utilise GIS. There are similarities in the field requirements for using GIS between forestry, ecology, archaeology and epidemiology that could provide substantial benefits by the sharing of experiences and the pooling of resources11. However, much of the spatial data collection efforts within Africa have been conducted in a decentralised and uncoordinated manner. Inter-sectoral collaboration initiatives should therefore be encouraged and receive funding priority. Africa could usefully build projects such as the Global Spatial Data Infrastructure80 (embedded within which is the SDI – Africa project) and the EIS – Africa 81 projects which aim to support ready access to geographic information to support decision making at all scales for multiple purposes. Geographic datasets are being developed for some countries in Africa through these initiatives, but a systematic programme is required to make geographic data readily available for the continent as a whole. A major programme (funded by an international body) is needed to take up this challenge. Priorities include, for example, the digitalisation of 1:250 000 and 1: 50 000 cartographic maps for countries that have them. Similarly, national geo-referenced health facility databases should be established. Inexpensive African data sets include the African data sampler (topographic, boundary and place data)82, long-term rainfall and temperature data 83 and raster population data 84. Development of such data sets are of paramount importance to ensure the growth of all sectors of GIS in Africa.

Widespread availability of small scale digital data (< 1: 50 000) for many countries within Africa is unlikely to ever become a reality. The most cost-effective answer to the data deficit and poor vital registration and health statistics problem in Africa is the establishment of sentinel geo-referenced demographic and health surveillance systems85. This will enable the elucidation of small-scale disease patterns (e.g. diffusion dynamics) that could be modelled using coarser resolution data and the coverage extended. The INDEPTH network is a network of these sentinel surveillance sites, 23 of which are in Africa86. The sites follow up a designated population intensively over time collecting highly accurate demographic, vital event (e.g. births, deaths, migrations) and health data on a routine basis. So far only a small proportion of the sites are fully geo-referenced but this is likely increase with the increase in GPS accuracy, falling prices and the obvious operational and research advantages of fully geo-referenced data. These sites can especially contribute (and have already contributed) to our understanding diseases with ill-defined relationships to the environment due to the detailed longitudinal collection of disease covariates. A recent spatial initiative in health is the West African Spatial Analysis Prototype (WASAP) that used geo-coded demographic and health survey (DHS) data to study the effects of climate on children's nutritional status, and the relationship between economic diversity and reproductive behaviour, as well as study the subnational geographic variation in health indicators at a regional level8788. Following the success of WASAP, more DHS sites have began to geo-code their survey data in an effort to facilitate cross-disciplinary analyses. The increasing availability of regional geo-referenced DHS data will facilitate a more comprehensive understanding of the patterns and processes of demographic and health changes and will lead to an increasing amount of GIS-based analyses of this important data in the near future.

In addition to the geo-coded household datasets outlined above, a large number of remotely sensed data sets, which have been already used extensively in health are available free of charge or at nominal cost. With the emergence of new technologies and techniques within remote sensing, there is likely to be a great improvement in the quality of such data sets and parallel improvement of GIS and related research products89. Nevertheless, it is also true to say that so far, our ability to extract meaning and make useful decisions from remotely-sensed data has not kept pace with the developments in this field.

The issue of scale is one that is poorly understood in the disease arena. Disease patterns and processes evident at one scale are not necessarily evident at another. Moreover, correlations between explanatory variables and outcomes may even be (seemingly) reversed at different scales. This has led to a significant amount of confusion when hypotheses are rejected at one scale and not at another. Sometimes it is advisable to use coarser resolution data to mask out small scale heterogeneity. For example, the malaria modelling at a continental level used climatic data at a resolution of 0.05° 3234. Higher resolution satellite data (sub kilometre) may obscure continental malaria patterns by exposing unnecessary small area variation. Ideally the resolution of the data should be driven by the application. However, given Africa's geographic data deficits, future research is needed to establish how applicable coarse resolution data sets are to modelling high resolution disease-specific dynamics and vice-versa. The above issues are as applicable to temporal resolution as they are to spatial resolution.

Another obstacle remaining to the growth of GIS in health in Africa is to convince role players (often from cash-strapped organisations) of the proven cost-effectiveness of GIS in the health arena90. Even amongst the international scientific community, significant scepticism still exists surrounding the use of GIS technology in health. This problem will diminish in size as GIS continues to evolve. The parallel with epidemiology again warrants mentioning: In the same way that scepticism greeted epidemiologists who hypothesised that a relationship existed between smoking and lung cancer in the 1950s 77, so to will scepticism continue to plague GIS until it is firmly established as a science.

It is encouraging to note that several of the issues cited as obstacles to the growth of GIS in Africa a decade ago 91 have been overcome to some degree. These included the incompatibility of different software formats (data conversion problems), the non user-friendly interfaces of many systems and the lack of good inexpensive/free GIS software. Other obstacles such as the prohibitive costs of hardware have also become less of an issue. Perhaps a review in a decade's time will describe the increasing availability of inexpensive spatial data sets for Africa?

The 'mapping malaria risk in Africa' (MARA) research collaboration is an African research endeavour that makes extensive use of GIS technology. The collaboration has been highly successful in collating malaria data from around the continent, and producing a large number of scientific publications on a limited budget. The outputs of the research were then disseminated to countries throughout Africa in the form of digital (via the stand-alone MARA lite software) and hard copy maps. The collaboration overcame significant data deficits by creating its own base data sets and created a significant amount of GIS capacity in its five regional centres throughout the continent. During the setting up of the collaboration, significant scepticism was expressed by influential malaria scientists as to the ultimate value of a GIS approach, its logistical feasibility and cost-effectiveness33. The collaboration is a testament to the fact that successful GIS initiatives can be undertaken in Africa.

The current software and hardware trends in combination with the realities faced in Africa have given rise to essentially, two broad categories of long-term feasible GIS health applications in Africa. The outputs of the categories will inform one another and are not mutually exclusive and may overlap. The first category involves the use of GIS as a research tool. These applications should seek to provide new insights into the spatial dimensions of disease and new methodologies to more cost-effectively allocate resources to health services. These types of applications will normally use high-end systems with significant analytical functionality and will usually involve a significant amount of additional data collection.

The second category of long-term viable GIS application concerns the use of GIS as a health planning and management tool and for exploratory data analysis. Generally speaking this kind of system will involve a low-end GIS. The primary goal of such a system will be to simply display and overlay basic health data concerning both health care facilities and disease patterns. These systems (normally vector-based) permit rapid manipulations of spatial data and display of the results so that the decision makers can use them for policy decisions. A further step could involve limited spatial queries and analysis such as buffering.

The outputs of the different categories of application will inform one another. As the data is geographically displayed using a management GIS and research questions are derived, collaborations can be initiated with institutions undertaking GIS research to test hypotheses and model disease distributions. Similarly, research GIS applications will inform GIS management applications to plan optimal resource allocation and intervention strategies, for example. The MARA collaboration is a successful example of this type of approach and is embedding several of its research outputs in the freely available GIS software HealthMapper (developed by WHO) for intervention planning in Africa at a district level.