GIS-technology may soon be controlling beer delivery in Kinshasa, Democratic Republic of Congo. It will enable the company to keep track of every lorry as a Global Positioning System (GPS) continuously sends its geographical co-ordinates to a central computer that displays its position on a digital town map. The growth rate of industrial GIS applications has been around 25–35% per annum over the last decade (Maguire 1991), and they are slowly but steadily invading the universe of tropical medicine, epidemiology and public health. The latter discipline seems, however, rather resistant to their conquest (Mott et al. 1995). Would beer distribution warrant more advanced technology than the delivery of vaccines and drugs to underserved populations? Or are GIS too complex and beyond the grasp of public health experts? GIS are nothing less than the spatial dimension attached to data, allowing for mapping and, to a varying degree, for analysis of spatial information. An important feature of most GIS is the multilayer structure of the database. If the basic map layer contains the administrative boundaries of an area, other information layers can be superposed and displayed simultaneously. Moreover, GIS allow for a number of spatial transformations of data, e.g. the selection of all events within a certain radius around a point location. As such, GIS facilitate the step from descriptive to analytical epidemiological work, and help to raise hypotheses about associations. Advanced features of GIS furthermore include tools for spatial statistical analysis, and for spatial modelling (Gesler 1986). Mapping through GIS can make a substantial contribution to the assessment of environmental health risk (Briggs 1996). Currently, GIS are also being introduced in tropical disease control programs against sleeping sickness, Chagas disease, leishmaniasis, schistosomiasis, guinea worm and malaria (Mott et al. 1995; De Savigny & Wijeyaratne 1996). For national health policy makers, cartographic display can indeed facilitate identification of risk areas, their subsequent targeting and monitoring of interventions to these areas. In their analysis of unmet obstetrical need in Morocco, De Brouwere et al. (1996) showed how mapping can become a tool for public health decision-making at national policy level. There is, of course, more to GIS than the production of brightly coloured maps. Again, the environmental epidemiologists are leading the way in exploiting GIS for research (Briggs & Elliott 1995). Climatic, vegetation, and other data obtained through remote sensing can be combined with epidemiological data to predict vector occurrence (Rogers & Williams 1993). An overview of GIS applications in infectious disease epidemiology can be found in Clarke et al. (1996). GIS applications for health systems research seem to have concentrated on catchment area research (Zwarenstein et al. 1991; Oranga 1995; Gordon & Womersley 1997). But what about health district management? Should district managers succumb to the GIS vendors? Glossy advertising obscures reality, especially at the scale of a district (100 000–150 000 inhabitants). If a district manager wishes to examine spatial differences in the tuberculosis (TB) programme, his or her first headache will be where to find a digital map of the district and corresponding, up-to-date population figures. Detailed spatial and demographic information is a scarce commodity at district level! Second, how to be sure that the geographical references in the TB surveillance data are valid? Residence is a socio-culturally defined concept and the question: 'Where do you come from?' can be interpreted in many ways. Once one has managed to plot TB patients on a map, another problem pops up: the rare disease/small areas issue. Whereas one would not dare draw conclusions about 3 TB patients in a village of 2600 inhabitants compared to 1 in a community of 1500, it looks as if by colouring both areas on the map according to trivially differing incidence rates, this difference becomes pervasively more 'real'. Bright colour palettes tend to silence a statistical conscience about fortuitous differences in the raw data. Despite the above, we do not want to challenge the idea that district management could benefit substantially from an analysis of spatial information. A team discussion on accessibility to health centres, for example, is certainly facilitated by a district map. Looking at a map stimulates the manager's mind more than struggling with figures in a table, just as a graph of an epidemic curve is more eloquent than a list of monthly incidence figures, and we regret that both map and graph are too often absent from the discussion. Still, the question remains: Does one need GIS to locate the Broad Street pump? If the district health team can work out a relevant map with pencil and paper, a computer might render the mapping job slightly more attractive and efficient. The added value (and the danger) of GIS is that it forces one to estimate the population denominators linked to the geographical areas of interest. Those fixed topographic denominators have the potential to facilitate spatial and temporal comparisons of disease occurrence as well as service utilization, but they are critically dependent on the accuracy of the source data (Twigg 1990). We do not need quantified garbage-in, garbage-out exercises, but reliable denominators coupled with extensive field knowledge of the district to produce meaningful information for action. An integrated mappable surveillance system could so provide ground for communication, and hopefully enhance collaboration, between peripheral and central levels of the health services on the one hand and health services and disease control programmes on the other hand (Arbyn et al. 1997). Finally, which of the GIS packages would be most appropriate at district level? The prospective GIS-user has everything to gain from a careful price/quality comparison of the available software and from making a well-informed choice. The most obvious need in routine health information systems is cartographic display of data on coverage and utilization rates, which does not require any advanced GIS features. Simple freeware, such as EPIMAP (Dean et al. 1993), which only contains the mapping function and is thus strictly speaking not a full-blown GIS, can suit perfectly the information needs of district management, and further development of this type of software should be encouraged. Our major unease about the GIS trend today is that it might well give rise to a new wave of technology-driven priority setting within the health sector. In an environment of constrained resources, introduction of GIS implies an important opportunity cost in terms of manpower, staff time, supplies and equipment. Moreover, the risk exists that GIS will be largely unused by health workers at the periphery, even if enjoyed by researchers. We cannot recommend GIS to any health district manager unless a thorough cost/benefit analysis is made. It will be important to evaluate the pilot schemes impartially and to appraise their values, uses and limitations. Still, in its drive to maintain its exponential growth, the GIS industry seems eager to expand its market into the health service sector. Can you resist?
Stored only when both papers are in the corpus — 18 citations exist upstream.