The studies of the spatial characteristics of disease mentioned in the introduction have one quality in common: they are all purely (verbal) descriptive accounts. The obvious tool for the study of spatial distributions, the map, is missing from these works. It was not until the second third of the 19th Century that maps began being used as tools to aid in the study of the spatial distribution of disease. The advantages of using mortality or morbidity maps in conjunction with verbal descriptions of the spatial variations are many. One of the early researchers to make use of maps in this way was Petermann, who constructed maps of cholera mortality for the British Isles in 1852. He described the advantages of using maps to display medical information as follows:
"The object, therefore, in constructing Cholera Maps is to obtain a view of the geographical extent of the ravages of this disease, and to discover the local conditions that might influence its progress and its degree of fatality. For such a purpose, geographical delineation is of the utmost value, and even indispensable; for while the symbols of the masses of statistical data in figures, however clearly they might be arranged in systematic tables, present but a uniform appearance, the same data, embodied in a map, will convey at once, the relative bearing and proportion of the single data together with their position, extent, and distance, and thus, a map will make visible to the eye the development and nature of any phenomenon in regards to its geographical distribution." (1)
The display of mortality or morbidity information on maps can be, in the eyes of at least one worker in the field, the most important contribution a researcher can make. (2) In the hands of a skilled map-maker, a spatial distribution can, in many instances, assume a new relevance when displayed on a map, and unnoticed information may be gleaned from what previously seemed to be a random collection of data. (3)
Disease maps, in their simplest form, communicate the location of a disease in some manner, and are thus an enhancement of verbal or tabular methods of communicating that information. That is, they serve to describe the spatial distribution of the disease in reference to basic benchmark features such as coastlines and political boundaries. A further advantage of maps is their suitability to the display of more than one factor simultaneously. For example, a map displaying both the distribution of a disease and also the topography of a region is easily constructed and interpreted.(4)
Multiple-factor maps have proved to be beneficial in establishing tentative etiologic hypotheses. Maps of the prevalence of Burkitt's lymphoma (a particular cancer common in central Africa) indicated high incidence in the warm and relatively moist areas around Uganda and in the coastal regions of Kenya and Tanzania. The maps also showed the rarity of the disease in the highlands of south and west Uganda, central Kenya, and northern Tanzania. The tumour was also shown to be rare in central Tanzania. Analysis of the maps has led to the conclusion that Burkitt's tumour is related to the presence of malaria, and this relationship has been shown to hold in maps of the world and in detail in maps of Uganda.(5)
Since the first disease maps appeared in the 19th Century, they have been used quite extensively for both descriptive and analytical purposes. In the following sections, a brief history of mortality mapping will be presented, with emphasis on the development of new mapping methods and on problems associated with those methods.
The general use of disease mapping seems to have been initiated by the great outbreak of cholera in Europe in the first half of the 19th Century. The most notable example of the mortality maps produced in this period was Dr. John Snow's map of the distribution of cholera deaths in the Broad Street district of London in the year 1854. On his map, produced in 1855, Snow portrayed each death from cholera as a black dot positioned at the residence of the victim. The map contained as base information the streets of the area and also, significantly, the location of the public water supply pumps in the area. Snow was then able to show that "the incidence of cholera was only among persons who drank from the Broad Street pump." (6) Following this disclosure, the Broad Street pump was made inoperative and the appearance of new cholera cases in the area ceased almost immediately.
Snow's map was significant in terms of its results, but was not the first of its kind. In 1795, Valentine Seaman published two dot maps to illustrate the distribution of yellow fever in New York. Dr. Robert Baker illustrated his "Report of the Leeds Board of Health" (1833) with a "cholera plan" which was a map of Leeds in which areas affected by cholera were shaded red. Based on the resultant pattern of red areas on his map, Baker observed "how exceedingly the disease has prevailed in those parts of the town where there is a deficiency, often an entire want of sewerage, drainage, and paving." (8) Although it is not known whether the construction of Snow's or Baker's maps actually resulted in their formulation of cause and effect relationships with environmental conditions, the maps certainly aided in providing clear support for certain hypotheses and clear counter evidence for others.
Many other maps of cholera incidence or mortality were constructed in England during this period. Among these were those by W. P. Ormerod (1848) of the situation in Oxford, and T. Shapter (1849) of Exeter. At the national scale, the distinguished German geographer Augustus Petermann, who was living in Enaland at the time, produced a "Cholera Map of the British Isles" in 1852. He shaded areas in proportion to the relative amount of mortality from cholera in those areas from 1831 to 1833. On the basis of this map, Petermann concluded that the districts with large amounts of cholera mortality all "seemed to lie in the lower ground and valleys". However, after further investigation he decided that these areas were attacked by cholera because of their large amounts of population, not their low elevation. Petermann's cholera map of the British Isles included an inset map of London divided into 43 districts in which each district was shaded to indicate the proportion of deaths from cholera to total population during the outbreak of 1832. This is a remarkable map in that it uses a crude mortality ratio as well as class intervals (e.g. less than 1/35, 1/35 to 1/100, etc.). On no other previous medical map could we determine the exact range of mortality in a given area of the map.
Following the publication of Dr. Snow's cholera map, Dr. Henry Acland produced innovative maps of Oxford in which he portrayed the location of cholera cases for various years. These maps were to be used in conjunction with maps he prepared of the topography of Oxford, the location of undrained areas, points of water contamination, and the contaminated rivers and streams in Oxford. Using these maps, Acland found that altitude played an important role in cholera incidence as did the drainage characteristics and the contamination of water resources. (9)
Cholera was not the only disease being investigated with the use of maps during this period. In 1839, Joseph-Francois Malgaigne, the eminent surgeon and surgical historian, presented before the Academie des Sciences in Paris a map of the incidence of hernia in France. What makes this map special is that it is considered, by some, the first "statistical map" of disease. (10) Malgaigne used his map to test various contemporary etiologic theories relating to hernia. For example, to test the theory that hernia had a low incidence in regions where cider was drunk rather than wine, Malgaigne drew a line on his map which represented the approximate boundary between wine growing areas and non-wine growing areas. Using his map, Malgaigne decided that the theory contained too many exceptions and therefore should not be accepted. He used similar cartographic techniques to evaluate the relationship between olive-oil consumption and hernia incidence, and between topography and hernia incidence. Malgaigne's map was the first to go beyond simple mapping and is considered a statistical map because of its use of a transformation of raw data (reciprocal transformation), its technique of agglomeration by departments, its use of shading to represent class intervals (as later done by Petermann), and its overlaying of external environmental information to test hypotheses.(11)
The end of the "golden age" of disease mapping came in the last few years of the nineteenth century. The spectacular discoveries in bacteriology (by the end of the century, the organisms causing tuberculosis, diphtheria, lobar pneumonia, erysipelas, Malta fever, cerebro-spinal meningitis, tetanus, plague, botulism, cholera, dysentery, and wound infection had been identified, isolated, and proved to be responsible for their associated diseases (12) seemed to make all further research in other directions superfluous (13). The last major work involving disease mapping during this period was Dr. Alfred Haviland's The Geographical Distribution of Disease in Great Britain, published in 1875. The maps contained in this book were technically far superior to what had been produced previously. Included were colored plates showing the distribution in 1851 to 1860 of heart disease and cancer, making this work one of the first to treat chronic diseases in this manner. (14) Dr. Haviland, in this work, coined the term medical geography to describe the kind of research represented by his book.
By 1940, the fortunes of environmental medicine had declined so much that the very concept of a geography of diseases had been forgotten by most medical researchers and practicing physicians. (15) It was earlier remarked that this trend has been, to some extent, reversing in recent years. The impetus for this reversal can, unfortunately, be traced to the experiences of World War II. The belligerents discovered that it was vitally important for them to know the affected areas and means of the dispersion of disease. Both sides in the conflict set up special organizations whose task it was to collect the relevant information. The work of these organizations culminated in the production of two "monumental" works: Global Epidemiology by J.S. Simmons et al. (1944-54) which represented the U.S. effort, and the Welt-Seuchen-atlas (World atlas for Epidemic Diseases) by E. Rodenwaldt et al.(1952-1961), the result of the German effort. (16)
The German work (17) is of much greater interest here because of its employment of maps which were not included in the U.S. effort. This difference probably reflects the strong cartographic tradition in Germany (especially at the University of Heidelberg, where the work was produced). The main body of this atlas was published in 1952, with supplements, for example on cholera, published on occasion. The complete Atlas contains 120 color plates, many of which contain more than one map. The maps are highly descriptive and contain much related information. For example, one map which deals with outbreaks of the plague in the twentieth century has a world map on which are (1) dated arrows to represent the routes of dispersion of all plottable epidemics during the century, and (2) shaded areas to represent regions with active endemic occurrence of plague to the present day, areas with rodent enzootic plague associated with pandemic spread, and regions with permanent enzootic plague over the last few centuries. Following each plate are extensive verbal analyses of the subject matter presented in the maps. The type of analysis used is similar to that employed by Snow in 1855; the method can perhaps be called "cartographic correlation". That is, different types of data are presented on a map or a series of maps, and the researcher observes the spatial relationship among the data types. The method is purely qualitative and can be thought of as a "subjective integration of very intensive map correlations and wide reading." (18)
It is surprising how, without the use of any quantitative statistical methods, the authors of the Welt-Seuchen-atlas come up with such a large number of rather complex cause and effect relationships. For instance, they claim to show a "definite correlation" between the spread of cholera and the amount of water flow. By observing their maps, they notice that "the reduction of the disease gradient of cholera stands in inverse correlation to the fall of the gradient of the velocity of flow of the water." Thus they conclude that "rapid water movement is unfavourable to cholera." (19) One has to admire the ability (or at least the brashness) of the writers of the Atlas for being able to establish so many correlations on the basis of visual "cartographic correlation" and external information.
The 1950's also witnessed the appearance, in the U.S., of a work similar in many ways to the Welt-Seuchen-atlas: Dr. Jacques May's World Atlas of Diseases (20). This atlas contains seventeen map plates dealing with the major infectuous diseases of the world (primarily tropical ones). While the analyses accompanying the maps are not nearly as extensive or intensive as those in the German work, the use of cartographic correlation is evident from the nature of the maps presented. For instance, a map illustrating the world distribution of spirochetal diseases (i.e. yaws and bejel), shades regions according to the intensity of the disease in that region and also illustrates the distribution of rainfall and heat through the use of isohyets and isotherms, respectively. This type of presentation was designed to enable the reader to undertake cartographic correlations of the information presented on the maps.
Following the publication of the two world disease atlases, researchers in the U.S. and Great Britain turned their attention to the mapping of medical information regarding the developed world. This process differed from that of the world atlases because of the greater availability of relatively accurate and disaggregated data, and because of the different nature of disease common in the U.S. and Britain (i.e. relative unimportance of many of the infectuous diseases important on the world scale, and the greater Importance of the chronic diseases). Because of the greater availability of data, it was possible, when dealing with the U.S. or the U.K., to use age-standardized mortality rates in the disease maps. C.M. Howe was one of the first to map age-standardized mortality rates in his article on the geographical distribution of cancer (21) mortality in Wales (1960) . He employed separate maps for male mortality and female mortality, and he shaded each administrative sub-area of Wales in accordance to its level of cancer mortality, for various types of cancer.
Howe's method of analysis and interpretation were similar to those of the writers of the atlases discussed above. By visually comparing his cancer maps to similar maps of air pollution, water supply characteristics, and other hypothesized cancer correlates, Howe was able to identify "possible environmental factors". These included air pollution for lung cancer and soil type for stomach cancer.
At the national level, Murray produced an article on the "geography of death" in England and Wales, (22) in 1962. The choice of England as a study region, by Murray and Howe, as well as by other researchers, reflects that nation's good quality biostatistical data and its large population which is distributed widely enough across the country to insure an adequate number of e deaths in each sub-area to allow for valid analysis. Murray mapped the distribution of mortality rates (age-standardized) for some of the major causes of death; cancer of the stomach and lung, heart disease, tuberculosis, and others. For each disease, the range of mortality rates was divided into class intervals according to an arithmetic division (i.e., each category included an equal range of values).
Murray uses his maps for analytic purposes in much the same fashion as his predecessors. Integrating a visual analysis of the maps with external information such as the location of urban agglomerations, pollution sources, soil types, etc. Murray posits etiologic hypotheses for most of the diseases treated in his article. His conclusions, for instance, concerning the cancers generally agree with the earlier findings of Howe. Murray believes his approach of analysis to be a valid and significant one: "Although the sensitivity limits of such an approach must be recognized, there is increasing awareness that the macroscopic technique of medical geography can assist in focusing the microscopic technique of the biologist." (23) This viewpoint, that the spatial analysis of disease can be useful as a guide for perhaps pointing medical researchers in the right direction, is one which will be echoed by nearly all medico-geographical investigators to the present day. It represents a humbler role for medical geography, and is a response to the tremendous progress made by the microscopic studies of health scientists.
The work of Murray was quickly followed by the ambitious National Atlas of Disease Mortality by Howe, (24) published, in its first edition, in 1963. A number of very important advances appeared in this work and these will be dealt with in later chapters. Most of the maps in this atlas are traditional ones; that is, for various disease types, the range of age-standardized mortality rates is divided into equal categories, and each sub-area is then shaded according to the category its associated mortality rate has been placed in. Unlike preceeding efforts, correlation with environmental or other factors was not an important concern in this atlas. The short accompanying commentary limits itself to very basic discussions of the shading pattern visible on the map. A typical comment by Howe, concerning the stomach cancer map, is "The outstanding feature of the maps of standardised mortality ratios for stomach cancer is the exceptional concentration of high ratios throughout the greater part of Wales." (25) Unlike his previous work, this atlas makes no attempt at etiologic hypotheses. Perhaps Howe, who in his work on the atlas was collaborating with medical people, had become less confident in the ability of visual cartographic correlation techniques to affirm etiologic theory.
The past thirty years have seen numerous changes in medical geography in general, and specifically in mortality mapping. These advancements will be discussed below. However, some traditional works have appeared recently and merit a mention. In the U.S.S.R. (where, incidentally, medical geography is well-supported and considered an academic field unto itself), the Atlas of Transbaykalial (26) published in 1967, contains an extensive section of medicogeographical maps. One reviewer has gone so far as to say that these maps are the first of their kind in the world." (27) The amount of information portrayed on these maps is enormous, probably reflective of the stated purpose of the maps: use by "local health organs for current and perspective planning of health protection measures ..." (28) Among the kinds of information found on these maps are: location of health facilities, mineral water springs, bio-chemical endemics, vector ticks, physicians, mammalian vectors, arthropod-borne infections, etc. The emphasis in these maps is clearly on presenting a complete picture of the health conditions of the region and how those conditions vary over space. Considering the stated purpose of the maps, this is an appropriate focus. There is no attempt made in the direction of etiology; therefore no correlations or similar analyses are undertaken or appropriate. The maps are simply excellent descriptive devices.
Traditional mortality maps can be found in many recent publications. Nearly all national atlases contain one or more mortality maps, and these generally conform to our description of a "traditional" map. As late as 1968, Dunham and Bailar produced 16 world maps, in the traditional fashion, showing the distribution of various types of cancers. (29) An extremely shallow explanatory text accompanied the maps. Notwithstanding the rather crude mapping method, the problems of using data aggregated at a national level, and the lack of adequate data in many parts of the world, the Dunham and Ballar study did serve to add evidence either in favor of or against many contemporary theories concerning causes of cancer. For instance, Dunham and Ballar's map of stomach cancer tends to affirm previous suggestions of high stomach cancer rates in areas where certain ethnic groups (e.g. Scandinavians) predominate.