Abstract
The history and development of public health was motivated, illustrated, and implemented with the use of maps. In this chapter, we provide a brief review of the use of maps and related concepts in public health spanning from medieval plague quarantine plans to remotely sensed satellite measures of air pollution. We begin with early maps of infectious disease (e.g., plague, yellow fever, and cholera) including an in-depth review of Dr. John Snow’s famed map of the 1854 London cholera epidemic. We also review the role of mapping in contemporary political debates regarding miasma versus contagion as underlying causes of disease and associated early public health responses such as sanitation and quarantine. We next highlight atlases of disease identifying, documenting, labeling, and mapping endemic areas of known diseases (e.g., yellow fever, cholera, and leprosy) across the globe . Finally, we outline the rise of quantification of observed patterns and the use of spatial statistical techniques to investigate epidemiologic hypotheses regarding geographic variations in disease risk and associations with potential local explanatory factors. Taken together, we find a rich history of mapping in the development, maturation, and modern implementation of public health science and practice.
Keywords
1 Introduction
Throughout human history, astute observers have noted that certain locations seem to be associated with particular outcomes. Western concepts linking health to location date at least to Hippocrates’ famous treatise On Airs, Waters, and Places (Miller 1962). Even today we refer to tropical diseases, cancer clusters, and sick buildings to denote links between areas and (real or perceived) increased risk.
Broadly speaking, the study of public health can be viewed as the study of the incidence (new cases) and/or prevalence (existing cases) of disease or other health outcome, placed in context of the local environment where the term “environment” is very broadly construed and can include (among many other things) exposures to pollutants, pathogens , insect vectors, local variations in social and behavioral determinants of disease, regional health policies, and the impact of interventions. In many cases, these motivating questions inherently include a spatial component, i.e., many questions of recurring interest in public health begin with “where”. As an illustration, consider the following:
-
Where, when, and within which population subgroups are risks of disease highest?
-
Where are environmental exposures the highest?
-
Where are health interventions implemented, and how consistently are they applied?
-
Where do we find concentrations of social determinants of disease (e.g., high poverty, low access to fresh fruits and vegetables, low access to health care, low access to transportation)?
-
Where is the strongest association between a local exposure and local health outcomes?
The key concept is that a person’s location links one to a surrounding local context of exposure(s) experienced by and impacting the health of the individual.
This notion that one’s location impacts disease risk motivates the mapping of exposures, cases, potential risk factors (e.g., age, race , sex), and the numbers and composition of the at-risk population. As illustrated throughout this text, we consider mapping to be an active process of arranging, summarizing, and visualizing geographic data in order to reveal patterns driven by location with the goal of improving insight, enabling predictions, evaluating real or potential interventions, and, ultimately, gaining new understanding regarding potential causes.
The path from pattern to process is long and winding (and often not unique) with many potential missteps and challenges when it comes to inferring true causality, but the history of mapping in public health illustrates that, while we may not jump directly from observed pattern to true process, the geographic analysis of spatially-referenced public health data often provides important steps forward in our understanding of and steps toward control of the risk of adverse health outcomes.
In the sections below, we review the early history of medical cartography and the mapping of disease , the expansion to the development of atlases of disease , and the growth and development of quantitative spatial analysis tools for public health. In each case, we find contributions of mapping, sometimes in the forefront, sometimes behind the scenes, involved in the study and practice of public health.
2 Early Examples: Miasma and Contagion, Quarantine and Sanitation
As summarized in Koch’s (2005) Cartographies of Disease, the earliest extant maps of disease on record include Arrieta’s 1694 maps of quarantine plans for plague in the province of Bari in the then-kingdom of Naples. The impact and fear of plague drove the local government to plan for and map a quarantined area, an extreme acknowledgement of the locality of disease risk , i.e., a strict interpretation that disease occurs here and must be kept from moving there. Note that quarantine does not necessarily acknowledge the cause(s) of the local increases, but presumes that by limiting interaction with the location and/or the individuals within it, one can limit risk in those outside the impacted area. The practical aspects of quarantine can involve severe consequences by condemning those inside to high risk as dramatized in Camus’ The Plague (1948) and observed quite recently in the response to the 2014 Ebola outbreak within poor, high-incidence neighborhoods of Liberia’s capital of Monrovia, which revealed that quarantine as a concept often fuels the polemics of political pundits more than providing any positive impact on the practice of public health.
Moving beyond the identification of high local risk for the purpose of containment, the 1700s and 1800s saw a use of disease maps in the debate between miasma (bad air) and contagion (transmission by infection with an infectious agent) as the cause of epidemic diseases. The debate not only involved the causes of disease but also the proper public health response to them, with rapid advances in both areas due to maps. Maps provided insight into the role of bacteria and other pathogens , e.g., in drinking water, in the cause of disease. The period saw the initiation of the “sanitation movement,” with an increased focus on clean water and the development of citywide sewer systems, a movement also clearly influenced by maps. However, as well known and documented examples illustrate, in neither case were maps free of misunderstanding and misinterpretation.
As a first example, New York public health official (and avowed sanitation supporter) Valentine Seaman’s 1798 “spot map” of yellow fever provides the first known map of resident locations of individual cases of disease (Stevenson 1965; Garfield 2013), revealing a concentration of cases near the docks where ships often brought cases (sick individuals infected in endemic areas and moved to New York) and where swampy ground and the build-up of sewage effluvia from “areas of convenience” led to foul odors and, as noted by Seaman, concentrations of mosquitos (Fig. 1). As a proponent of sanitation, Seaman interpreted the concentration of mosquitos as further evidence of local miasma, not as related to the transmission of disease. The recommended responses of cleaning up the areas under the docks were effective, even if the primary route of transmission was missed through focus on preconceptions and assumptions.
Following, but better known than Seaman’s work, is the famed investigation of the 1854 London cholera outbreak by Dr. John Snow. Snow’s detailed, data based, and highly influential work (Snow 1855) results in his listing among the founders of modern epidemiology . A recent London travel guidebook, in a note regarding the John Snow pub (a landmark for epidemiologists worldwide) summarizes the story as follows:
In 1854, Londoners were dropping like flies from cholera until Dr. Snow figured out that the bacteria were carried by water. The water pump he turned off, thereby saving countless lives, was near the site of this pub. (Wurman 2006)
Snow’s innovative use of maps, and his theory of a water-borne cause of cholera make for a compelling story that is often serves as the introduction to discussions of the role of maps in public health . Naturally, the real story is a bit more nuanced but ultimately even more interesting than the brief summary above. Brody et al. (2000), Koch (2005), Johnson (2006), and Garfield (2013) provide readable, detailed, and fascinating summaries of Snow’s work and the role and impact of the maps involved. Briefly, Dr. Snow, as both a physician with a suspicion that water was involved in transmission and a resident of the Soho neighborhood strongly impacted by the 1854 outbreak, interviewed the families of, and mapped the residence of, each cholera death in the neighborhood. He also mapped the location of public water pumps and noted which individuals accessed each pump. Brody et al. (2000) and the others note that Snow’s map was not the first map of the Soho cases of the 1854 outbreak, nor the only one considered by the London Board of Health, a group largely in the miasma camp (noting that the disease appeared concentrated in poorer sections of town where one could smell the filth).
The first spot map of cholera cases for the outbreak was provided by Edmund Cooper, an engineer for the Metropolitan Commissioner of Sewers, who was investigating a public concern that the sewer system had somehow connected to a site of a mass burial of victims of the 1665 plague outbreak in London. Cooper’s map presented deaths by address with respect to “gully-holes” where the sewer vented to the surface. Comparing Cooper’s gully-hole map to Snow’s map with public water pumps reveals that the gully-holes and pumps are largely in the same locations, particularly around the Broad Street pump which was the focus of Snow’s investigation and, ironically, which was slightly misplaced (two doors from its actual location) in Snow’s original map. Both maps show a high concentration of deaths in the vicinity of the Broad Street pump on Snow’s map and near the Broad Street gully-hole in Cooper’s. Both Snow and Cooper supplemented their maps with additional field work. Cooper, arguing against the plague burial theory, noted that sewer workers within the tunnels near the gully-hole had not become ill. Snow, arguing for the impact of the pump, noted that (a) some individuals far from Broad Street who used water from the Broad Street pump had fallen ill and (b) that no residents of a workhouse nearby with its own well had fallen ill. Snow (and others) effectively lobbied the London Board of Health to remove the handle from the Broad Street pump, a dramatic public health response that occurred as the epidemic was abating.
The fact that Snow’s theory was correct makes it tempting to suggest that the spot maps of disease and potential causes provided overwhelming and irrefutable evidence indicating the underlying cause of disease, but it is important to note that proponents of the miasmist theory of disease viewed the same maps and saw clear evidence for their own argument. In Seaman’s case, Seaman himself sees a clear miasmist interpretation of the observed concentration of cases near the docks and sewage, noting the concentration of mosquitos as further evidence of his theory noting “circumstances favoring the rise of putrid miasmata, equally favor the generation of these insects” (quoted in Stevenson 1965). In Snow’s case, Brody et al. (2000) note both (a) that the Committee on Scientific Inquiries of the General Board of Health rejected Snow’s theory in favor of a miasmist interpretation, and (b) that contemporaneous cholera expert Edmund A Parkes in his review of Snow’s monograph noted:
On examining map given by Dr Snow, it would clearly appear that the centre of the outburst was a spot in Broad-street, close to which is the accused pump; and that cases were scattered all round this nearly in a circle, becoming less numerous as the exterior of the circle is approached. This certainly looks more like the effect of an atmospheric cause than any other; if it were owing to the water, why should not the cholera have prevailed equally everywhere where the water was drunk? (Parkes 1855)
In both cases, maps provided important clues but not necessarily incontrovertible evidence alone, and map-readers (Seaman, Snow, and Parkes) clearly continued to bring their own assumptions, presupposition, and bias into their interpretation of the maps. However, while the maps did not provide definitive proof of the underlying cause, they certainly framed both the discussion and the public health response in a manner leading to subsequent discoveries, a more complete understanding, and best response practices.
3 Documenting Spatial Variation in Disease Risk: Atlases of Disease
While Seaman and Snow set the stage for the use of detailed maps of case residences in neighborhoods to identify factors related to local outbreaks of a disease , another branch of medical geographers aimed to define the reach of disease across districts, countries, continents, and the globe . The earliest examples of global maps of the prevalence of various diseases were modeled after similar maps of species of plants and animals and include Friedrich Schnurrer’s 1827 “Charte über die geographische Ausbreitung der Krankheiten” (Brömer 2000) and Heinrich Berghaus’s 1848 map of human diseases in his first major atlas of global thematic maps Physikalischer Atlas (Camerini 2000; Fig. 2). Barrett (2000) provides evidence that a German physician, L.L. Finke, may have generated a single copy of a global disease map as early as 1792. These early efforts spawned an ongoing effort to map the incidence and prevalence of disease across administrative areas in order to identify areas at increased risk for disease.
In the 20th century, Walter (2000) notes a transition of interest from mapping infectious disease to mapping chronic disease with a particular emphasis on cancer, beginning in the United Kingdom as early as the late 1920s and into the 1930s with Stocks’ series of individual maps of regional cancer mortality rates adjusted for local differences in the age and sex distributions of the at-risk population, an important adjustment allowing better comparability between rates from regions with different population compositions (Walter 2000). Howe (1977) extended these efforts to produce an atlas of thirteen major causes of death, leading the way to a proliferation of printed disease atlases , particularly cancer atlases, in the 1970s, 1980s, and 1990s (Walter and Birnie 1991). Updated atlases of chronic and infectious disease continue to appear, with an increasing number of on-line resources available with custom basic mapping, query, and data download capabilities.
The development of disease atlases in the late 20th century corresponded to the rapid increase of three contributing factors: (1) the availability and portability of administrative health data, (2) the development of geographic information systems (GIS) to manage, link, and display large quantities of georeferenced data, and (3) the development of spatial statistical methods to address analytic challenges to the analysis of local disease rates. These three areas of development provided access to increasing amounts of information, the ability to link data from disparate sources (e.g., health outcomes from medical records, local exposures from air monitors, and demographics from the census) onto the same geography , and a growing set of analytic tools to address increasingly spatial questions of interest including:
-
Are there “clusters” or “hot spots” of disease incidence?
-
Do high local rates correspond to high local exposures?
-
What local demographic groups are impacted the most for particular health outcomes?
-
What areas are good habitats for vectors that transmit disease?
While atlases provide relevant data summaries and displays, questions such as those above provided the motivation for development of specialized statistical approaches representing a toolbox for the spatial analysis of public health data, our next topic of review.
4 Quantifying Pattern in Public Health
In order to provide quantitative responses to the questions motivated by the maps described above, public health researchers focused on the development, assessment, and application of spatial statistical tools to address a growing list of geographic questions in public health (Waller and Gotway 2004; Lawson 2008). Cressie’s (1993) landmark Statistics for Spatial Data brought together theory and applications of point process models (is an observed pattern of events more clustered or more regular than we would expect under random allocation of events to possible locations?), geostatistical prediction (using observed values at a fixed set of locations, what is the best prediction of the value associated with a new location?), and regression models for regional data (what is the estimated association between an outcome and covariates measured on the same set of zones?). While not focused specifically on public health, each element of Cressie’s (1993) typology of methods has application in quantifying geographic distributions in public health research as illustrated by linking to the questions above.
Point process methods can address whether an observed pattern appears more clustered than one might expect. A spatial point process defines a probabilistic model of data patterns, and a spatial Poisson process often defines the pattern we would observe under a null hypothesis of no local variation in disease risk . The general mathematics of spatial point patterns tend to compare observed patterns to a uniform distribution in space (wherein an event is equally likely to be observed at any location), but in application to the distribution of disease, it is more common to compare observed patterns to an equal risk setting where a case is equally like to occur in any individual. Since individuals aggregate spatially into villages, towns, and cities , we not only seek spatial concentrations of cases, but we are looking for clusters of cases above and beyond the level of spatial aggregation of individuals observed in the population at risk. Waller and Gotway (2004: Chapters 5 and 6) review general principles for approaches to detect spatial clustering of disease in spatially heterogeneous populations, and outline two families of approaches. The first compares the spatial distribution of cases to that of controls (or non-cases), while the second compares the level of clustering observed in cases (as measured by the number of additional cases expected within a given distance of each case) to the level of clustering observed in controls. The first approach can identify the location of specific clusters, while the second provides a summary of the spatial scale of clustering. Both concepts are of interest and both can detect deviations from the null hypothesis, but the two approaches cannot distinguish between underlying causes of the pattern. More specifically, Bartlett (1964) notes that, without additional information, it is mathematically impossible to distinguish between a pattern generated by local increases in risk for otherwise independent cases (e.g., a cluster due to a shared local environmental exposure), and a pattern generated by interactions between cases (e.g., a cluster due to a set of cases infected by nearby cases) from a single observed pattern of point locations. Interestingly, this challenge is exactly a mathematical manifestation of the public health debates regarding miasma and contagion during the time of Snow and Seaman, and a part of the reason that mapping alone could not distinguish between the two theories.
In order to move past this challenge, two primary types of additional information are of interest in public health . The first is time, since a similar cluster occurring in the same location across multiple time periods suggests a local source of increased risk rather than spatially non-specific contagion, while similar sized clusters at different locations over time provides the opposite conclusion. Exposure data provide a second source of additional information, wherein one links local exposures to local observed disease risk , and assesses associations, often through the use of linear or, more commonly for count-based outcomes, logistic or Poisson regression.
Spatial statisticians addressed two challenges in the application of regression techniques to disease rates observed in small administrative areas. The first, referred to as small area estimation, addresses the tension between geographical and statistical precision in presenting maps of regional rates of disease. Geographically, one prefers small areas to provide a map with local detail, while, statistically, rates based on data from smaller numbers of individuals become increasingly unstable. This is a particular problem for maps of local rates of rare diseases. For example, if public health records suggest a rate citywide of one case per 100,000 individuals, local estimates from a city of 100,000 individuals divided into neighborhoods of 100 individuals will include, on average, one neighborhood with a local rate of 1 per 100, even with no local increase in risk. To address this situation, statisticians developed techniques to “borrow information” from neighboring regions through the use of local weights, typically based on a hierarchical random effects models (Clayton and Kaldor 1987; Besag et al. 1991).
The inclusion of a spatial random intercept within these models also provides a mechanism to address a second challenging analytic problem, namely, adjusting regression estimates to account for residual spatial autocorrelation, the often-observed feature of spatial data that nearby observations tend to be positively correlated, perhaps due to a shared unmeasured (or even unmeasureable) local feature. Allowing residuals to be correlated complicates statistical inference, again due to Bartlett’s result: in this case, from a single data set, it is very difficult to know if an observed pattern is due to correlation or to covariate effects. The hierarchical models mentioned above provide a specific component of the model defining spatial covariance and identifies the combination of covariate values and spatial correlation that provides the best fit to the model.
Collectively, these methods for stabilizing rates and fitting spatial regression models define a class of methods referred to in the public health literature as disease mapping and have seen wide application in the public health literature to provide stabilized estimates of local disease rates and to assess associations with local covariate values such as demographics and exposures (Waller and Gotway 2004; Lawson 2008; Banerjee et al. 2014).
The next area of spatial statistical developments relating to mapping in public health includes the geostatistical prediction of exposure values. To illustrate, consider the measurement of air pollutant levels at fixed monitors located at particular locations across a city. These values offer detailed measurement of pollutant levels but only for those locations. The field of geostatistics grew out of exploratory mining analysis seeking to predict the yield of mines in new locations, and works as follows: Given measurements at a set of fixed locations, we predict a measurement at a new location by calculating a weighted average of the observed data with weights relating to how similar we expect our prediction to be to each of the observations. Assuming positive spatial correlation, i.e., nearby observations are more similar than those far apart, we give more weight to observations near to the prediction location. Geostatistical methods formalize the definition of optimal weights in order to minimize uncertainty in the prediction (typically by minimizing the mean square prediction error or other summary of prediction error). By predicting a measurement value for each of a set of points across the study area, we can now construct a contour map of predicted pollutant levels across the study area and the estimated prediction error associated with each location.
Predicted exposures and estimated associations between exposure and outcomes allow the analyst to make maps of predicted risk for any location. Such maps provide powerful communication tools in public health , by pooling information from multiple sources, linking them geographically, and analyzing them statistically. Placing the results on a map places predicted risk in context for residents and policymakers alike, and clearly identifies the impact of living or working here versus there, a strong motivating question for the first maps of disease . Such risk maps can be basic regression predictions based on exposures and demographics, or more complex, e.g., incorporating satellite images to define locations of favorable habitats for disease vectors and identifying areas where potentially infected vectors and people coincide (Kitron 2000), or identifying areas at high risk for violent crime based on alcohol outlet density (Gruenewald et al. 2006).
The methods above provide a means to quantify spatial patterns for statistical analysis, but this need not mean the map has been subsumed by quantitative, statistical analysis. In fact, Waller (2014) suggests quite the opposite, identifying and calling for new uses of maps to keep a spatial perspective within spatial statistics . As an example, geographically weighted regression (Fotheringham et al. 2002) and spatial varying coefficient (Gelfand et al. 2003) models allow the associations between exposures and health outcomes to vary by location, e.g., the association between local socioeconomic status and birthweight could be stronger in some neighborhoods than others. Such methods use maps to reveal where associations are stronger and where associations are weaker, offering insight into areas where targeted interventions may be more effective (Fig. 3). Waller (2014) also suggests maps of model residuals to reveal where a statistical model fits well and where it does not. Such application require spatial thinking, statistical thinking, and, increasingly, spatial statistical thinking to gain the full public health benefit of mapped data.
5 Conclusion
Our whirlwind tour of mapping in public health ranges from medieval plague quarantines to GIS -linked satellite and health record data, all enabling and addressing spatial questions in public health. As illustrated above, the act of mapping is a critical component of public health research, but rarely a stand-alone solution. Maps identify public health issues, suggest quantitative associations, evaluate the impact of interventions, and communicate results to the neighborhood residents, public health researchers, and policy makers. The range of applications is as broad as the field of public health itself. While our review focuses on chronic and infectious disease , similar examples of the impact of mapping exist in other areas of public health. For example, health services research contains a considerable literature exploring geographic variations in health care delivery and marketplaces (Wennberg and Gittelsohn 1973; Wennberg and Cooper 1998), one goal of which is to enable location-specific valuation of services for Medicare reimbursements in the United States (IOM 2012a, b). Clearly, mapping is essential to public health research, practice, and policy , and the mapping process begins anytime a neighbor, reporter, student, investigator, or government official raises a public health question that begins “Where….”.
References
Banerjee, S., Carlin, B. P., & Gelfand, A. E. (2014). Hierarchical modeling and analysis for spatial data (2nd ed.). Boca Raton: Chapman & Hall/CRC Press.
Barrett, F. A. (2000). Finke’s 1792 map of human diseases: The first world disease map? Social Science and Medicine, 50, 915–921.
Bartlett, M. S. (1964). The spectral analysis of two-dimensional point processes. Biometrika, 51, 299–311.
Besag, J., York, J., & Mollié, A. (1991). Bayesian image restoration, with two applications in spatial statistics (with discussion). Annals of the Institute of Statistical Mathematics, 43, 1–59.
Brody, F., Rip, M. R., Vinten-Johansen, P., Paneth, N., & Rachman, S. (2000). Map-making and myth-making in Broad Street: The London cholera epidemic, 1854. Lancet, 356, 64–68.
Brömer, R. (2000). The first global map of the distribution of human diseases: Friedrich Schnurrer’s 1827 ‘Charte über die geographische Ausbreitung der Krankheiten’. Medical History, 44(20), 176–185.
Camerini, J. R. (2000). Heinrich Berghaus’s map of human diseases. Medical History, 44(20), 186–208.
Camus, A. (1948). The plague. New York: Alfred A. Knopf.
Clayton, D. G., & Kaldor, J. (1987). Empirical Bayes estimates of age-standardized relative risks for use in disease mapping. Biometrics, 43, 671–682.
Cressie, N. A. C. (1993). Statistics for spatial data (2nd ed.). New York: Wiley.
Fotheringham, A. S., Brunsdon, C., & Charlton, M. (2002). Geographically weighted regression: The analysis of spatially varying relationships. Chichester: Wiley.
Garfield, S. (2013). On the map: A mind-expanding exploration of the way the world looks. New York: Gotham Books. Chapter 13.
Gelfand, A. E., Kim, H. J., Sirmans, C. F., & Banerjee, S. (2003). Spatial modeling with spatially varying coefficient processes. Journal of the American Statistical Association, 98(462), 387–396.
Gruenewald, P. J., Freisthler, B., Remer, L., LaScala, E. A., & Treno, A. (2006). Ecological models of alcohol outlets and violent assaults: Crime potentials and geospatial analysis. Addiction, 101(5), 666–677.
Howe, G. M. (Ed.). (1977). A world geography of human diseases. London: Academic Press.
IOM (Institute of Medicine). (2012a). Geographic adjustment in medicare payment: Phase I: Improving accuracy. Washington, DC: The National Academies Press.
IOM (Institute of Medicine). (2012b). Geographic adjustment in medicare payment: Phase II: Implications for access, quality, and efficiency. Washington, DC: The National Academies Press.
Johnson, S. (2006). The ghost map. New York: Riverhead Books.
Kitron, U. (2000). Risk maps: Transmission and burden of vector-borne diseases. Parasitology Today, 16(8), 324–325.
Koch, T. (2005). Cartographies of disease: Maps, mapping, and medicine. Redlands: ESRI Press.
Lawson, A. B. (2008). Statistical methods for spatial epidemiology (2nd ed.). New York: Wiley.
Miller, G. (1962). “Airs, waters, and places” in history. Journal of the History of Medicine and Allied Sciences, 17(1), 129–140.
Parkes, E. A. (1855). Review: Mode of communication of cholera by John Snow. British and Foreign Medico-Chirurgical Review, 15, 458.
Snow, J. (1855). On the mode of communication of cholera (2nd ed.). London: John Churchill.
Stevenson, L. G. (1965). Putting disease on the map: The early use of spot maps in the study of Yellow Fever. Journal of the History of Medicine, 20, 226–261.
Waller, L. A. (2014). Putting spatial statistics (back) on the map. Spatial Statistics, 9, 4–19.
Waller, L. A., & Gotway, C. A. (2004). Applied spatial statistics for public health data. Hoboken: Wiley.
Waller, L. A., Zhu, L., Gotway, C. A., Gorman, D. M., & Gruenewald, P. J. (2007). Quantifying geographic variations in associations between alcohol distribution and violence: A comparison of geographically weighted regression and spatially varying coefficient models. Stochastic Environmental Research and Risk Assessment, 21, 573–588.
Walter, S. D. (2000). Disease mapping: A historical perspective. In P. Elliott, J. C. Wakefield, N. Best, & D. J. Briggs (Eds.), Spatial epidemiology: Methods and applications (pp. 223–239). Oxford: Oxford University Press.
Walter, S. D., & Birnie, S. E. (1991). Mapping mortality and morbidity patterns: An international comparison. International Journal of Epidemiology, 20, 678–689.
Wennberg, J. E., & Cooper, M. M. (Eds.). (1998). The Dartmouth Atlas of health care 1996. Chicago: American Hospital Association.
Wennberg, J., & Gittelsohn, A. (1973). Small area variations in health care delivery: A population-based health information system can guide planning and regulatory decision-making. Science, 182, 1102–1108.
Wurman, R. S. (2006). Access London (10th ed.). New York: Harper Collins.
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2017 Springer Science+Business Media B.V.
About this chapter
Cite this chapter
Waller, L.A. (2017). Mapping in Public Health. In: Brunn, S., Dodge, M. (eds) Mapping Across Academia. Springer, Dordrecht. https://doi.org/10.1007/978-94-024-1011-2_9
Download citation
DOI: https://doi.org/10.1007/978-94-024-1011-2_9
Published:
Publisher Name: Springer, Dordrecht
Print ISBN: 978-94-024-1009-9
Online ISBN: 978-94-024-1011-2
eBook Packages: Earth and Environmental ScienceEarth and Environmental Science (R0)