Not for the first time—and not for the last—a new disease appeared
to ignore the theories of the day. In the 18th century, everyone knew
that
yellow fever was a climatic illness: “Distempers of very hot
southerly countries, and natural to those climes”, wrote one authority
in 1755, were “unnatural to other countries situated in a northern
latitude”. And yet, during the 18th century yellow fever travelled up
the American coast to attack with increasing frequency cities like
Boston, New York,
and Philadelphia. Figuring out yellow fever was a matter not only of
medical urgency but also of economic survival. In one outbreak, in 1793,
Philadelphia lost 10% of its population. “Why should cities be
erected”, asked Noah Webster in 1796, “if they are only to be the tombs
of men?” Some believed this new fever was a plague-like contagion that
somehow travelled in the holds of cargo ships. Others insisted it was
generated locally in the foul, fetid airs of unsanitary cities built for
trade but not for cleanliness. If it was contagious then only
quarantine could tame it. If it was, however, a miasmatic illness then,
as sanitarian Benjamin Rush wrote in 1797, the medical response was
obvious: “Offal matters, especially those which are of a vegetable
nature, should be removed from the neighborhood.”
To test the theory of local miasmatic generation,
New York
physician Valentine Seaman etched on a copperplate map the street
location of patient deaths at the epicentre of an outbreak. On a second
plate he located the waste sites of “putrid effluvia” that were
suspected sources. Published in the inaugural issue of
The Medical Repository in 1798, the result revolutionised disease studies. Just as Andre Vesalius’
De Humani Corporis Fabrica challenged
16th-century physicians to see in dissection the real connections
between the skeleton and its musculature, Seaman’s mapping invited his
contemporaries to see the relation between the environment and a disease
generated in it. The
miasmatic theory became visible in the apparent correlation of disease incidence and proximate waste sites of odiferous, “furry miasmata”.
This is the unique thing about mapping: it takes a collection of
individual cases and makes of them a uniform event class. Each case is
related to all others—inviting measures of density and proximity—and to
other elements of the mapped environment. In this way medical maps
visualise Hippocrates’ great insight in On Airs Waters, and Places that
disease events are related, at one or another level, to the environment
in which patients live. Of course, the resulting map is only as good as
the theory it tests. In applying a miasmatic theory of local disease
generation to yellow fever, Seaman assumed it was spawned in the
odiferous urban air. In identifying waste sites to be mapped he
therefore identified the origin but not the source of the outbreak. He
observed dense populations of mosquitoes buzzing at waste sites—“never
before known, by the oldest inhabitants, to have been so numerous as at
this season”—but, lacking a theory of insect-born illness, Seaman missed
the true vector of transmission.
Mapping is also only as good as the technologies of its
presentation. Seaman wrote that he regretted the copperplate technology
would not permit more than a few of the many reported cases to be
imaged. Nor was it possible to overlay his two maps to make clear
comparisons. Quickly, however, these problems were overcome as the old
copperplate technology gave way to cheaper, faster more precise printing
methods that allowed for the inclusion of more data, and greater ease
of publication.
The mapping techniques developed for yellow fever were later used
in myriad attempts to understand the 19th century’s great recurring
pandemic, cholera. By mid-century it would become the most studied and
the most mapped epidemic in history. In England, the first pandemic
killed more than 50 000 citizens between 1831 and 1834, spawning a
generation of research papers. The study of this disease necessitated
better data and a bureaucracy to administer it. In answer, Parliament
created the
General Register Office
in 1836. Beginning in 1837, the map of the British nation was redrawn
into a series of registration districts and subdistricts—each supervised
by a
Medical Officer—to
permit ever more precise collection of mortality and morbidity reports.
Under the General Register Office, the first modern census of 1841
created a population database that would provide the population
denominator in the future study of disease at every scale. For the first
time, density analysis and mortality ratios could be argued across the
mapped surface on which individual cases became invisible—the specific
subsumed by the general. Still, specificity was important. In 1831
The Lancet published a “
Map
of the Progress of Cholera” occurring “through 700 irruptions” in 2000
towns to argue cholera’s independence from local environmental
conditions; it was a “poison which progresses independently” and thus
could not be stopped by quarantine or sanitary measures. From
New York City to
St Petersburg,
others mapped cholera as a local miasmatic illness, a contagion, or
both. In all these maps, the graphic argument was increasingly combined
with numerical and statistical analytics that were either produced in
the map or the commentary accompanying it.
Detail from Valentine Seaman’s map of cases of yellow fever in New York, published in The Medical Repository (1798)
Detail from The Lancet‘s map of the progress of the cholera in Asia, Europe, Africa published in 1831
Famously, John Snow argued that cholera might be spread by
contaminated water. He sought to prove his thesis in two studies, one of
a ferocious outbreak in
St James, Westminster, in 1854, and the other
across South London. In the first he created a non-statistical dot map
centred on a single water source: the Broad Street pump. In the second
he attempted to assign cases of cholera in south London to water supply
company jurisdictions in an attempt to show a correspondence between
water quality and disease intensity. Most of Snow’s contemporaries
agreed water seemed to be implicated, but that did not mean foul
miasmatic odours were irrelevant. Since the Broad Street map, like
Seaman’s map, was devoid of statistics or statistical arguments, it left
room for other interpretations. Worse, his more ambitious south London
study lacked—as Snow would later admit—a solid statistical base. Even
with one, as John Simon of the Board of Health concluded in 1856, the
maths of the day was incapable of a definitive analysis of the problem.
Robert Koch finally settled the question of cholera’s nature when,
applying the methods of his new bacteriology, he identified the
waterborne
Vibrio cholerae in 1883.
If bacteria were the cause of disease, mapping remained the means
by which its local source was most easily identified. In the late 19th
and 20th centuries, mapping became the tool public health experts used
to identify the probable source of a range of outbreaks including
cholera, typhoid, typhus, and yellow fever. The map also served to
identify new, non-bacterial epidemics. In the 1870s, Alfred Haviland
mapped cancer rates in Britain to argue a national epidemic. Within 20
years, maps of cancer mortality from across the British Empire were
being published in national health surveys and medical journals. Cancer
replaced cholera as the disease to study at every scale, from that of
the neighbourhood to the nation. New statistical methods were
developed—including confidence ratios, which allowed ever more precise
renderings of its prevalence. Once again, these maps served only as well
as the theories they embodied. US public health physician W H Frost,
for example, attempted to map influenza and poliomyelitis epidemics in
the early 20th century. These maps did not serve because, as we now
know, viral disease has a different profile from diseases that are
bacterially based. It was not until the computer revolution of the 1960s
that complex statistical models capable of describing—and thus
mapping—viral progression were possible. Finally, predictive models to
map viral progression could be constructed. In the 1980s, for example,
geographer Peter Gould married gravity and distance decay models with
data on HIV/AIDS incidence in the USA to create the first powerfully
predictive model of the disease. The resulting maps were disseminated as
a video. Still, the old dot map of incidence remained important; it was
through mapping the individual geographies of his patients that Abraham
Verghese came to see the dynamic of HIV as it affected the rural
Tennessee population he served.
By the first decade of the 21st century, maps of a range of new
conditions—from H1N1 influenza to West Nile virus—were presented as
statistical surfaces in which disease variances were embedded in
regional, national, and international geographies. Often presented on
the internet, these were typically maps of maps, aggregates of local
case studies collected by regional authorities that formed the basis of
national investigations. Irrespective of the nature of the disease
mapped, or the scale of its presentation, the old Hippocratic insight
remained: where cases aggregate its origin is likely to be found in the
human and geographical environment. To understand disease, therefore, we
map its community of incidence and its environment.
