Social Science & Medicine 69 (2009) 1246–1251
Contents lists available at ScienceDirect
Social Science & Medicine
journal homepage: www.elsevier.com/locate/socscimed
Short report
Crediting his critics’ concerns: Remaking John Snow’s map of Broad Street
cholera, 1854
Tom Koch*, Kenneth Denike
University British Columbia, Vancouver, BC, Canada
a r t i c l e i n f o
a b s t r a c t
Article history:
Available online 27 August 2009
Few cases in the history of epidemiology and public health are more famous than John Snow’s investigation of a neighborhood cholera outbreak in the St. James, Westminster, area of London in 1854. In this
study Snow is assumed to have proven that cholera was water rather than airborne through a methodology that became, and to a great extent remains, central to the science and social science of disease
studies. And yet, Snow’s work did not satisfy most of his contemporaries who considered his proof of
a solely waterborne cholera interesting but unconvincing. Uniquely, this paper asks whether the caution
of Snow’s contemporaries was reasonable, and secondly, whether Snow might have been more
convincing within the science of the day. The answers significantly alter our understanding of this
paradigmatic case. It does so in a manner offering insights both into the origins of nineteenth century
disease analysis and more generally, the relation of mapping in the investigation of an outbreak of
uncertain origin. The result has general relevancedpedagogically and practicallydin epidemiology,
medical geography, and public health.
Ó 2009 Elsevier Ltd. All rights reserved.
Keywords:
Cholera
Medical cartography
Medical history
Methodology
John snow
Introduction
Few cases in the history of epidemiology and public health are
more famous, or more cited, than John Snow’s investigation of
a neighborhood cholera outbreak in the St. James, Westminster,
area of London in 1854 (Snow, 1855a). Since William T. Sedgwick’s
(1901) textbook on sanitary science, Snow’s study of the ‘‘Broad
Street outbreak’’ has served as a foundational example in epidemiology, medical geography (Koch, 2005; Koch & Denike, 2004),
and public health (Vinten-Johansen et al., 2003). This interest is not
limited to health professionals. A spate of recent popular books
(Hempel, 2006; Johnson, 2006) and articles (Shapin, 2007) has
retold the story of the Broad Street study. The interesting question
therefore is not, ‘‘Who made John Snow a hero?’’ but perhaps, who
has not (Vandenbroucke, Elkman, & Beaukers, 1991). One answer
would be Snow’s contemporaries who, in the main, were unconvinced by his argument. The default assumption, challenged in this
paper, has been that Snow’s evidence was convincing and his
contemporaries should have credited his evidence, as do we today.
* Corresponding author. University British Columbia, Department of Geography
(Medical), 1984 West Mall, Vancouver, BC, Canada V6K 2S1. Tel.: þ1 647 351 0810;
fax: þ1 604 822 6150.
E-mail addresses: [email protected], [email protected] (T. Koch).
0277-9536/$ – see front matter Ó 2009 Elsevier Ltd. All rights reserved.
doi:10.1016/j.socscimed.2009.07.046
Here we ask whether Snow might have been more convincing
within the science of the day, and if so, how?
John Snow and cholera
In August 1854 began what Snow would describe as ‘‘the most
terrible outbreak of cholera which ever occurred in this kingdom’’
(Snow, 1855a: 38). For Snow, it was an opportunity to test his
theory of waterborne cholera, first articulated in an 1849 monograph (Snow, 1849). Naturally, Snow looked first for a water source
near the epicenter of the outbreak. ‘‘Within two hundred and fifty
yards of the spot where Cambridge Street joins Broad Street, there
were upwards of five hundred fatal attacks of cholera in ten
days.on proceeding to the spot I found that nearly all the deaths
had taken place within a short distance of the pump.’’
Snow applied for and received mortality reports from the
General Registrar Office (GRO) (Snow, 1855a: 39) to develop an
analysis reported in two separate venues, an expanded edition of
his 1849 monograph (MCC2) and a report to an inquiry committee
at St. Luke’s parish whose parishioners lived at the epicenter of the
outbreak (Snow, 1855b). Other researchers engaged in parallel
investigations that similarly relied on GRO mortality reports
included reports by London Sewer Commission engineer Edmund
Cooper (Cooper, 1854), St. Luke’s Parish curate Rev. Henry Whitehead (Whitehead, 1854) (who also contributed to the parish inquiry
T. Koch, K. Denike / Social Science & Medicine 69 (2009) 1246–1251
(Whitehead, 1855)), and members of the London Board of Health
(Board of Health, 1855).
Snow’s studies relied on two very different types of evidence.
First, he amassed a wealth of epidemiological data concerning the
relation between specific cholera deaths and decedent water usage,
much of it collected by collaborators more intimately connected with
the neighborhood than was Snow. Informants included a Greek
Street surgeon, Mr. Marshall, and the medical officer of health for
nearby Dean Street, Dr. J. Rogers. Dr. Fraser of Oakley Square provided
perhaps the most compelling case of the outbreak (Parks, 1855) in
reporting that the widow of a former percussion cap maker, who
died of cholera in the West End, ‘‘daily imported a large bottle of the
water from the pump in Broad Street’’ (Snow, 1855a: 38–39). In
addition, Rev. Henry Whitehead, who published the first monograph
on the outbreak and who had visited the home of every cholera
victim in his parish (Whitehead, 1854), shared his work with Snow
(Whitehead, 1865).
Second, Snow mapped the homes of cholera decedents in what
he described as ‘‘a topography of the outbreak’’. Mid-nineteenth
century researchers understood topographies as describing relationships between two event classes (Koch, 2005: 48). In a map of
local streets and landmarks Snow located 596 deaths reported to
the GRO from late August through September. To these Snow added
the location of public water pumps in the cholera study area. Snow
then sought to argue a relationship between the perceived dense
cluster of cholera deaths and one or more water sources at its
geographic center.
The combined use of individual case histories with a map
locating a class of reported deaths in relation to suspected disease
sources was a common methodology in nineteenth century disease
studies originating, at the latest, with Seaman’s 1796 study of a New
York City yellow fever outbreak (Koch, 2005: 23–33; Seaman, 1796).
From the first report of cholera in the garrisons of British soldiers in
India (Jameson, 1819) researchers in France, Germany, the United
Kingdom and the United States published articles on cholera with
maps of choleric incidence, typically in relation to suspected
sources of the disease (Koch, 2007). Simply, Snow’s mapped
evidence was a common analytic of his day.
‘‘It might be noticed,’’ Snow wrote in MCC2, ‘‘that the deaths are
most numerous near to the pump in Broad Street’’ (Snow, 1855a:
47). In a second version of the map prepared for the St. Luke’s
parish inquiry in 1855 Snow added an irregular polygon, presumably based on pedestrian distance, to distinguish those cholera
deaths nearer to the Broad Street pump than others in the study
area. Snow neither quantified the visual impression (for example,
‘‘two-thirds of all deaths occurring in the study area were within
the pump area’’) nor sought to develop a method by which
mortality in the Broad Street service area could be compared to
mortality in other pump catchments.
Snow’s critics
As Eyler has noted ‘‘Snow had not eliminated other explanations
or the role of coincidence’’ (Eyler, 2001: 226). It was not Snow’s
readers opposed his theory but that they objected to his dismissing
without detailed consideration other potential sources of contagion. Popularly and professionally two alternative explanations
were of special importance to Snow’s contemporaries. First, it was
believed that bad airs originating from a 1665 plague burial site on
which housing had been built and under which new sewers had
been laid might be a source of cholera contagion. Second, it was
believed the sewers themselves, as independent conduits of foul
odors generated by humane waste sites, might be complicit.
Snow dismissed both as possible sites of contagion, citing
a London Sewer Commission report in which engineer Edmund
1247
Cooper inventoried the existing sewer lines and gratings in a map
of 315 cholera deaths occurring between late August and midSeptember (Cooper, 1854). With no analytic but his impression of
the pattern of deaths evidenced in the map Cooper concluded there
was no relationship between the cholera incidence and local sewer
lines. In his map Cooper symbolized the seventeenth century
plague burial site as a relatively small oval area distant from the
outbreak’s epicenter. Unfortunately for Snow, Cooper’s map was in
error. The former plague burial site, on which nineteenth century
housing had been built, was far larger than Cooper has assumed
and extended to within a block of the Broad Street pump.
Snow surely knew this because of his then close personal and
working association with Rev. Henry Whitehead (Whitehead,
1865). Whitehead’s report for the parish inquiry, published with
Snow’s, included a map of 684 cholera deaths reported to the Board
of Health across the entire outbreak, each death locateddunlike
Snow’s but like Cooper’sd by house number on streets where all
houses were mapped. Whitehead’s map also included Cooper’s
detailed survey of sewers lines, with the year of their construction,
as well as both Cooper’s incorrect location of the old plague burial
site and its correct location in close proximity to the Broad Street
pump (Fig. 2).
Remaking Broad Street
Snow’s argument could have been strengthened using then
existing data lodged in the other contemporary maps. This would
have required at most several days of pedestrian labor. In 2008,
using only then existing maps and records, I did this work by hand
and then replicated the work in a GIS computerized mapping
program.
Snow could not count the number of deaths in the observed
cluster in the map prepared for MCC2 because its boundaries were
unclear. Did Carnaby, King, or Marshall Street define its western
boundary, for example? Where did the cluster begin and end to the
south? This problem was solved in Snow’s second map for the
parish inquiry in which the irregular polygon based on walking
distance was created a subset of all cholera deaths based on proximity to the Broad Street pump (Fig. 1). Surprisingly, Snow never
counted the number of deaths within this Broad Street pump
service area.
Within this service area were a total of 381 deaths mapped in
223 houses. In other words, two-thirds of Snow’s 596 mapped
deaths were in the Broad Street pump service area. Calculated
another way, over half of all houses in which cholera occurred were
sited in the Broad Street water service area. A conclusion expressed
in this way transforms Snow’s argument based upon a visual
impressiondwhat he saw in the mapped clusterdinto a more
forceful, numeric conclusion. While suggestive, the result was less
than meaningful without a population denominator capable of
translating general numeric incidence into a population mortality
figure.
Both Cooper’s and Whitehead’s maps contained the data
necessary to transform Snow’s raw mortality figures into then
common mortality ratios per 1000 persons. To demonstrate this I
counted the number of houses on each street segment in first
Cooper’s and then Whitehead’s maps (there was in this no difference between them), transferring the resulting sum in pencil to the
street segments in a photocopy of Snow’s second map. The sum of
all houses on all street segments in Snow’s Broad Street service area
created the denominator for a mortality ratio Snow elsewhere
employed, deaths per number of houses. The 1851 census recorded
an average of 10 persons per house in the registration sub-district
(Farr, 1852) Multiplying the number of houses by ten gave the
denominator of a rough population mortality ratio. While
1248
T. Koch, K. Denike / Social Science & Medicine 69 (2009) 1246–1251
Fig. 1. Rev. Henry Whitehead mapped almost 700 cholera deaths, sewer lines, and both the incorrect location of the old plague burial site (oval) and its correct size and location
a block from the Broad Street pump in this 1855 map.
imprecise, perhaps, by modern standards this would have been an
accepted technique in Snow’s day. The mortality in Snow’s Broad
Street area was 149.41 per 1000 persons: 381 deaths in 223 houses.
Because Snow argued mortality was greatest in the Broad Street
pump service area he needed, to be convincing, comparative
mortality ratios for adjacent pump catchments. Here was
a problem. Snow created only one water service area based on
pedestrian walking time and, given the difference between the
London of the mid-1850s and the modern, automobile-dominate
city, it would be difficult to attempt equivalent irregular polygons
based on pedestrian walking distance today. Street traffic and
access have changed too much. To demonstrate the potential for
comparative analysis I therefore created by hand a set of Thiessen
polygons centered on adjacent mapped water pumps, each
enclosing all mapped deaths nearer to a pump than all others. The
lines joining all service areas of a contiguous area are called
a Dirichlet tesselation after Snow’s contemporary, the nineteenth
century mathematician P. Lejeune Dirichlet who first described
their construction (Bailey & Gattrell, 1996: 156). The resulting
catchments serve here to demonstrate the effect of comparing
mortality based on population within different service areas in
a manner consistent with mid-nineteenth century mathematics
and science.
For demonstration purposes, only the five water service areas
with the highest raw mortality are included in Fig. 3. .936% of all
deaths mapped by Snow occurred in these catchments. Focusing on
these service areas avoided certain technical problems, for example
edge boundary concerns, and differences in the total study area of
Cooper’s and Whitehead’s maps which were, unlike Snow’s,
bounded by Regent and Oxford streets. The results for the principal
catchments in and around Broad Street are summarized visually in
Fig. 3. The results strongly supported Snow’s thesis with 28.71
deaths per thousand in the Rupert Service area, 25.88 deaths per
thousand in the Little Marlborough Street pump service area, and
12.34 deaths per 1000 person in the Warwick Street Pump service
area. Similarly, the number of deaths per house, a simple nineteenth century measure of intensity, dropped as one moved
outward from the Broad Street service area.
With this approach Snow also could have discounted the likelihood that other sites of potential contagion were the source of the
outbreak. To demonstrate this I first drew a 10 m buffer around the
burial site area mapped by Whitehead to allow for winds that some
believed carried miasmatic odors, mentioned in popular reports,
into neighboring streets. Separately, I counted deaths and houses
along streets served by sewer lines built after 1850 and implicated
in the popular literature. In this way I was able to calculate rough
mortality ratios for both the plague burial area and the sewers.
In the buffered burial site 113 cholera deaths occurred in 57
affected houses resulting in mortality per 1000 persons of 104.63.
And while the number of deaths was higher along streets across the
study area serviced by sewer lines built after 1850 (184 deaths) than
other streets, the population of those streets was much higher as
well (541 houses), yielding a mortality of only 34.01 per 1000
people. Deaths along sewer lines built after 1850 only in Snow’s
T. Koch, K. Denike / Social Science & Medicine 69 (2009) 1246–1251
1249
Fig. 2. Mortality per 1000 persons was calculated for the central water service catchments by taking the number of deaths per area, dividing by the number of houses on street
segments in each area, and multiplying by an estimated 10 persons per house.
irregular polygon included 139 cholera deaths in 73 houses among
a population of 2190 persons. The mortality reported was 63.47 per
1000 persons for that subset of post-1850 sewers suspected as
cholera conduits.
Calculating the combined effect of both the old plague burial site
and the post-1850 sewer lines raises a series of technical problems
in the analysis of overlapping but spatially non-commensurate
areas. These would have been beyond the science of Snow’s day. For
example, because all sewer lines, irrespective of age, were joined in
an integrated disposal system, analyzing the effect of only post1850 linesdor only those within Snow’s irregular polygondwould
have been relatively meaningless. And because the then new sewer
lines ran under both Broad Street, the old plague burial site and
elsewhere in the greater study area, correctly analyzing their effect
would have required a form of analysis not available to researchers
of the day. Simply averaging the cholera deaths per 1000 persons,
a likely recourse in that age, would not have returned an accurate
assessment of their effect.
Nor would this have been necessary. Fig. 3 presents the data
returned by the mapping for the major water service catchments in
the central study area considered by Cooper, Whitehead, and Snow.
To it we have added risk ratios that while not a common form of
1250
T. Koch, K. Denike / Social Science & Medicine 69 (2009) 1246–1251
Fig. 3. Calculating mortality in Broad street water service areas.
analysis in Snow’s day are too common in present epidemiology to
be excluded here. The 21 deaths reported in the Newman Street
service area have been excluded because neither Cooper’s nor
Snow’s maps provided street segment population data for the area.
Relative risk is therefore calculated for the other areas on the
assumption of a total 569 deaths within a population of 11,400.
Using either the then standard figure of mortality per 1000
persons or more modern relative risk ratios the result is clear.
Mortality, calculated as deaths per 1000 persons or as relative risk of
7.08 indicating how much greater the risk for the Broad Street pump
service population was compared to the remainder of the population. It was higher as well than mortality and risk of 1.95 in the
housing built on top of the old burial site, even with the buffer added
in this study, and far higher as well than along the post-1850 sewer
lines inventoried by Cooper for the London Sewer Commission.
Conclusion
The counting and analysis for these calculations took about four
days of on and off deskwork (it then took four days to do create the
files permitting the work to be done in the GIS computer mapping
program). We believe Snow’s contemporaries would have perceived
this comparative, numerical argument as persuasive, and certainly
more persuasive than the analytic approach employed by Snow.
Why Snow did not take an extra few days to do an analysis similar
to the one argued here? The simple answer is we do not know. Snow
left no record of his deliberations during this research period. One
reason was likely that Snow’s time was as a practicing anesthesiologist and physician during a period of fierce epidemic occurrence
Snow’s time was at a premium. In addition, he was simultaneously
engaged in the ambitious and time consuming study of cholera in
South London as well also reported in MCC2. Simply, And, reading
between the lines of Snow’s writings, we suspect Snow believed his
argument so utterly convincing he did not need to consider carefully
the alternate theories others put forward in separate studies.
Reading the reviews of MCC2 by critics like E.A. Parks (Parks, 1855),
and the studies of others like John Simon (1856), in this Snow was
clearly in error.
References
Bailey, T. C., & Gattrell, A. C. (1996). Interactive spatial data analysis. Essex, UK:
Longman Scientific & Technical Publishers. p. 156.
Cooper, E. (1854). Report on an enquiry and examination into the state of the drainage
of the homes situated in that part of the Parish of St. James, Westminster.
Metropolitan Commission of Sewers 478/21. London Metropolitan Archives.
Eyler, J. M. (2001). The changing assessments of John Snow’s and William Farr’s
cholera studies. Soz Praventivmed, 46, 225–232.
Farr, W. (1852). Report on the Mortality from Cholera in England, 1848–1849. London:
HMSO. (Also published as Farr, W. 1852. Registrar General’s Report on Cholera
in England 1849–1850. London: W. Clowes & Son).
General Board of Health (UK). (1855). Report of the committee for scientific inquiries
in relation to the cholera-epidemic of 1854. London: HMSO.
Hempel, S. (2006). The medical detective: John Snow and the mystery of cholera.
London: Granta Publications.
Jameson, J. (1819). Report on the epidemick cholera morbus, as it visited the territories
subject to the Presidency of Bengal, in the years 1817, 1818 and 1819. Drawn up by
order of the Government, under the superintendence of the Medical Board. E.
Balfour. Government Gazette Press.
Johnson, S. (2006). The ghost map: The story of London’s most terrifying epidemicdand
how it changed science, cities, and the modern world. NY: Riverhead Books.
Koch, T. (2005). Cartographies of disease: Maps, mapping and medicine. Redlands, CA:
ESRI Press. pp. 28–33.
Koch, T. (Nov 16 2007). Haynes lecture in geography: John Snow and his contemporaries, the Broad Street experience. College Park, TX: Texas A&M University.
Koch, T., & Denike, K. (2004). Medical mapping: the revolution indteaching and
usingdmaps for the analysis of medical issues. Journal of Geography, 103(2), 76–85.
Parks, E. A. (1855). Review XIII. Mode of communication of cholera, by John Snow,
M. D. The British and Foreign Medico-Chiurgical Review, 15, 449–463.
Seaman, V. (1796). An account of the epidemic yellow fever, as it appeared in the city of
New-York in the year 1795. Containing, besides its history, &c., the most probable
T. Koch, K. Denike / Social Science & Medicine 69 (2009) 1246–1251
means of preventing its return, and of avoiding it, in case it should again
become epidemic. New York: Hopkins, Webb & Co.
Sedgwick, W. T. (1901). Principles of sanitary science and the public health, with special
reference to the causation and prevention of infectious diseases. New York: Macmillan.
Shapin, S. (2007). Sick city. The New Yorker. http://www.newyorker.com/archive/
2006/11/06/061106crbo_books 6 November. 2006. Accessed 28.03.08.
Simon, J. (1856). Report on the last two cholera-epidemics of London as affected by the
consumption of impure water. London: George E. Eyre and William Spottiswoode
Printers.
Snow, J. (1849). On the mode of communication of cholera. London: Churchill.
Snow, J. (1855a). On the mode of transmission of cholera (2nd ed.). London: Churchill.
(Reproduced in W H. Frost (Ed.), Snow on cholera: A reprint of two papers by John
Snow, M.D., NY: The Commonwealth Fund, 38–56 (1936)).
1251
Snow, J. (1855b). Dr. Snow’s Report. In Parish, St. James. (Ed.), Report of the
cholera outbreak in the Parish of St. James, Westminster, during the autumn of
1854 (pp. 97–120). London: J. Churchill.
Vandenbroucke, J., Eelkman, H., & Beukers, H. (1991). Who made John Snow a hero?
American Journal of Epidemiology, 233, 967–973.
Vinten-Johansen, P., Brody, H., Paneth, N., et al. (2003). Cholera, chloroform, and the
science of medicine: A life of John Snow. NY and Oxford: University of Oxford Press.
Whitehead, H. (1854). The cholera in Berwick Street. London: Hope & Co.
Whitehead, H. (1855). Mr. Whitehead’s report. In Parish, St. James. (Ed.), Report of
the cholera outbreak in the Parish of St. James, Westminster, during the autumn
of 1854 (pp. 120–167). London: J. Churchill.
Whitehead, W. (1865). The Broad Street pump: an episode in the cholera epidemic
of 1854. Macmillan’s Magazine 113–122.