A spatial analysis of the smallpox epidemic in Sheffield, United Kingdom, 1887-1888

A spatial analysis of the smallpox epidemic in

Sheffield, United Kingdom, 1887-1888

CHAPTER 1: INTRODUCTION AND LITERATURE REVIEW

The data and methodology make this project unique. To date no

researcher has examined Barry’s Report on an epidemic of smallpox at

Sheffield during 1887-8, which contains complete demographic information

for all of the 590 deaths that occurred. The first step in analyzing the data set

was to create a geographic information system (GIS) and analyze the

resulting information on the epidemic. This paper details the research done

on Barry’s text.

Barry noted the unique nature of epidemic diseases and the necessity

of understanding that nature for the protection of a city’s inhabitants.

…no statistics can estimate the amount of

suffering caused by so wide-spread an epidemic. In

many houses, especially those inhabited by the

unvaccinated, where death has spared the inmates, a

legacy of blindness, or permanent disfigurement, or

weakened health and impaired usefulness, has been

left behind –of such losses no account can be

rendered.

(Barry 1889, 294).

The horrors associated with the epidemics of the past can still generate fear.

The threat is magnified when one realizes how incredibly vulnerable modern

populations are to the possibility of such destructive diseases.

Chapter 1 will review the literature on geography and disease, spatial

analyses of medical data, and studies of smallpox epidemics. Chapter 2 will

discuss the data source, Barry’s text, in detail as well as problems with

historical medical data. Chapter 3 will discuss the creation of the GIS, the

spatial analyses used on the data, and the result of those analyses. Chapter 4

will compare the results of this study to a study done by P.H. Curson on the

1881-1882 smallpox in Sydney, Australia, including comparisons of the age

structures of the epidemics and cluster formations. Chapter 5 will discuss the

applicability of this study to various academic and practical fields and

further analyses that could be done on Barry’s text.

Literature Review

Many areas of research coincide in the area of medical geography,

especially when discussing infectious disease diffusion through urban

populations. The exploration of the 1887-1888 smallpox epidemic in

Sheffield, Yorkshire, United Kingdom highlights the importance of those

research areas. The areas of relevant study include geographic theory on

disease diffusion (Smallman-Raynor and Cliff 2001; Haggett 2000; Becker,

et al. 1998; Pyle and Rees 1971; Hagerstrand 1969), spatial analyses of

medical data (Edsall 2003; Haggett 2000; Barrett 2000; Cliff, Haggett, and

Smallman-Raynor 1998; Curson 1985; Morrill and Angulo, 1981, 1979;

Angulo, et al. 1980a; Angulo, et al. 1980b), and studies of smallpox

epidemics in various disciplines (Carrell 2003; Preston 2002;Tucker 2001;

Fenn 2001; Banthia and Dyson 1999; Craddock 1995). These three areas of

study will be addressed in this chapter.

Geography and Disease Diffusion

Infected spaces and populations and the changes in those spaces and

populations have been the concern of researchers since the earliest days of

civilization. Hippocrates, in his Of Airs, Waters, and Places, noted that

disease and regions of the globe often shared close ties (Brachman 2003;

Bewell 1996). Suspicion of far-away places ran rampant among Asians and

Europeans through the Dark Ages and Middle Ages because of plague

epidemics (Wills 1996). In 1377, the citizens of Raguna, Italy invented

quarantine, a forty-day isolation period, to restrict the spread of plague

(Meade and Earickson 2000). This was because certain areas of the world

were known to harbor plague and the ships coming from them often carried

the disease. Before the advent of germ theory, disease was believed to be

caused by pathogenic places. Many scientists of the day were convinced that

most diseases, especially infectious diseases, were environmental in origin;

hence, the naming of malaria, which literally translates as “bad air.” (Bewell

1996) Scientists of this period between the beginning of the Scientific

Revolution and the advent of germ theory were concerned with place

because their observations seemed to indicate that places made people sick

with certain diseases. Leeuwenhoek’s invention of the microscope and

subsequent discovery of microbes created a basis for theories on disease

causation (Brachman 2003). Eventually, Robert Koch’s announcement of the

germ theory of disease in the late nineteenth century caused a radical change

in medicine and the study of disease in all disciplines. However, the idea of

pathogenic places remained prominent in popular thinking for many decades

(Meade and Earickson 2000). Still today, medical geography and disease

mapping tries to determine the relationship between environmental factors

and disease patterns, even when investigating infectious diseases (Jones and

Moon 1991)

In 1792, L.L. Finke created the first cartographic images of the world

divided by disease. The map he made showed areas of the world according

to the diseases that were endemic and epidemic to each region (Barrett 2000;

Bewell 1996). Valentine Seaman created a local point map of a New York

yellow fever epidemic in 1798 (Barrett 2000). Between 1820 and 1836 over

36 scholars produced maps of cholera mortality and morbidity distributions.

Frederich Schurrer in 1827 created a world map of yellow fever, cholera and

plague distributions. Heinrich Barkhaus’s Physical Atlas from 1848 details

the diseases that affected humans around the world (Barrett 2000). Yellow

fever, the scourge of the Western tropics and subtropics, was feverently

tracked as it swept from port to port each year throughout the eighteenth and

nineteenth centuries. The cartographic progress of yellow fever often

determined the economic and medical health of most cities in the Western

hemisphere throughout the nineteenth century (Barrett 2000; Meade and

Earickson 2000; Wills 1996; Carrigan 1994). In the late twentieth and early

twenty-first century, the tradition continues with cartographic techniques

being used to analyze the various infection rates and demographics of the

HIV/AIDS epidemic across the world from the African savanna to the

American heartland (Meade and Earickson 2000; Smallman-Raynor 1995;

Lam and Liu 1994). Maps and their creation have been an integral part of

human understanding of disease (Haggett 1992).

All infectious diseases spread from one organism to another. The

agent of the disease encounters a host, multiplies, and spreads to other

potential hosts (Cliff, Haggett, and Smallman-Raynor 1998). Scientists

describe the manner in which a disease spreads as diffusion, the expansion of

infection from a starting point or center (Haggett 2000; Meade and

Earickson 2000; Pyle and Rees 1971; Hunter and Young 1971). Diffusion

theory has evolved into an eminently scientific explanation for the spread of

disease. Many variables can affect the speed at which a disease infects an

area, and its population. Nevertheless, at its heart, “diffusion is patterned by

the configuration of the networks that encourage movement and barriers that

discourage it” (Meade and Earickson 2000, 163).

Geographers have documented two major types of diffusion:

relocation diffusion and expansion diffusion (Haggett 2000). There are also

several subtypes of diffusion, but only a few will be discussed here.

Relocation diffusion occurs when the agent jumps a great distance, usually

due to host migration. An example of this type of diffusion was the

importation of smallpox from Europe to the Native American populations in

the fifteenth and sixteenth centuries (Haggett 2000; Meade and Earickson

2000). Expansion diffusion has several subtypes including hierarchical

diffusion, cascade diffusion, and contagious or contact diffusion.

Hierarchical diffusion allows an agent to move through an area in an ordered

sequence. For example, cholera epidemics will usually move from the

original source to locations farther down a river’s course because it is a

water-borne disease. Cascade diffusion is a specific type of hierarchical

diffusion in which the agent moves only from large urban areas to

progressively smaller areas. The expansion of HIV through Europe and the

United States is an ideal example, as the disease was imported first to first-

tier urban centers, like New York, then it diffused out through second- and

third-tier centers. Contact or contagious diffusion, the type with which this

research is most concerned, depends on direct contact between hosts and is

strongly affected by physical proximity. Smallpox epidemics are an

excellent example because smallpox requires close contact for transmission

(Haggett 2000; Meade and Earickson 2000).

In general, three realms shape the patterns of disease. These factors,

environmental influences, social influences, and epidemiological influences,

interact to create various disease patterns in a population. Environmental

influences of the modern era are not the cause of the disease but they present

an opportunity for contact between an agent and a host (Gatrell 2002; Pyle &

Reese 1971). For example, it is unlikely that a traveler to Moscow would

acquire yellow fever there because the climate in Moscow is hostile to the

mosquito that transmits yellow fever (Carrigan 1994). Social influences,

such as income, education, and class, also are not direct causes of a disease

but present opportunities for the agent to invade the host. Lower income,

education, and class all affect the amount and quality of nutrition and health

care an individual is likely to receive (Gatrell 2002; Pyle & Reese 1971).

Epidemiological factors include infectious agents, genetic conditions, and

other physical factors. These are often cited as the primary cause of disease.

Some diseases, like cholera, measles, and smallpox, rely on interaction

between all three spheres of influence; other diseases, like cancer or mercury

poisoning, rely on only one realm to wreak havoc on populations (Gatrell

2002; Pyle & Reese 1971).

Epidemics are a singular phenomenon in the world of medical

geography and much debate has occurred over the exact definition of an

epidemic. The word was originally used in Hippocratic texts from 500 B.C.

and comes from the Greek words epi meaning ‘upon’ and demos meaning

‘people’ (Haggett 2000, 10; Cliff, Haggett, and Smallman-Raynor 1998, 3).

The modern definition of the term “epidemic,” an unusually high incidence

of disease in a particular place and time, was not used until about 1603

(Haggett 2000; Cliff, Haggett, and Smallman-Raynor 1998). In Benenson

(1990, 499), the most complete and modern definition of an epidemic to date

is:

The occurrence in a community or region of cases

of an illness (or an outbreak) clearly in excess of

expectancy. The number of cases indicating

presence of an epidemic will vary according to the

infectious agent, size, and type of population

exposed, previous experience or lack of exposure

to the diseases, and time and place of occurrence;

epidemicity is thus relative to usual frequency of

disease in the same area, among the specified

population, at the same season of the year (cited in

Haggett 2000, 11).

This definition is inherently spatial in its discussion of epidemics and their

occurrences. This leads to the spatial analysis of disease and using statistical

information of disease outbreaks to attempt to catalog epidemics and the

factors that cause them (Haggett 2000).

Spatial Analysis of Disease

Spatial patterns on maps are directly linked to where they are located

on the earth’s surface. “The geographical location of any area on a map is

unique to it as its locational relationship to other areas, and it is the

interactions between areas which create the distinctive patterns of disease

incidence frequently seen on maps” (Cliff and Haggett 1988, 33). This is

why research into new spatial analysis applications, technologies, and

methods is vital to advances in medical geography and epidemiology (Edsall

2003; Flint 2003; Lobben 2003; Theophilides, et al. 2003; Bengtsson and

Lindstrom 2003; Grais, Ellis, and Glass 2003; Eichner 2003; Eichner and

Dietz 2003; Kreiger 2003; Cromley

2003; Brachman 2003; Ricketts 2003; Kistemann. Dangendorf, and

Schweikart 2002; Smallman-Raynor and Cliff 2001; Borroto and Martinez-

Piedra 2000; Haggett 2000).

Spatial analysis of disease and geography take many forms. GIS, point

pattern analysis, regression analysis, cartographic animation and

manipulation are all examples of modern spatial analysis techniques. There

are four different scales of mapping, referred to as nominal, ordinal, interval,

and ratio scales. Nominal scale data is merely classification level data, such

as male or female. Ordinal scale data is set on a ranking scale. Interval scale

data is set on a continuum with no natural origin; while ratio scale data, the

most common in mapping, is set on a continuum with a natural origin, like

zero (Cliff and Haggett 1988). Maps generally display three types of

information. The first map type, called a real-valued map, shows where the

value of a variable is mapped. The second map type displays point data, the

location of a point with no other information added. The third map type

traces networks, the connections between points or areas that have some

common factor. These maps, also called flow maps, are frequently used

when summarizing patterns from records built from newspapers and death

records (Cliff and Haggett 1988).

GIS has become an integral part of spatial analyses in medical

geography. Most of the research done in the last thirty years has used GIS to

explore data or explored the uses of GIS methods and technology. The early

years of GIS research were restricted to simple chloropleth maps, point

maps, and flow maps that illustrated the numbers and directions of epidemic

movements (Mielke, et al. 1984; Morrill and Angulo 1981, 1979; Angulo, et

al. 1980a; Angulo, et al. 1980b; Pyle 1973; Hunter and Young 1971; Pyle

and Rees 1971). However, most GIS research did not occur until after the

mid-1980’s when the increase in personal computing power and increase in

GIS programs written made experimentation easier (Ricketts 2003). After

that, much spatial research exploring various diseases, both infectious and

non-infectious, was done using GIS (Edsall 2003; Flint 2003; Lobben 2003;

Smallman-Raynor and Cliff 2001; Haggett 2000; Meade and Earickson

2000; Becker, et al. 1998; Cliff, Haggett, and Smallman-Raynor 1998; Cliff

and Ord 1995; Gatrell 1995; Smallman-Raynor 1995; Haggett 1993; Wilson

1993; Cliff and Haggett 1988). Many other health-related fields are also

exploring the uses and applications of GIS in examining their data. For

example, epidemiologists and public health officials have published the

majority of the research on the feasibility of using GIS to analyze health data

(Theophilides, et al. 2003; Bengtsson and Lindstrom 2003; Grais, Ellis, and

Glass 2003; Eichner 2003; Eichner and Dietz 2003; Krieger 2003; Cromley

2003; Brachman 2003; Ricketts 2003; Kistemann, Dangendorf, and

Schweikart 2002; Borroto and Martinez-Piedra 2000).

Infectious diseases have distinct patterns when the locations of their

occurrences are mapped and analyzed. Two divisions occur when discussing

infectious diseases: those that are spread by real contagions, or propagated

diseases, and those that are spread by apparent contagions, or common-

vehicle diseases. Real contagions are diseases in which contact is directly

responsible for passing disease to others, spawning new generations of

disease (Haggett 2000; Cliff and Haggett 1988). This disease type results in

a pattern described as “a cluster centre (the parent), surrounded by offspring

whose geographical density will decline with increasing distance from the

parent.” (Cliff and Haggett 1988, 57) Apparent contagions are diseases

where contact is not directly responsible for transmission; these diseases are

also called vector-borne diseases (Haggett 2000; Cliff and Haggett 1988).

This disease type results in “geographically restricted areas of high and low

levels of deaths,” reflecting the presence of the vector (Cliff and Haggett

1988, 57).

A geographic study of the 1970-1971 measles epidemic in Akron,

Ohio revealed similar patterns and pathology to smallpox epidemics.

Measles and smallpox are similar infectious diseases; both can be eliminated

by rigorous vaccination, mostly affect children, and confer lifelong

immunity upon survival of an attack. This study examined the morbidity

and mortality of measles by mapping cases by census tract (Pyle 1973).

After visually examining the patterns presented in the maps, Pyle uses

correlation analysis to examine the statistical relationships between various

socio-economic factors and incidence of measles in this epidemic. He found

that census tracts with low incomes and low educations correlated to higher

incidences of measles. He speculates that these correlates were due to poor

quality vaccination material available and lower percentages of children

vaccinated in the census tracts (Pyle 1973).

Haggett (1992) explores possible applications of Sauer’s historical-

cultural geography concepts to the diffusion of disease and medical

geography. The Sauer methods of using deductive locational principles and

using maps as hypotheses were evaluated for their utility in medical

geography, specifically in tracing the origins of diseases (Haggett 1992). The

use of maps to present theories on the various diffusion patterns of disease is

a method used by many geographers in various situations (Mortimer 2003;

Mortimer and McVail 2002; Williams 1994; Haggett 1992; Angulo, et al.

1980a,

1980b; Morrill and Angulo 1981, 1979; Thomas 1974; Hunter and Young

1971; Pyle 1971; Pyle and Reese 1971).

Lobben (2003) gives an overview of the world of cartographic

animation and a classification system for types of cartographic animations.

She defines four major categories of cartographic animation: time-series,

areal, thematic, and process. The first category, time-series animation, shows

the change of a spatial element of a variable over a period of time (Lobben

2003). The generally acknowledged components in a timeseries are trend,

seasonality, cyclical fluctuations, and residual/random patterns. Smallpox

epidemics would provide good data for such animations because they have a

seasonal pattern (Banthia and Dyson 1999; Cliff and Haggett 1988).

Smallpox favored periods of low humidity and low temperature, and

consequently, the United Kingdom’s greatest numbers of smallpox cases

tended to occur in winter (Cliff and Haggett 1988).

The second category, areal animation, shows a change in location at one

point in time with a static variable. The third category, thematic animation,

holds the spatial component steady while the time and variables fluctuate.

The fourth category, process animation, has all three elements, space, time,

and variable, in flux. Lobben (2003) has also designed a decision-tree

diagram to aid readers in classifying projects. This development in

cartographic animation seems to present a new manifestation of Sauer’s

“maps as hypothesis” concept (Haggett 1992).

Smallpox: Pathology and Study

Smallpox, caused by the virus Variola major, was the cause of death

and disfigurement among the human population since the beginning of

civilization (Brachman 2003; Tucker 2001; Fenn 2001; Henderson 1978;

Mack, et al. 1970). Smallpox belongs to the Orthopoxvirus genus, which

also includes monkeypox, cowpox, and variola minor (Tucker 2001; Cliff

and Haggett 1988). Virus-laden droplets spread from the nose, throat, and

skin of infected people to susceptible people which formed the primary

transmission path for smallpox. Also, smallpox could be spread by airborne

virus particles and contaminated inanimate objects, like clothing and

bedding (Tucker 2001; Fenn 2001; Cliff and Haggett 1988; Henderson 1978;

Mack, et al. 1970). When a susceptible person was infected, the disease

would incubate for 11 to 14 days. Then, for 1 to 3 days, the patient was

attacked by aches, fever, and other flu-like symptoms. As the flu-like

symptoms eased, the patient developed a rash that matured into pus-filled

blisters. The blisters lasted about two weeks; most patients who died of the

disease, did so during this stage. After 12 to 14 days, the blisters formed

scabs and the patient recovered.

However, the patient remained contagious until the last scabs fell off

because the scabs carried live virus particles (Tucker 2001; Fenn 2001;

Henderson 1978). Unvaccinated individuals suffer case fatality of 15-25%,

up to 40-50% among infants and elderly. Those who did survive were

frequently disfigured and blinded. In 1797, Edward Jenner invented a

vaccination method for smallpox and protection from smallpox was possible

(Mortimer 2003; Mortimer and McVail 2002; Tucker 2001; Williams 1994;

Flinn 1981;

Angulo, et al. 1980a, 1980b; Morrill and Angulo 1981, 1979; Henderson

1978; Thomas 1974; Trobridge 1897; Barry 1889; Davidson 1889;

Hainsworth 1889).

Various academic disciplines have produced longitudinal studies of

smallpox in Scandinavian populations throughout the seventeenth,

eighteenth, and nineteenth centuries. The impeccable and thorough records

kept by many parish administrators and ministers detailed major life events,

including causes of each death (Bengtsson and Lindstrom 2003; Wilson

1994; Mielke, et al. 1984). Wilson (1993) addresses the utility of modeling

smallpox epidemics to predict the pattern of an epidemic at the meso-scale.

His study of an epidemic of smallpox in one parish in Finland used church

records of births, marriages, and deaths, supplemented with local census

records to document and develop a model for contagious disease

transmission. He discusses an interesting aspect of early vaccination records,

called Communion Books that were used by the Church to record vaccination

histories for everyone in the community who had received their first

communion (Wilson 1993). He also cites a 15% mortality rate from

smallpox in the eighteenth and nineteenth centuries, which is significantly

lower than most epidemics (Wilson 1993; Flinn 1981).

Another study using Finnish records tracked the epidemiology of

smallpox in sixteen administrative units for 140 years. This demographic

study tracked the fluctuation of smallpox in this area using detailed parish

records that required entry of a cause of death. Over the study period, which

extended from 1751 to 1890, only three significant trends were noted. The

first was the seven-year periodicity of smallpox outbreaks before the advent

of vaccination with mostly children suffering from the disease. The second

was the significant drop in smallpox morbidity and mortality after 1805, the

year vaccination was introduced to Finland. The third was the change in

periodicity of smallpox outbreaks to an eight-year period, where both adults

and children were affected (Mielke, et al. 1984).

A third study, located in Sweden, was more general in scope,

addressing the correlation between the incidence of airborne infectious

diseases among infants and the likelihood of mortality from airborne

infectious diseases as an adult. Bengtsson and Lindstrom (2003) examined

records from four rural parishes in southern Sweden from 1766-1894. The

data examined for smallpox showed that infants, defined as children less

than one year old, who survived smallpox infections did not suffer different

mortality rates of airborne infectious diseases as adults than other infants.

However, none of the infants that survived smallpox, which was unlikely,

later died of smallpox since one smallpox infection grants lifelong immunity

(Bengtsson and Lindstrom 2003).

Conclusion

This chapter examined geographic theory on disease diffusion, spatial

analyses of medical data, and studies of smallpox epidemics in various

disciplines. Only superficial consideration has been given to these three

areas, as each area could generate a thesis in itself. These areas of medical

geography are especially important when discussing infectious disease

diffusion through urban populations. The exploration of the 1887-1888

smallpox epidemic in Sheffield, Yorkshire, United Kingdom will clarify the

importance of those areas. The next section will examine the primary source

document written by Frederick Barry that detailed the events of the Sheffield

epidemic.

CHAPTER 2: BARRY’S REPORT ON AN EPIDEMIC OF

SMALLPOX AT SHEFFIELD, DURING 1887-8

Introduction

The original source for the case data in this study was Dr. Barry’s

Report on an epidemic of small-pox at Sheffield, during 1887-8. This text is

rare, as only ten libraries in the world own copies and an unknown number

are in private hands (OCLC: First Search 2003). The copy used in this study

is the only copy known to exist in a private collection.

Despite an extensive literature search, there appears to be no modern study

of the

Sheffield smallpox epidemic of 1887-8 on record. Only a few passing

references to Barry’s text seem to exist in modern studies of disease

patterns, most of which are references by researchers studying smallpox

epidemics in Brazil in 1954 (Angulo, et al. 1980a; Angulo, et al. 1980b;

Morrill and Angulo 1979). These and other studies that mention Barry’s text

or include it on their reference lists only refer to statistics calculated in the

text as benchmarks for that period or as an example of a method for

recording epidemics (Mortimer 2003; Mortimer and McVail 2002; Williams

1994; Thomas 1974). No researcher has completed an intense study of the

1887-1888 Sheffield smallpox epidemic since Barry completed his record

and his contemporaries commented on it (Davidson 1889; Hainsworth

1889).

The British government commissioned Barry to investigate the

epidemic soon after it began. The city of Sheffield was home to a population

of approximately 316,000, 87% of whom were vaccinated against smallpox.

However, over 6,000 of Sheffield’s inhabitants fell ill and nearly 600 died.

His mission was to document and analyze the epidemic using various maps,

charts, and statistics. He reported his observations on possible causes for the

epidemic and reasons for its severity. In his Report, Barry recorded the

demographics of all 590 smallpox deaths in the Sheffield area between April

1887 and March 1888. All the death records consist of the district, victim’s

name, the date of death, age, residential address, occupation, vaccination

status, and identity of the person verifying this information. Barry and his

team of assistants, who were mostly local doctors and public health officials,

collected and catalogued this information. The creation of the geographic

information system (GIS) and the spatial analyses of the data will be

discussed in detail in Chapter 3.

The City of Sheffield

Sheffield is an ancient city situated 150 miles north of London in Yorkshire

(See

Figure 2.1). Originally founded by the Saxons sometime during the ninth

century, Sheffield has long been a center of metalworking and is often seen

as the birthplace of the modern steel industry. From 1750 to the First World

War, Sheffield was home to many of the innovations that revolutionized the

modern steel industry, including the cast steel process, the crucible steel

process, and the Britannia metal process (Tweedale 1987). Through the mid-

nineteenth century, Sheffield reportedly “produced 90% of all British steel

and 50% of all European” (Tweedale 1987, 2).

The city of Sheffield was home to many skilled craftspeople and trade

unions. Most of these jobs were well paid but very dangerous. Over-crowded

and unsanitary living conditions were the largest of the city’s problems

during the nineteenth century due to the rapid population growth (Tweedale

1987). Sheffield’s population expanded from 31,000 in 1800 to 135,000 in

1850 to 316,000 at the time of the epidemic (Tweedale 1987; Barry 1889).

Today, Sheffield is England’s fourth largest city and still home to steel

production and skilled metalworking. The 2002 census reports its population

as 512,200 with tourism, sports, high tech manufacturing, and the largest

shopping center in Europe now supporting the economy (Atherton 2003;

Sheffield City Council 2003).

The Text

Barry’s text contains four major sections plus an introduction by the Chief

Medical Officer, Dr. Buchanan. These four sections give detailed

information on all aspects of the district surrounding Sheffield, the

epidemiological history of Sheffield in the nineteenth century, the path of the

1887-1888 epidemic, and the reactions of local authorities to the epidemic.

The introduction by Buchanan presents a few statistics, both demographic

and epidemic, in addition to possible causes behind the extent of the

epidemic. Buchanan gives the total population as 316,288; and the

vaccinated population as 275,878, 87.2% of the total population. The

epidemic’s raw numbers included 6,088 cases and 590 deaths from

smallpox, 9.7% of the total cases. Buchanan claims that smallpox spread

through Sheffield despite the efforts of the Sanitary Authority because those

suffering mild cases did not feel the need for medical help and, therefore, did

not report their illness to doctors. Cases that were not reported to doctors

were not reported to government agencies. Also, several doctors did not

comply with the mandatory notification program established by the Sanitary

Authority despite a reward program for reporting (Barry 1889). No reason

was given for this refusal to report, though the doctor/patient privilege has

been cited by other sources as the reason for non-compliance (Williams

1996; Barry 1889). According to the local authorities, crowding was the only

“sanitary circumstance” that affected the differences in smallpox incidence

in the city (Barry 1889). During this period of England’s history, the

crowding that exacerbated many epidemics was due to poverty. The lower

classes were much more likely to live in tenements and other poorly

maintained housing (Mooney 1999; Williams 1996; Craddock 1995;

Williams 1994; Curson 1985; Hunter and Young 1971). However, Buchanan

maintained that no single factor affected infection rates except vaccination

(Barry 1889).

The first section, titled “Description of District,” details the

background information of Sheffield. It begins with the general topography

and geology of the district of Sheffield, with specific accounts of landscape

types and climate. The Borough of Sheffield at the time of the epidemic was

divided into six townships; three in the

Sheffield District, Sheffield, Attercliffe, and Brightside, and three in the

EcclesallBierlow District, Ecclesall-Bierlow, Nether Hallam, and Upper

Hallam. The townships also served as sub-districts for the purposes of this

text, except in the case of Sheffield, which was further divided into Sheffield

North, Sheffield West, Sheffield East, and Sheffield South Sub-district. This

made a total of nine areas of study in Barry’s text.

Barry described the general demographics of the population, citing a

population of about 316,288 people in 1887 with a population density

averaging 14.5 people per acre. The main industry of Sheffield at the time

was steel manufacturing and various high quality metal works. Barry even

describes the gender divisions in labor. Generally, men were involved in the

mining and refining of iron and the manufacture of metal goods and women

were working either in the home or in factories doing polishing and file

cutting work (Barry 1889).

The second section, titled “Origin and Progress of the Smallpox

Epidemic in the Borough,” describes the history of smallpox in the area from

1872 to the 1887-1888 epidemic. There were less than five deaths per year

from 1872 to 1884. From that point on, a steady increase in morbidity and

mortality was evident through the 1887-1888 epidemic. The Sanitary

Authority reported 34 deaths in 1884 and 67 deaths in 1885, which

threatened to develop into an epidemic, but it was controlled. In 1886, three

small outbreaks across the city were noted and controlled. The last cases of

the 1886 outbreaks were believed to be the origin of the 1887-1888

epidemic. Barry tracks two possible points of origin both beginning in

March 1887 and eventually expanding to consume the city. The first possible

source, referred to as Outbreak A, occurred the third week of March with a

man living in Sheffield Park Sub-District near a house that was infected in

1886. He worked in Brightside Sub-District, and carried the disease between

the two, spreading it to co-workers and family members. The second

possible source, referred to as Outbreak B, began in early March with a man

living in Nether Hallam Sub-District who probably contracted the disease

from smallpox particles caught in a second-hand waistcoat. He spread the

disease into his neighborhood, which was spread farther by a man who

worked as a trolley conductor in Sheffield. After cases began appearing

across the city, sometime in August and September, many people were

infected as patients or visitors at the Borough Hospital (Barry 1889).

The third section, titled “Action of Local Authorities in Reference to

the Epidemic,” describes and summarizes the actions of the Sheffield Town

Council, who administered the Public Health Acts. It also summarizes the

actions of the Sheffield and Ecclesall Bierlow Boards of Guardians, who

administered the Vaccination Acts. The Town Council met to address the

burgeoning epidemic on 9 June 1887, with 40 cases reported, and 1 August

1887, with 160 cases and 7 deaths reported. At these meetings, they decided

to establish a public disinfection station to reduce infection risks. However,

this disinfection station was not set up until June 1888, several months after

the end of the epidemic. On 10 November 1887 the Town Council voted to

print and post 2,000 placards “recommending revaccination, and advocating

other precautions against infection from small-pox” (Barry 1889, 8).

Decisions were made at subsequent meetings from December 1887 to

February 1888 reinforcing the earlier recommendation for revaccination. In

addition, at the meeting held on 26 January 1888, the Town Council

requested that the Local Government Board advise Parliament of the need

for compulsory notification of infectious diseases (Barry 1889).

The fourth section, titled “Circumstances influencing the spread of

smallpox in Sheffield,” contains many maps, tables, and diagrams of

population and epidemiological data. Here, Barry considers four major

factors in the epidemic: vaccination; general sanitary circumstances;

hospitals; and various influences on disease communication between the

infected and the uninfected (Barry 1889). Thirty-nine maps were created by

the author and drawn for publication by Dangerfield Lithographers of

Covent Garden. There are three divisions among the maps. The first map,

“Map 1: Showing the position of houses first invaded by small pox in the

epidemic of 1887” (Figure 2.2) presents a view of the whole of Sheffield

with Outbreak A marked in red and Outbreak B shown in green (Barry,

1889, 4). The second set of maps, Maps 2 through 26 are titled ”Map of a

Portion of the Borough of Sheffield” (Figure 2.3) and shows the progress of

the epidemic from January 1887 to 17 March 1888, with bimonthly updates

from 4 June 1887 to 17 March 1888 (Barry, 1889, 276). The third division of

maps, Maps 27 through 39, are titled, “Plan Shewing (sic) District 1000 Ft.

Round Winter Street Hospital Sheffield” depicting cases that occurred near

the Winter Street Hospital, the designated smallpox hospital for the borough

at different intervals between April 1887 and March 1888 (Figure 2.4)

(Barry 1889, 280).

Barry calculated many statistics on the population affected by the

epidemic, and nearly all of it was catalogued in the 156 tables in the text.

Most of the data catalogued were the mortality and morbidity returns for

each sub-district. The mortality tables (See Figure 2.5), contain detailed

demographic information including victim’s name, the date of death, age,

residential address, occupation, vaccination status, and identity of the person

verifying this information. Also, included were the vaccination information

of large sections of the population, such as the overall vaccination and

revaccination rates by sub-district. Several are statistical analyses of the

various data collected, including the differences in death rates for various

age groups and vaccinated and unvaccinated persons. He also included

diagrams showing graphs of death rates for smallpox, measles, scarlatina,

diphtheria, whooping cough, fever, and diarrhea from 1861 to 1887. His

stated purpose for including this information was the examination of changes

in infectious disease mortality (Barry 1889). He correlated these changes to

“certain progressive changes in the sanitary and vaccination circumstances

of the population” (Barry 1889, 248).

Barry used diagrams to better show the most important statistical

representations of mortality and morbidity data for vaccinated and

unvaccinated classes by sub-district (Figure 2.6). Also, Barry included the

floor plans of the various workhouses and hospitals in the area. These

diagrams were included because much of the infected population was housed

in the workhouses for the Sheffield and Ecclesall-Bierlow Districts. The

hospital diagrams are informative as to the design of infectious disease

hospitals of the day and the lack of isolation when compared to modern

hospitals’ infectious disease wards.

Three appendices that followed the main text contain various

information Barry considered important to the analysis but not vital enough

to include in the main body of text. These appendices were titled,

respectively, “Meteorological Returns”, “Rules of Small Pox Associations”,

and “Account of Public Meeting Held on February 8th, 1888.”

“Meteorological Returns” contains tables of barometric readings, wind

direction, cloud formations, hydrometer readings, maximum and minimum

temperatures, rainfall amounts, and general remarks about the weather from

May, June, and July 1887. “Rules of Small Pox Associations” records the

policy information for people that bought insurance against loss of income

from illness and death from smallpox, scarlet fever, and Asiatic cholera.

Most of these associations required vaccination for coverage to be valid and

reporting of any cases in the household to the Medical Officer. “Account of

Public Meeting Held on February 8th, 1888” describes the rejection of the

services of a Mr. Herring who clamed to be able to rid the town of Sheffield

of the disease within a week and the public clamoring for compulsory

notification of infectious disease outbreaks. A section of citizen’s questions

to the Mayor and other town dignitaries were also included here.

Contemporary Sources that Cite Barry

Hainsworth (1889) held a very derogatory view of Barry and the work

done by the Local Sanitary Officers, largely due to his affiliation with the

Yorkshire Union of Anti-Compulsory Vaccination Leagues. He tells of

several interactions with Barry and his associates in attempts to gain access

to hospital records for the purpose of examining their findings. He goes on to

compare the cases and deaths reported between March and December 1887

in an effort to undermine the validity of Barry’s statistics. His information

sources were the Superintendent Registrar’s Books and the Sheffield

Borough Health Report, neither of which contained the complete and

verified information that was used in Barry’s text. He, also, advocated an

anti-vaccination standpoint throughout the text, maintaining that smallpox

was spread by poor sanitary conditions (Hainsworth 1889).

Davidson (1889), in contrast to Hainsworth, takes a more balanced

view of Barry’s statistics and comments. Davidson advances his own theory

on vaccination scarring and the amounts necessary for protection from

smallpox. He uses Barry’s statistics to bolster his argument against the

theory that the number of scars a patient had increased his protection.

Davidson professed that it was the interval at which the vaccinations were

given, with a periodicity of no more than ten years.

Modern Studies that Cite Barry

Morrill and Angulo (1979) cited Barry’s text, along with a series of

studies of poliomyelitis, while addressing a lack of spatiality in discussing

epidemics in the past. They also discussed the importance of spatial analysis

of contagious disease diffusion and demonstrated several methods of spatial

analysis using smallpox data from a 1956 epidemic of variola minor in

Bragança Paulista, São Paulo, Brazil. Angulo, et al. (1980a) called Barry’s

work “intuitive” for analyzing epidemic data in terms of time and space,

over the entire period of an epidemic (Angulo, et al. 1980a, 278). They also

mentioned the pattern of disease transmission in communities, basing the

methodology for their study on the information on Barry and a host of other

studies. Angulo, et al. (1980b) cited Barry when discussing the susceptibility

of persons living with an infected person as compared to persons working

with an infected person. They also pointed out problems with Barry’s maps

of the epidemic, as well as other studies done on other diseases, which was

mainly a lack of mathematical modeling or analysis of spreading

mechanisms.

Mortimer’s commentary section mentioned Barry’s text in a

discussion of other examples of revaccination’s effectiveness in comparison

to the Glasgow epidemic of 1900-02 and the Gloucester epidemic of 1895-6.

The Sheffield epidemic reported 6,088 cases of smallpox and 590 deaths

from smallpox, for an attack rate of 19.3% and a death rate of 1.86% when

compared to the Sheffield population (Barry 1889, 7). A large majority of

the deaths in Sheffield were unvaccinated people or people who had not

been vaccinated in over ten years (Barry 1889). The unvaccinated people

suffered a death rate of 37.2% whereas vaccinated people suffered only a

1.1% death rate (McVail and Mortimer 2002; Barry 1889). The Gloucester

epidemic reported over 6,000 cases and 2,036 deaths, for an attack rate of

15% and a death rate of 5.09% when compared to estimates of the

Gloucester population (McVail and Mortimer 2002; Trobridge 1897). Again,

the vast majority of the deaths were unvaccinated persons, which were more

common in the latter part of the nineteenth century because of inconsistent

enforcement of compulsory vaccination laws (McVail and Mortimer 2002).

The Glasgow epidemic reported 2,555 cases in a population of 675,000 for

an attack rate of 0.4%, however, the case rate for the epidemic is 0.95%

when using the unvaccinated population. When the first cases were reported,

the city began a vigorous revaccination program to control the epidemic.

Their efforts were rewarded; none of the 404,855 revaccinated people

suffered from smallpox. None of the smallpox cases in the Glasgow

epidemic had been recently revaccinated (McVail and Mortimer 2002).

Thomas (1974) in his discussion of airborne transmission of smallpox,

cited Dr. Barry and many other studies. He made the point that there are

many historical epidemics where no apparent contact occurred between

infected individuals and susceptible people. However, new cases of smallpox

would be reported to medical authorities and eventually a connection

between the infected and the new cases was discovered. Mack, et al (1970)

cited Barry briefly in a discussion of discrepancies between death rates for

reported epidemics in Europe and their study in West Pakistan, now

Bangladesh. Mortimer (2003) referred to Barry’s text in passing when

defining mortality statistics during large smallpox epidemics, his article

discussed the dangers of wide spread vaccination in the wake of bioterrorism

threats.

Williams (1994) did not mention Barry’s text, though she made

several mentions of Sheffield’s vaccination and epidemic history during the

height of the Vaccination Acts. She discussed how Sheffield was the home

of a strong vaccination resistance movement. However, there were

conflicting statements on how effectively those who refused to vaccinate

their children were prosecuted. Williams also noted that most children were

vaccinated until the 1880’s when anti-vaccination leagues began growing in

numbers and volume. She cited Barry, but did not mention the text

specifically when discussing the ameliorating effect of any smallpox

vaccination on the length and severity of an attack, even if it was not totally

prevented. She, like Barry, cited a lack of revaccination for adults and any

vaccination for children as the major factor behind the high attack and

mortality rates in 1887-88.

Conclusion

In 1887, the British government sent Barry on a mission to the town

of Sheffield to study the smallpox epidemic and find ways to prevent its

reoccurrence. With meticulous records and detailed analysis, he succeeded

in giving a complete picture of the Borough of Sheffield, its inhabitants, and

its illness. Unfortunately, within a few years of the epidemic Barry’s work

was nearly forgotten. Modern scholars have not deeply analyzed the records

contained in this rich primary document. The few studies that have examined

the document used it only as supplemental documentation to support their

discussion points. The next chapter will present the first modern analysis of

the data recorded in Barry’s text.

CHAPTER 3: CREATION OF THE GIS AND SPATIAL ANALYSIS

This chapter contains two sections. The first section contains a

detailed account of the creation of the GIS in GeoMedia. The purpose of this

section is meant for readers not familiar with the GIS apparatus described.

The second section examines the results of the various spatial analyses

applied to the data using CrimeStat II. All readers should feel free to skip the

section they are not interested in, as both sections are extremely detailed.

Creation of the GIS

The maps used to locate the cases were 1896 and 1905 Charles Goad Fire

Insurance Maps of Sheffield, Yorkshire, United Kingdom. Louisiana State

University

(LSU) funds purchased paper copies of maps. The map set, acquired from

the Sheffield City Archives, consisted of twenty-eight cadastre maps (40 feet

= 1 inch), one city map (400 feet = 1 inch), and a street guide listing the

businesses and streets featured in the map set. The entire package of maps

was scanned in digital format using the large scanner in the LSU Computer-

Aided Design and Geographic Information Systems (CADGIS) lab. Saved as

grayscale “.tif” files, the map images were imported into GeoMedia

Professional 5.1 for modification and analysis. Due to the restriction of the

street-level maps to the city center, only 80 deaths were located and

analyzed.

This project used GeoMedia Professional 5.1 to create the point-

pattern map files. Each cadastre map was brought into GeoMedia

Professional as an interactive image and saved as a separate feature class.

Care was taken to maintain the same scale when importing all 30 images to

insure proper image registration when the time came. Then, two point

feature classes were created. The first point feature class, the ground control

points, was used to reference the twenty-nine cadastre maps to the city map.

The ground control point feature class contained only the intersection or

landmark used and the identification number. The second point feature class,

the deaths, was plotted at the address indicated in Dr. Barry’s text as the

place of death. The death feature class included an identification number, the

address of the death, the date of death, the deceased’s name, age, and

vaccination status (see Table 1). This researcher assigned a date number for

all 181 days of the study period from 30 September 1887 to 31 March 1888.

To georeference the city map, the projection was set for the United

Kingdom Ordinance Survey 1980, which uses longitude and latitude. This

setting was decided upon after experimenting with the Universal Transverse

Mercator, longitude and latitude, and British National Grid projections.

Then, after finding and matching the coordinates for five reference points in

the city map (see Table 2), GeoMedia processed and registered the map.

This allowed for the assignment of longitude and latitude coordinates to the

death points through vector registration of the cadastre maps to the city map.

The city map was geo-referenced to acquire longitude and latitude

values for each death using the vector registration process. This process

allowed the transference of the ground control points and the death points

from the cadastre maps to the city map. In GeoMedia, creating spatial filters

for each cadastre map and registering them individually was the simplest and

most effective method for handling the large volume of points and maps in

this project. Using the “Create Spatial Filter by Fence” function under the

“Warehouse” menu, a filter was created for each cadastre map. Those filters

were applied to restrict vector registration to the cadastre map currently

under examination. This prevented duplicate registration of the points. Then,

under the “Tools” menu, the “Vector Registration” function was activated. A

new registration file was created for each cadastre map with four to five

ground control points assigned to each map. The first point was identified on

the cadastre map then, matched to its point on the city map. Once all the

ground control points were matched, the points were transferred to the

registered city map using the “Transform” function. At this point, both the

control points and the death points were transferred to the georeferenced

image. By processing all twenty-nine cadastre maps this way, a complete

image with all the deaths and control points was created (Figure 3.1).

However, ten death points and six control points transferred incorrectly and

were corrected in a second run of the analysis. These errors skewed the first

run of CrimeStat results, as discussed in the Analysis section of this paper.

The correction of the incorrectly transferred points was fairly simple.

Using the

“Move” button in the points toolbar, then clicking on the incorrect point and

the correct location, the death point was transferred . The same method was

used to transfer the ground control points but no other steps were taken as

they were not required for the spatial analysis. After the points were

physically moved, the geometry of those points needed to be updated in the

data tables because the tables are exported for spatial analysis. That was

accomplished by going to the “Edit” pull-down menu and selecting “Update

Attributes.” Then, by setting the conversion equation for longitude to read

“text(X(CoordGeometryPoint,TrueMeas,deg))(“00.00000”)” and for latitude

to read “text(Y(CoordGeometryPoint,TrueMeas,deg))(“00.00000”)” then

clicking “OK”, the table was updated and exported.

Using GeoMedia allowed for the creation and exportation of a

database file that contained the coordinates for each death (See Appendix 1).

Using the “Analyze Geometry” function in the “Analyze” menu, GeoMedia

created this coordinate file. Databases containing coordinates in both

decimal degrees and “degrees:minutes:seconds” formats were created. Then,

under the “Warehouse” menu, a Microsoft Access database file was created

and exported. The file was then converted to a Microsoft Excel file. Once in

Excel, the file was adjusted to fit all data in the columns and hide the

seemingly empty geometry columns created by GeoMedia. It is important to

remember to never delete the geometry columns in either Access or Excel

because this can erase the registration of the points when GeoMedia is

reopened. Labeling of the coordinate columns with longitude and latitude

also occurred here. The complete file was saved in “.dbf” format to meet the

CrimeStat requirements.

Analysis

The “.dbf” file using decimal degrees, taken from GeoMedia, was

imported into and analyzed by CrimeStat 2.2, a program written by Ned

Levine & Associates under grants from the National Institute of Justice.

This spatial statistics program, designed for the analysis of crime incident

locations, is Windows-based and creates outputs that are easily imported into

most desktop GIS programs. The purpose is to “provide supplemental

statistical tools to help law enforcement agencies and criminal justice

researchers in their crime mapping efforts” (Levine and NACJD 2003, 1).

However, the CrimeStat program has been used by disciplines for

applications outside of crime studies. The program analyses incident

locations in “.dbf”, “.shp” or ASCII formats with either spherical or

projected coordinates for input files. It then calculates various spatial

statistics and writes graphics files for ArcView, GeoMedia, MapInfo, and

other imaging programs (Levine and NACJD 2002). For this project, the

calculated statistics were imported back into GeoMedia to visualize the

statistics.

For the primary examination of the data, a variety of the statistics

available in CrimeStat were used to analyze this data set because of the

exploratory nature of this project. Upon opening the program, the user has to

set the source file and variables under the “Data Files” menu. This was done

in the “Primary File” tab by using the “.dbf” file and setting the “X” as

longitude, “Y” as latitude, and “Date” as date number. The “Reference File”

tab was set with the Lower Left coordinates reading “X” as “-1.45” and

“Y” as “53.30” and the Upper Right coordinates reading “X” as “-1.48” and

“Y” as “53.40.” The “Measurement” tab was set to an area of 100 square

miles, 500 miles of roads, and direct distance measurement.

The statistics selected on the first trial were more extensive than those

on the second trial. This was the first time that Dr. Barry’s dataset had been

analyzed by modern spatial software and there was no driving hypothesis.

For the first trial, many statistics were used. From the “Spatial Description”

menu, at least one statistic from each subset was selected. From the “Spatial

Distribution” subset, the mean center and standard distance, the standard

deviational ellipse, the median center, and the center of minimum distance

were selected. From the “Distance Analysis” subset, nearest neighbor

analysis was selected. From the “Hot Spot One” subset, nearest neighbor

hierarchical analysis was selected. From the “Hot Spot Two” subset, the K-

mean statistic was selected. Only three statistics from the “Spatial Modeling”

section were run on this data set for the first trial.

From the “Interpolation” subset, kernel density analysis was selected and,

from the “Space/Time Analysis” subset, the Knox and Mantel tests were

selected.

Mean center and standard distance calculations are used to determine

the location of the arithmetic mean and dispersions of points in the data set.

The direct mean center can also be considered the center of gravity of the

data set; it is where the mean of the X and Y coordinates intersect (Levine

and NACJD 2003, 109). The geometric mean center is the mean center of

the logarithms of the data set and reduces the effects of any extreme values

(Levine and NACJD 2003, 127). The harmonic mean center is the “inverse

of the mean of the inverse of X and Y respectively” and also reduces the

effects of extreme values (Levine and NACJD 2003, 130). A standard

deviational ellipse shows the dispersion, shape, and orientation of the points

in a data set, usually around a mean center. (Levine and NACJD 2003, 120).

The triangulated mean center and the weighted triangulated mean

center are products of the directional mean center calculations. The

triangulated mean center is the intersection of the maximum of the X and Y

coordinates and the minimum of the X and Y coordinates in the dataset. The

weighted triangulated mean center is the intersection of the maximum of the

X and Y coordinates and the minimum of the X and Y coordinates in the

dataset but weighted according to a variable, usually population (Levine

and NACJD 2003, 150). Median center is the intersection of the median X

coordinate and the median Y coordinate of all the points in the dataset

(Levine and NACJD 2003, 114). The geometric median center is the point at

which the lines that divide the dataset in half intersect (Levine and NACJD

2003, 114). The center of minimum distance defines the point at which the

distance between all points is at a minimum. This is important because this

point is ”the point at which the sum of the distance to all other points is the

smallest” (Levine and NACJD 2003, 120).

Nearest neighbor analysis gives an estimate of the amount of

clustering or dispersion among the points in a dataset by comparing the

distance between nearest points to the distance that could be expected by

chance (Levine and NACJD 2003, 171). Nearest neighbor hierarchical

clustering analysis groups points together based on proximity, the number of

cluster groups, and the statistical significance level. The grouping is assigned

a threshold distance and minimum number of points; if the points in the

dataset cannot meet the requirements, then no cluster is shown. This

continues from first-order clusters through the number of clusters specified

by the person running the analysis (Levine and NACJD 2003, 216).

K-means clustering is a clustering routine that assigns the points to

seed clusters according to the original seed cluster numbers decided by the

person running the analysis. It is similar to nearest neighbor hierarchical

clustering analysis because it assigns the points to only one cluster.

However, K-means clustering only uses first-order clusters in the analysis.

The user-set seed clusters allows the person running the analysis to control

the number of groups and, possibly, manipulate the results (Levine and

NACJD 2003, 273). Kernel density analysis examines the density of points

in a single distribution when overlaid with a symmetrical surface. It allows

for the incidents to be generalized over an entire region (Levine and NACJD

2003, 301).

The Knox index shows the relationship between ‘closeness in time’

and

‘closeness in distance’ for pairs of points in the data set (Levine and NACJD

2003, 46). This simple test can quickly show any correlation between the

closeness in distance and the closeness in time but, it can be difficult to tell if

the significance comes from the distance element or the time element

(Levine and NACJD 2003, 423). The Mantel index shows “the correlation

between ‘closeness in time’ and ‘closeness in distance’ across pairs” (Levine

and NACJD 2003, 46). It solves some of the problems with the Knox test

and makes it easier to determine the source of significant results by

calculating the amount of correlation between the distance and time elements

(Levine and NACJD 2003, 424).

The mean center and, the median center, and the center of minimum

distance all created points that centered in cadastre map 20. Also located

here are the central portions of the standard distance and the standard

deviational ellipse calculations (Figure 3.3). While there was only one death

in map 20, upon visual examination, these centrographic statistics do lie

between three areas of denser deaths. The areas figures created by the

standard deviational ellipses run along a northwest to southeast orientation

that covers the major death concentrations in the southern part of the map

(Figure 3.4). However, both the centrographic statistics and the standard

deviational ellipses were skewed because of the thirteen incorrectly plotted

deaths from the first trial. The misplaced points were to the northwest and

southeast of the study area which accounts for the skewed statistical

displays. It was at this point where this researcher realized the death points

were misplaced and sketched the plan for the second trial. But, first, the rest

of the first trial analysis will be addressed.

Nearest neighbor and nearest neighbor hierarchical cluster analyses were run

multiple times with various settings (Figure 3.5). The trials included a

minimum of five and ten points per cluster and small (0.001), mid-range

(0.1), and large (0.99) significance levels (Levine and NACJD 2003, 218).

The five-point minimum clusters for large (0.99) and mid-range (0.1)

significance revealed a number of clusters that covered much of the same

area, as seen when examining the yellow and green ellipses in Figure 3.5.

The large (0.99) significance setting created five clusters and the mid-range

(0.1) significance level created six clusters. However, two of the mid-range

(0.1) clusters covered the same area as the one of the large (0.99)

significance clusters. The ten-point minimum clusters used the small (0.001),

mid-range (0.1), and large (0.99) significance levels in its analysis of the

data. There was no output for the large (0.99) significance level, but the

small (0.001) and mid-range (0.1) significance levels each showed one

cluster. The small (0.001) significance level, seen in pink, overlays one set

of the fivepoint minimum clusters while the mid-range (0.1) significance

level, seen in blue, overlays another. The blue cluster was the only cluster

affected by the corrections made when moving misplaced deaths because

three of the deaths in this cluster were relocated to their proper place in

Maps 2 and 4. Without the blue cluster, the analysis indicates that one

statistically significant cluster of smallpox deaths existed in this study of the

epidemic. This cluster centered in maps 2, 15, 16, and 17 in the northern

portion of the study area, which had several blocks that suffered multiple

deaths. This area was a particularly densely populated area due to the wharfs

and docks of the River Don, the industrial areas immediately to the west and

north, and the city center to the east and south (Barry 1889). In map 2, the

clustering of deaths was noticeable with three members of the same family

dying at the same address, 68 Snig Back Hill, within one month of each

other. This pattern of spatial clustering within homes, courtyards, or blocks

was common in this epidemic. It seemed that the interaction between

residents was a critical factor in the epidemic’s diffusion.

For the second trial, the same tests were run with the same settings. The only

difference between the trials was the updated database created after the

misplaced deaths were corrected. A slight shift in the centrographic statistics

and the orientation of the elliptical measurements were expected as a result

of these corrections. This was the result for most of the statistical tests run.

The mean center, the standard distance, the median center, and the center of

minimum distance all shifted slightly northeast into Map 5 (Figure 3.6).

Even the mean center ellipse, in red, and the rectangular mean center, in

yellow, shifted only slightly northeast. However, the complete reorientation

of the standard deviational ellipses, the first standard deviation in black and

the second standard deviation in green, was surprising (Figure 3.7). There

were nine deaths in the northwestern portion of the map and four deaths off

of the map to the southeast that were corrected which accounts for the

dramatic change.

The most surprising change of the second trial was the reduction in

the number of clusters from the nearest neighbor hierarchical clustering

analysis (Figure 3.8). The blue cluster from Figure 3.5, that was expected to

change with the corrections, did disappear. However, so did all of the other

small clusters from the previous analysis. Despite using the same parameters

for five- and ten-point clusters and small, mid-range, and large statistical

significance, only three clusters were produced. The mid-range significance,

five-point cluster, in dark blue, and the mid-range significance, ten-point

cluster, in light blue, and the large significance, ten-point cluster, in light

green, covered the study area and had a different orientation than the

standard deviational ellipses. The marked northeast-to-southwest orientation

is similar to the orientation of many of the small clusters from Figure 3.5 but

that is the only real similarity. The other interesting area that was consistent

in both trials was the open space in the center of the city, specifically all of

maps 3, 4, and 5, the southern portion of map 2, the western portion of map

8, and the northern portion of map 20. These areas were free of any deaths

because this was the business center of the city and very few people lived in

this part of the city. Many churches, government buildings, and market

spaces were built in that area of the city. Much of the industrial and

residential districts of Sheffield were built in the southern, far western, and

far eastern portions of the city, as well as the vast industrial complex north

of the River Don (Barry 1889).

For both trials, no matter the variation in settings, some of the tests

attempted gave no valid results. The K-mean statistics were run with one and

two seed clusters. The ellipses that resulted from the first trial were skewed

similarly to the standard deviational ellipses in Figure 3.5. The ellipses that

resulted from the second trial were identical to the nearest neighbor

hierarchical clusters in the second trial and were not included in the images

for this study. The kernel density analysis was not successful for the first

trial because the boundary coordinates acquired in GeoMedia need to be

converted from “degrees:minutes:seconds” format to decimal degrees

format. The kernel density analysis was not successful for the second trial

because the CrimeStat program does not seem to accept the Ordinance

Survey projection that was selected for this data. For the Knox and Mantel

tests, both of which are time-series analyses, the time format in the database

did not correspond to the correct CrimeStat format in the first trial. In the

second trial the creation of the “DateNumber” column was an attempted

solution to this problem. The Mantel test gave results but when examined

against the mapped points and the original data, they were inconclusive. The

Knox test continued to produce irrecoverable errors for no reason that this

researcher could discover. Unfortunately, the more advanced statistical tests

produced no results but, further research, discussed in Chapter 5, will

hopefully address these issues.

Conclusion

Significant areas of interest have been discovered in the progress of

this project, including the variation in cluster results between the two trials

of the study. The periodicity of the deaths in relation to their location will be

of significant interest in further analysis of this dataset. However, the most

important achievement of this project was the creation of a fully

georeferenced map with a database of all deaths. This project began as a set

of tables listing deaths and demographics of people who died one

hundredfifteen years ago. Barry, the author of that text, analyzed the data he

collected to the best of his ability. Included in his study are hundreds of

tables of statistics on death rates and infection rates for vaccinated and

unvaccinated classes, as well as dozens of maps tracking the deaths across

the city of Sheffield. This project was an extension of that endeavor. When

that data set was combined with the street level maps, which were not

available until decades after the epidemic, and a GIS, allowing for a more

sophisticated spatial analysis. If modern GIS technology and techniques had

been available to Barry in 1887, this is the most probable avenue of research

he would have pursued.

CHAPTER 4: COMPARISON TO CURSON’S TIMES OF CRISIS

Introduction

In the past, two general approaches have emerged in the field of

historical medical geography. These include the analysis, sometimes spatial,

of historical data in order to examine the pattern of a historical epidemic,

like this project. The second method examines the social and cultural effects

of an epidemic on a city’s population, like Craddock’s 1995 study of disease

in nineteenth century San Francisco. P.H. Curson combined both approaches

as he examined a series of epidemics in nineteenth century Sydney,

Australia, including the 1881-2 smallpox epidemic (Curson 1985). This

chapter will compare Curson’s analysis of the 1881-2 Sydney epidemic to

this study of Sheffield’s 1887-1888 epidemic, in order to determine

similarities and differences in the spatial epidemicity.

Curson’s Times of Crisis

P.H. Curson, a senior lecturer of Human Geography at Macquarie University

Australia, surveyed the epidemic history of Sydney from 1788- 1900 in his

book Times of Crisis. By using numerous historical sources, including

church burial records, death records from the Registrar-General of New

South Wales, convict burial records, and hospital records, he created a series

of maps for several of the major epidemics in Sydney. The six epidemics

Curson examined were the 1789 great epidemic, believed to be smallpox, the

measles epidemic of 1867, the 1875-1876 scarlet fever epidemic, the

smallpox epidemic of 1881-1882, the 1890-1891 Asiatic flu pandemic, and

the plague epidemic of 1900. Curson also used newspapers and public

records to reconstruct the social and political repercussions of these

epidemics (Curson 1985).

The chapter of the book that related to this project discusses the

smallpox epidemic of 1881-1882 in Sydney. Curson presented the history of

smallpox in Australia and analyzed the spatial aspects of the epidemic. Other

sections in the chapter discussed the official reactions to the epidemic by the

Sydney government, including the methods by which they tried to control

the epidemic. He also reviewed various primary sources to discover the

population’s fears and opinions about smallpox, the source of the epidemic,

and the issue of vaccination. By examining the spatial and social aspects of

this epidemic, he gave a unique picture of how diseases in general, and

epidemics, specifically, affect communities (Curson 1985).

Fifteen smallpox outbreaks occurred in Australia between 1789 and

1917, which marked the period between the arrival of Europeans on the

continent and the full establishment of vaccination in Australia. Sydney

usually had a few dozen cases and only a few deaths in each outbreak. The

notable exceptions were the 1789 epidemic, the 1829-1845 outbreaks, the

1860-1869 outbreaks, and the 1913- 1917 outbreaks, which reported cases in

the thousands and deaths in the dozens or hundreds. The 1881-1882 outbreak

was exceptional because of the accuracy and completeness of the records of

the 163 cases and 41 deaths. However, Curson did note that the numbers of

cases and deaths were probably higher than reported for two reasons.

Primarily, citizens were hiding cases from the authorities for fear of

quarantine. Secondarily, doctors were misdiagnosing cases of smallpox as

chickenpox, which was also epidemic at the time. Another remarkable

circumstance of the epidemic was the violence of the public reaction toward

the Chinese immigrants, which was unique to this epidemic in Sydney’s

history (Curson 1985).

The first case in this epidemic was reported on 25 May 1881 in the

son of a prominent Chinese merchant. It was believed that the child was

infected by his nurse but there was no information on where she was

infected. The usual method of infection for a city like Sydney was

importation from another port city. The epidemic was declared over on 19

February 1882. The worst periods of the epidemic were from early August

1881 to mid-September 1881 and from mid November 1882 to the end of

January 1882. Curson stated that the reasons for the epidemic’s long stay in

Sydney were “closely related to the slow smouldering nature of the disease,

the length of incubation period, the number of susceptibles and the intimacy

of personal contact in high risk areas” (Curson 1985, 94). According

to the map and charts Curson created, as seen in Figure 4.1, there were five

clusters of cases and deaths, circled in red, in this epidemic. All of these

clusters had death rates between 21.1% and 30.8% (Curson 1985). This was

consistent with death rates for unvaccinated populations at this time, as seen

in contemporary texts (Barry 1889; Davidson 1889; Hainsworth 1889;

Trobridge 1897). Curson continued with more in-depth discussions of the

population segments most strongly affected by the epidemic, including age

groups and occupational classes. Curson’s data showed that the age group

most affected by the epidemic was children infected by smallpox under the

age of 10 years, who suffered a 30.4% mortality rate and contributed 41.6%

of the total epidemic deaths (Table 4.1 and Figure 4.2). The second most

affected group was the 40 to 50 year old group, who suffered a higher

mortality rate of 37.5% but only represented 14.6% of the total epidemic

deaths (Curson 1985). For the study area, Barry reported a mortality rate of

31.3% among children under 10 years of age infected with smallpox. The

older population suffered from only a 5% mortality rate in Sheffield, as seen

in Table 4.2 and Figure 4.3 (Barry 1889).

The difference between Sydney and Sheffield seems to be the number

of people who had undergone vaccination. The majority of Sydney’s

population was not vaccinated because Sydney did not have an established

compulsory vaccination program. It was likely that only a small percentage

of the population had been previously infected, especially since the last

major outbreaks were in 1869 (Curson 1985). Sheffield’s older population

had lived in times of both compulsory vaccination and frequent smallpox

outbreaks, resulting in a low number of susceptibles in the older population

(Barry 1889). Between 1870 and 1890, the Sheffield area became a center of

anti-vaccination groups, which meant a significant minority of Sheffield’s

youth had never been exposed to smallpox (Williams 1994). This was the

cause cited by Barry and his contemporaries as the probable reason for the

high mortality rate for the population under age 20 (Barry 1889).

In addition, Curson stressed that all of the areas under discussion were

predominantly low-income, low-status, and crowded areas of the city with

exceedingly poor sanitation. He pointed out that only two of the 41 official

deaths occurred among members of the merchant class; the rest were mainly

in the laborer and tradesman classes. This comparison is illustrated in Table

4.3 and Figure 4.4 (Curson 1985). When the occupational classes of the

deaths in the Sydney epidemic are compared with those in Sheffield (Table

4.4 and Figure 4.5), it is apparent that the unskilled and skilled labor classes

suffered a significant share of the mortality in both epidemics. These classes

also represented a significant share of the overall population because Sydney

was a booming colonial port city and Sheffield was the top steel-producing

city in Europe during the latter part of the nineteenth century (Tweedale

1987; Curson 1985; Barry 1889).

Table 4.1: Age-sex structure of smallpox

deaths in Sydney 1881-2 and in

Sheffield, 1887-8.

Age Sydne

Male

Sydne

Fema

Sydne

Total

Sydney

Mortali

Sheffie

Male

Sheffie

Female

Sheffie

Total

Sheffiel

Mortali

yrs

9 8 17

41.6%

14 11 25

31.3%

102

yrs

1 3 4

9.6%

12 6 18

22.5%

203

yrs

5 4 9

22.0%

10 7 17

21.3%

304

yrs

5 1 6 14.6% 10 4 14

17.5%

405

yrs

5 1 6

14.6%

3 1 4

5.0%

50+ 0 0 0 0.0% 1 1 2 2.5%

Tot

25 16 41 50 30 80

Table 4.2: Table of deaths by occupational class in

Sydney smallpox epidemic, 1881-2.

Occupational Class Number of

Deaths

Unskilled 9

Semi-skilled 2

Skilled 6

Shopkeeper/

Merchant

Housewife 1

Orchardist 0

SS Garrone / Military 2

Not Stated 19

Total 41

Table 4.3: Table of deaths by occupational class in

Sheffield smallpox epidemic, 1887-8.

Occu

patio

nal

Clas

Unsk

illed

Skill

Shop

keep

Milit

ary

none 9

give

Total 80

To verify the differences or similarities between the numbers of deaths

in Sydney and in Sheffield, chi-square tests were run on the mortality

statistics using SPSS 11.0. Even when comparing the total deaths, the male

deaths, and the female deaths, there was no significant difference between

the numbers of deaths in the two epidemics. None of the statistical tests gave

statistically significant results, i.e., below the 10% or 5% thresholds. All of

the test resulted in significance levels of 68.3% or 98.2%.

Curson also created maps that tracked the spatial diffusion and person-

to-person contacts that spread the disease (Figure 4.6). However, there is no

discussion in the text of the methods used to create his various maps, spatial

diffusion diagrams, and statistical tables. However, he did discuss the

various contacts between the infected and the susceptibles and how the

disease spread from one neighborhood to another in detail. There are three

major factors Curson isolated in the spread of this epidemic. The first factor

was the “spatial mobility” of the patients and their contacts (Curson 1985,

99). The second factor was the socio-demographic influences on movement

patterns in the community. The third factor was government policies on

vaccination, quarantine, and sanitation (Curson 1985).

The method by which smallpox was spread through Sydney was of

particular importance in Curson’s discussion of the epidemic. Through his

close examination of the records used in his reconstruction, he identified the

five most likely situations for the contracting of the disease, seen in the

quote below.

1. Intra-household diffusion where the disease

spread from one family member to another or a

lodger within the house.

2. Transfer of the infection from one house to

the next- door or neighbouring house via patterns

of neighbouring and social intercourse (such as the

sharing of basic sani- tary facilities).

3. Transfer via the medium of longer distance

visiting (between kin or friends).

4. Transfer associated with recreation,

shopping, or work. 5. Where close association with

infected persons in quar- antine produced the

disease. (Curson 1985, 99-100).

It was apparent in Curson’s text that close personal contact was the most

important link between cases because close personal contact “accounted for

almost all cases of smallpox in Sydney where it has proved possible to

identify the medium of infection” (Curson 1985, 100).

The remaining sections of Curson’s study discussed the various

reactions to the epidemic and included samples of government policies,

medical actions, and popular attitudes on the subjects of vaccination,

quarantine, and the source and spread of the epidemic. This epidemic caused

a major shift in governmental policies. Before this epidemic, there was no

public health system and there were no policies for dealing with infectious

disease notification, vaccination, and quarantine. Uncertainty as to the extent

and location of the infected persons caused public hysteria throughout the

city. The public outcry resulted in the government requiring public

notification of infectious diseases. Quarantine was the primary method the

Sydney government used to control the epidemic. Unfortunately, the

government was draconian in its application of the quarantine policy,

resulting in far more people being quarantined than necessary. Also, racial

bias was a major factor in the stringency of enforcement, with the Chinese

population suffering the worst treatment. This epidemic resulted in increased

demand for public vaccination and government supply of optional

vaccinations for those who had contact with smallpox patients (Curson

1985). However, Curson did not mention the various Vaccination Acts

passed by the British Parliament that mandated smallpox vaccination for all

children within three months of birth, nor if these laws were enforced in

Sydney (Curson 1985).

Sheffield had no formal quarantine policy. The city relied on the

Vaccination Acts and a strong public campaign for universal compulsory

vaccination and revaccination to control the spread of the epidemic. Infected

people were only removed from their homes to the local hospitals if their

doctors decided it was necessary to care for them. Most of the infected

remained in their homes and received visitors in their sick beds. Barry

specifically noted:

Wherever a case of sickness is known to exist there they

congregate, in some instances led by curiosity, but in the

majority no doubt from a genuine desire to be of service.

Heedlessness of danger was almost universal.

[Emphasis added] (Barry 1889, 386).

There seemed to be no public outcry for a quarantine policy, which is

markedly different from the situation in Sydney.

Sydney’s medical community did take action, some beneficial and

others detrimental to the public. The negative aspects of medical care during

the epidemic were varied and shocking in their disregard for the patients’

health. Some of these horrors included the abandonment of quarantined

patients who were shut in their homes by government order and the poor or

non-existent medical care for those in hospital quarantine. Misdiagnosis,

which resulted in unnecessary quarantining, was the rule, not the exception,

in this epidemic. Part of this was due to the concurrent chickenpox epidemic,

but most misdiagnosed cases resulted from a combination of incompetence

and panic. However, there were a few positive moves made by the doctors in

Sydney for the long-term health of the city. The Sydney branch of the British

Medical Association requested a government public health authority for the

whole colony, which was instituted by mid-July 1881. This helped institute

some order in dealing with the epidemic. They also supported a bill for the

compulsory notification system for infectious disease cases, which was

passed in December 1881 (Curson 1985).

Popular attitudes during the epidemic were critical in Curson’s

evaluation of the severity of the epidemic. He surveyed newspaper articles,

opinion essays, and transcripts of public meeting to determine the public’s

perception of the epidemic’s origins, pattern of diffusion, and possible cures.

The cartoon in Figure 4.7 was published in July 1882 and summarized

contemporary public opinion. The cartoon addressed fears about vaccination

and quarantine among the white population, as well as the over-zealous

pursuit of the Chinese population as a cause of the epidemic. Even the

medical communities were berated for their apparent reluctance to treat the

smallpox cases due to their own health concerns (Curson 1985).

The issue that was of greatest concern for the government, the medical

community, and the white population of Sydney was the perceived threat of

contact with the Chinese community. These groups firmly believed that the

Chinese immigrant community brought smallpox with them when they

arrived in Australia and hid cases from the government for a period of time,

resulting in the 1881-1882 epidemic. However baseless this concern may

have been, it still resulted in the appalling treatment of the Chinese

population during the epidemic. The Chinese were subjected to compulsory

vaccination with no option to refuse and government inspections to

determine the infection status of each household. When the Chinese were

forced into quarantine, they were taken with no notice and no possessions

and their homes were fumigated or burned to the ground. This treatment was

markedly different from the treatment of the white population, who when

confined to quarantine were given notice and the option to remain confined

to their homes. The white citizens were also given the option of voluntary

vaccination by the government. They objected to any measures for

compulsory vaccination despite the medical communities overwhelming

support of the measure. Later that year, this epidemic, with its origins in the

Chinese community, was used to support the restriction of Chinese

immigration to Australia after 1882 (Curson 1985). The panic and

xenophobic reaction that this epidemic caused in Sydney was not unusual

because as Humphreys (2002, 846) stated, “The diseases that cause panic are

not usually the diseases that kill the most people on a daily basis.”

In striking contrast, Barry’s text makes no mention of a particular

group that suffered blame for the Sheffield epidemic. Because the relative

uniformity of the population, mostly white, skilled labor, no cohort in the

city suffered from excessive real or imagined culpability for the epidemic.

Only the unvaccinated population was under any excessive public pressure

to change their behavior (Barry 1889). Sheffield’s government authorized a

campaign to inform the public of the need for everyone to update their

vaccination. The city posted placards and placed notices in the local

newspapers with information on free public vaccinations and methods for

preventing the spread of the disease (Barry 1889; Hainsworth 1889). Instead,

the Winter Street Hospital, the city hospital designated as the smallpox

hospital during the epidemic, was blamed for higher rates of smallpox cases

and deaths in the surrounding neighborhoods. Barry calculated that the

homes within a 4,000-foot radius of the hospital were twice as likely to

contract smallpox as the rest of the city. The homes within a 2,000-foot

radius of the hospital were thrice as likely to contract smallpox as the rest of

the city.

This situation was used to justify the cost of constructing a new

isolation hospital four miles from the city center in the last months of the

epidemic. When the remaining patients were moved to the new hospital, in

late March, the high number of cases and deaths in the Winter Street

neighborhood “disappeared” (Barry 1889, xii). These differences indicate

that the majority of the damage from the Sydney epidemic resulted from

social disorder and government mismanagement, while the Sheffield

epidemic had distinct spatial features related to the location of public

facilities and interactions in the urban environment.

Conclusion

The Curson text does provide an unusual method for examining

historical epidemics. The reconstruction of the dataset from various sources

and the creation of the flow maps describing the epidemic are impressive.

This and the social impacts of epidemics are areas of study that have come

under increasing scrutiny with the rise in modern bioterrorism issues. This

study of the Sheffield epidemic developed a unique method of creating a

GIS for analyzing and modeling historical epidemics. Comparing the two

projects has given a picture of the range of possible studies of historical

epidemics and a possible avenue for further research on the Sheffield 1887-

1888 smallpox epidemic.

CHAPTER 5: FURTHER RESEARCH AND CONCLUSIONS

This chapter addresses three issues, leading from a discussion of this

project to issues for future research. First, an overview of the common

problems associated with moving a historical data set from its primary

source documents to modern databases for further analysis. The second issue

that will be addressed is the way historic epidemics can be used in planning

for a bioterrorist attack. The third section will enumerate future research that

could spring from this project. The concluding section will make some final

remarks on Barry’s intentions for his text and how this study has begun

fulfilling those wishes.

Problems with Historical Epidemic Data

All datasets, modern and historical, have their problems and

inconsistencies. Hardy (1993) believes it is necessary to be aware of the

possible limitations of these data sets before becoming too engrossed in

construction of a GIS and subsequent analysis. Many writers in geography

and disease agree and expound upon the data problems that accumulate

when dealing with historical epidemic data. Common problems with

historical epidemic data sets include errors in locational and temporal

aspects of the deaths, poor or incomplete death records, inaccurate patient

data, and the misdiagnosis of the disease (MacKellar 1993). Geographers

have even created requirements for analysis of a historical epidemic data set

including the need for a contemporary understanding of the disease, a

clinical definition of the disease, and an adequate reporting mechanism in

the location of the outbreak (McEvedy 1988; Cliff and Haggett 1988).

Cliff and Haggett (1988) wrote a detailed discussion of historical

epidemic data

problems in their Atlas of Disease Distributions. Most of these issues deal

with the reliability of the data over a period of time. The first major problem

was insuring that all reporting areas were reporting the same disease. In

some cases, diseases can be similar and it was difficult for medical personnel

in historic periods to make a definitive diagnosis without modern

pathological methods (Cliff and Haggett 1988). In the Sydney epidemic,

Curson noted the misdiagnoses of smallpox cases as chickenpox and vice

versa and how that resulted in inaccurate case and death reports (Curson

1985). This type of misdiagnosis was less likely in Barry’s text because

there were no other epidemics occurring in Sheffield at that time. However,

Barry did note that some cases were unreported due to their mildness or fear

of stigmatization. However, the deaths that were reported had virtually no

chance of being diagnosed as anything but smallpox (Barry 1889).

The second major problem is spatial in nature. This involves changes

in the sizes and areas of statistical reporting districts over time. The

inconsistency in the districts’ size and area can make plotting and analyzing

cases and deaths difficult because the areas should be adjusted so that they

are uniform over time (Cliff and Haggett 1988). The third major problem

involves the location and the year that the data was collected and how the

demography of a population changes over time (Cliff and Haggett 1988). By

this, the authors are addressing the fact that time passes and emigration,

immigration, births, and deaths alter the structure of the population. These

changes must be taken into account when dealing with data sets that cover

several years. The fourth major problem is that a geographical location is

often missing from records in a time-series when tracking a disease over

several years or decades (Cliff and Haggett 1988). Natural disasters and

human errors can occur, which can lead to the loss of that data. The fifth

problem of historical epidemic data requires that “changes in disease data

need to be assessed in terms of a reference or an ‘at risk’ population” (Cliff

and Haggett 1988, 65). In this the authors were referring to the need to

normalize any morbidity or mortality numbers against a population for the

region under discussion. This allows for comparisons that are more accurate

across years and regions. These last four problems are most troubling when

discussing a multi-annual epidemic or series of epidemics (Bengtsson and

Lindstrom 2003; Cromley 2003; Edsall 2003; Flint 2003; McVail and

Mortimer 2002;

Smallman-Raynor and Cliff 2001; Borroto and Martinez-Piedra 2000;

Haggett 2000;

Banthia and Dyson 1999; Becker, et al. 1998; MacKellar 1993; Wilson

1993; Cliff,

Haggett, and Smallman-Raynor 1998; Cliff and Ord 1995; Cliff and Haggett

1988;

Curson 1985; Mielke, et al. 1983; Pyle and Rees 1973; Pyle 1971; Hunter

and Young 1971). Many of these issues are not noticeable for the Sheffield

data, as no multi-year temporal comparison is made.

Applications Today

In the years immediately preceding the eradication, smallpox killed

over 2 million people a year worldwide and maimed up to three times that

number (Ryan 1997). Since the World Health Organization’s (WHO) official

eradication of smallpox in 1980, the fear and horror the disease inspired in

people has waned (Netys and DeClercq 2003; Tucker 2001; Henderson

1978). However, this former scourge of humanity has experienced a surge in

concern among government officials, epidemiologists, and military

personnel in the past few years because of its potential as a biological

weapon. In the twenty-first century, terrorism has become a critical concern

of all nations and acts of bioterrorism are perhaps the most frightening of all

(Eichner and Dietz 2003; Crosse 2003; Humphreys 2002; Heritage

Foundation 2002; Katz 2002; Preston 2002; Tucker 2001; Meltzer, et al.

2001). This concern about potential bioterrorist attacks is the most critical

reason for research into historical epidemics. The more information

documented on the behavior of diseases, especially highly contagious

diseases like smallpox, the greater the ability of governments to protect

citizens from the work of terrorists.

The United States spends millions of dollars per year to prepare for

potential bioterrorist attacks. Programs like the National Pharmaceutical

Stockpile Program (NPSP), the Health Alert Network (HAN), and the

Epidemic Information Exchange (EpiX) have been created to provide

emergency medical supplies, general communication system upgrades, and

secure medical communications for public health departments, respectively.

All of these programs are designed to monitor and control public health

situations and notify officials in case of a bioterrorist event. These systems

were designed to bolster the existing biological weapons surveillance

agencies like the Centers for Disease Control (CDC) and the Epidemic

Intelligence Service (EIS). These agencies have trained many of the public

health leaders in the past 50 years and continue to be influential forces in the

world’s public health community (Katz 2002; Meltzer, et al.

2001).

Biological warfare, both unintentional and intentional, has a long

history in human warfare. In modern times, most governments view the use

of such weapons as inhumane and unethical. However, most of these

governments have sponsored biological weapons research at some time in

their history. During World War II, both Allied and Axis nations developed

biological weapons in response to intelligence reports that their enemies

were researching such weapons (Miller, et al. 2002). The weaponization of

diseases like botulism, tularemia, and smallpox consumed military resources

throughout the Cold War, despite the public condemnation by all nations

involved. For nearly 50 years, biological weapons research was the primary

function of two of the largest medical research facilities in the world,

Russia’s Vector laboratories in Koltsovo on the Siberian plains, and

America’s Fort Detrick in Maryland (Miller, et al. 2002). Both Gulf Wars,

in 1992 and 2002, were fought because of Iraq’s possession of biological

weapons material, including weaponized variants of anthrax and camelpox

(Heritage Foundation 2002). In 1999, at a U.S. counter-proliferation

briefing, the presenters compared the Russian and American production of

germ agents by metric ton per year. The U.S. had much smaller amounts of

weapons-quality germs, including 1.6 metric tons of tularemia and 0.9 metric

tons of anthrax. Russia had incredible amounts of weapons-quality germs,

including 4,500 metric tons of anthrax and 100 metric tons of smallpox. The

concern of most of the world’s governments is that the disposal of these

weapons cannot be completely verified and, former Russian scientists cannot

guarantee that the remaining weapons were not sold to terrorist organizations

(Miller, et al. 2002, 254). Currently, the only known stores of smallpox virus

are contained in top-secret freezers at the CDC in Atlanta, Georgia and the

Vector laboratory in Koltsovo, Russia. It is unknown if any other nations

have secret stores of weaponized smallpox but, the CDC and the WHO have

published information suggesting that the possibility is real (Netys and

DeClercq 2003; Grais, et al. 2003; Katz 2002).

Smallpox research is also considered important in epidemiological

research concerning emerging infections diseases. Monkeypox outbreaks

among human populations in central Africa have concerned epidemiological

researchers because of the potential for a full-fledged species jump, similar

to the mutation that allowed HIV to enter human populations. Some

epidemiologists hope that studies of smallpox pathology can give some

insight into possible infection patterns and medical responses to these new

variants of monkeypox. Some researchers believe that this emergence of

new monkeypox variants is similar to the emergence of smallpox thousands

of years ago (Crosse 2003; Katz 2002; Ryan 1997).

The CDC and the WHO have also recommended smallpox

vaccination of military personnel in high-threat areas and medical personnel

in many areas (Crosse 2003; Katz 2002). Some researchers argue that any

history of smallpox vaccination could still protect people from the brunt of a

bioterrorist release of smallpox. Eichner (2003) has created statistical models

that credit vaccinations that are over 20 years old with a 79% residual

protection rate. However, most smallpox records show that the vaccinations

lose their effectiveness after 10 years (Crosse 2003; Baldwin 1999;

Craddock 1995; Willias 1994; Trobridge 1897; Barry 1889; Hainsworth

1889; Davidson 1889). That would leave the millions of people born since

the cessation of popular vaccination unprotected. In addition, models have

shown that a bioterrorist release of smallpox would have disastrous

consequences (Humphreys 2002; Meltzer, et al. 2001). If a bioterrorist event

infected 1,000 people with smallpox and no vaccination or quarantine

actions were taken, over

6,000 people would be infected in 30 days and over 447,000 people would

be infected in 90 days (Meltzer, et al. 2001, 963). Modern studies predict

mortality rates among unvaccinated populations between 30% and 40% and

many predict that a weaponized variant of smallpox would probably have

much higher mortality rates (Fenn 2001; Tucker 2001; Meltzer, et al. 2001).

The movement of smallpox through urban environments is unique and

has not been thoroughly examined since the development of GIS and spatial

analysis techniques (Curson 1985; Morrill and Angulo 1981, 1979; Angulo,

et al. 1980a; Angulo, et al. 1980b; Pyle 1973; Pyle and Rees 1971; Mack, et

al. 1970). Some researchers have begun research into historical smallpox

epidemics using GIS and spatial analysis techniques but, they have been

restricted to rural environments or regional studies (Eichner and Dietz 2003;

Williams; 1994; Wilson 1993). However, historic data sources that could be

used to explore epidemic diffusion in urban environments are being under-

utilized.

Geographers tend to focus on SIS analysis of modern health statistics instead

of exploring historical data (Ricketts 2003; Edsall 2003; Krieger 2003;

Kistemann, et al. 2002; Borroto and Martinez-Piedra 2001; Becker, et al.

1999: Gatrell 1995; Lam and Liu 1994). Research on the geographic

patterns of urban historical epidemics is necessary to understand how future

epidemics, either natural or intentional, move through modern urban

environments. James Nordin, a researcher at HealthPartners Research

Foundation and a member of the Harvard Consortium, discussed the

necessity for the exploration of historical data for use in new bioterrorism

detection systems.

It [the data] will be compared with historical

data to determine, for instance, if physicians in

Boston have seen patients exhibiting odd

symptoms for the time of year, the geography, and

the population, which could indicate early signs of

smallpox or anthrax. (McGee 2002, 26).

This thesis has begun the exploration of the unique manner in which

smallpox travels through urban environments. Barry (1889) noted that the

interactions between Sheffield’s citizens were numerous, frequent, and

irrespective of infection status. In contrast, Curson (1985) discussed the

extensive and restrictive quarantine methods imposed on Sydney’s

population. This variation in the level of interaction between infected and

susceptible population seems to be a key factor in the spread of smallpox.

The flow maps that track smallpox through the neighborhoods of Sydney

(Curson 1985) and the initial tracking of smallpox outbreaks in Sheffield

(Barry 1889) showed that hospitals, crowded house tenements, close

working conditions, and social interactions were critical in the urban

diffusion patterns of smallpox.

Future Research

This thesis is the first concentrated study of the 1887-8 smallpox

epidemic in Sheffield and only samples the potential of this data set. There

are several levels of additional research that could be performed using these

data. The possibilities under consideration include an expanded GIS with all

deaths from the epidemic, more detailed and sophisticated statistical and

spatial analyses, and investigations into the social aspects of the epidemic in

Sheffield.

The first step would be to expand the GIS to include the entire data set

of 590 deaths. The entire dataset was not used in this study because

contemporary street-level maps were not available. The Charles Goad Fire

Insurance maps for Sheffield were created in 1890 and updated in 1910.

These maps only covered the central city and allowed the mapping of only

80 deaths. Many of the deaths were in the outlying suburbs of Attercliffe and

Brightside and the more residential portions of the outer city core. The

Charles Goad maps did not cover the majority deaths as described by Barry,

including the cluster he noted at the Winter Street Hospital and in the

surrounding neighborhood (Barry 1889). Another set of maps needs to be

acquired, one that provides street-level coverage of the entire area

surrounding Sheffield, including the Ecclesall, Attercliffe, Nether Hallam,

and Brightside sub-districts during the late 19th century. There is no such

map set stored in the Sheffield City Archives. The Sheffield City Archive

workers have agreed to help look for such maps in collections held by

nearby universities and other town archives (Smith 2003).

The next step in the research on Barry’s text would be a more detailed

and complex statistical analysis. The expanded statistical analyses would

include three major areas: non-spatial, spatial, and animation. The non-

spatial techniques would be similar to the techniques that were described in

Chapter 3, but they would be re-run on the full dataset that will have been

plotted in the GIS. They would include the centrographic techniques, like the

mean and median. The next round of spatial techniques to be run on the data

that would be similar to those described in Chapter 3 but expanded to

include more detailed clustering analysis. The data would be refined to allow

for correct Knox and Mantel tests to be run for better estimation of the

space-time relationships between the deaths. In addition, diffusion models

could be run to extract space-time methods of spread.

The third layer of analysis attempted for the examination of the data

would be the creation of a time-series animation of the entire Sheffield

epidemic. This animation process would allow viewers to visualize the

epidemic’s progress through the borough of Sheffield. Also, one advantage

of using the GeoMedia program to create the GIS for this project would be

the ease of creating the animation as it is an option in the program

(Intergraph 2003). A time-series animation is a map series that shows how

variables change over time in a particular geographic area (Lobben 2003).

Dorling considered a time-series animation as a model “where the map is

held still and the action played out upon it” 1992, 218).

Another facet of the research on Barry’s text on the Sheffield

epidemic would be exploring the social aspects of the epidemic, similar to

the work done by Curson (1985) and Craddock (1995). Locating and

examining newspapers, notes from public meetings, and posted city notices

in the Sheffield City Archives and other locations would be the first step in

re-creating the atmosphere that permeated the city during the epidemic.

Barry incorporated some accounts of the city’s demeanor during the

epidemic in the text, including a letter to the editor of a local newspaper, The

Sheffield and Rotherham Independent, the rules of a few local smallpox

insurance associations, and the transcript of a Sheffield public meeting from

8 February 1888 (Barry 1889, 287-304). Perhaps some of the information

uncovered could be used to create flow maps similar to those in Curson’s

analysis of the 1881-2 Sydney smallpox epidemic (Figure 4.10).

Consultation with Curson, who is still teaching at Macquarie University in

Australia, on that matter will be conducted at a later date.

This data was purchased on Ebay (www.ebay.com), a popular public

auction website, and the source was largely unknown to the academic and

popular realms. Some researchers take a negative view of the private

purchase of data sources. Although these researchers see this “hoarding” of

data as being non-public minded, data sharing can be beneficial to all

researchers. Therefore, the data can be scanned into digital files and the

owners can grant researchers web-based access to share these rich resources.

Conclusion

Chapter 1 reviewed the literature on geography and disease, spatial

analyses of medical data, and studies of smallpox epidemics. Chapter 2

discussed the data source, Barry’s text, in detail, as well as some problems

with historical medical data. Chapter 3 discussed the creation of the GIS, the

spatial analyses used on the data, and the result of those analyses. Chapter 4

compared the results of this study to a study done by P.H. Curson on the

1881-1882 smallpox epidemic in Sydney, Australia, including comparisons

of the age structures of the epidemics and the cluster information. Chapter 5

discussed problems associated with historical epidemic data, the

applicability of this study to various academic and practical fields, and

further analyses that could be done on Barry’s text.

Barry included the following statement in the concluding remarks of

his text:

In the foregoing report it has been my object to

set out all the facts respecting the recent

epidemic in Sheffield, -- to give a complete

history of the behaviour there of the small-pox,

as well, a wholly unbiassed [sic] account of

every important condition capable of being

thought of as having influenced the disease,

whether in its extensions or its limitations. How

far this object has been attained, I must leave

others to judge.

At the same time I have abstained from any but

the broadest inferences, -- preferring rather that the

competent student of details should draw his own

conclusions from the abundant data thus placed at

his disposal, which data I have, I can truly say,

done my best to render absolutely trustworthy.

(Barry 1889, 294; all emphases are Dr. Barry’s).

It is this researcher’s belief that today’s research in medical geography and

epidemiology would meet with Barry’s heartiest approval. When Barry

wrote the above statement, it is not likely that he realized how long it would

take for his efforts to be analyzed and studied. However, this researcher

believes that the quality of his “abundant data” will not continue to be

unappreciated. By showing how important these data are, by detailing how

such data can be analyzed in a modern spatial environment and in creating

the first GIS of these data, this thesis has taken the first step to addressing

the goals set by Barry.

A spatial analysis of the smallpox epidemic in Sheffield, United Kingdom, 1887-1888

Students also viewed