Spanish Influenza in the City of Vancouver, British Columbia, 1918-1919
bySarah Buchanan
B.Sc., Queen‟s University, 2007
A Thesis Submitted in Partial Fulfillment of the Requirements for the Degree of
MASTER OF SCIENCE
in the Department of Geography
Sarah Buchanan, 2012 University of Victoria
All rights reserved. This thesis may not be reproduced in whole or in part, by photocopy or other means, without the permission of the author.
Supervisory Committee
Spanish Influenza in the City of Vancouver, British Columbia, 1918-1919
by
Sarah Buchanan
B.Sc., Queen‟s University, 2007
Supervisory Committee
Dr. Aleck Ostry, (Department of Geography, University of Victoria) Supervisor
Dr. Denise Cloutier, (Department of Geography, University of Victoria) Departmental Member
Dr. Mary-Ellen Kelm, (Department of History, Simon Fraser University) Additional Member
Abstract
Supervisory Committee Supervisor
Dr. Aleck Ostry, Department of Geography, University of Victoria Departmental Member
Dr. Denise Cloutier, Department of Geography, University of Victoria Additional Member
Dr. Mary-Ellen Kelm, Department of History, Simon Fraser University
During the last year of World War I (1918), a second deadly foe was causing
mortality around the world. Spanish Influenza killed an estimated 50-100 million people
worldwide, including 50,000 people in Canada during the 1918-1919 pandemic. This
thesis examines the impact of Spanish Influenza on people living in Vancouver, British
Columbia, Canada between June of 1918 and June of 1919. Statistical analysis with SPSS
was used to determine the association between influenza-caused deaths and
socio-demographic characteristics such as age, gender, immigration status, and employment. In
Vancouver, those who were between the ages of 19 to 39, and those who were employed,
Table of Contents
Supervisory Committee ...ii
Abstract... iii
Table of Contents ... iv
List of Tables ... vi
List of Figures ...vii
Acknowledgments ... viii
Dedication ... ix
Chapter 1: Introduction ... 1
1.1 Purpose/Objectives ... 1
Chapter 2: Literature Review ... 4
2.1 Understanding Spanish Influenza through the Lens of Medical Geography ... 4
2.2 Spanish Influenza around the World ... 5
2.3 Public Health and the Response to Spanish Influenza ... 7
2.4 Origins, Timing, and Mortality from the Epidemic ... 10
2.5 Groups Most Vulnerable to the Epidemic ... 20
2.6 Spanish Influenza in British Columbia and Vancouver ... 21
2.7 Summary ... 23
Chapter 3: Methods ... 25
3.1 Data Collection and Determining Population Size ... 25
3.2 Boundary of Analysis ... 26
3.3 Calculating Mortality Rates ... 27
3.4 Managing the Association between Pneumonia and Influenza ... 28
3.5 Analysis using the Statistics Package for Social Sciences (SPSS) ... 28
3.6 Visual Representation ... 30
3.7 Limitations ... 31
Chapter 4: Results ... 34
4.1 Boundaries ... 34
4.2 Case Definition of Influenza Deaths... 35
4.3 Socio-demographic description of influenza versus non-influenza deaths in both time periods ... 38
4.4 Temporality of Influenza Deaths within Vancouver ... 46
4.5 Distribution of Influenza Deaths in Vancouver ... 47
4.6 Mortality Rates in the City of Vancouver ... 49
4.7 Statistical Analysis of Links between social and geographic characteristics of residents and influenza deaths in Vancouver ... 51
4.7.1. Univariate analyses... 51
4.7.2 Multivariate Analysis ... 58
Chapter 5: Discussion ... 61
5.1 Timing and Intensity ... 62
5.2 Spatial Distribution of Influenza Deaths in Vancouver ... 63
5.3 Socio-Demographic Characteristics of Influenza in Vancouver ... 64
5.4 Mortality Rates for Spanish Influenza in Vancouver ... 65
5.5 Contributions to Medical Geography ... 66
Chapter 6: Conclusion ... 69
Bibliography ... 71
Appendix A: Codebook ... 78
A.1 Coding for Cause of Death and Secondary Cause of Death ... 78
A.2 Coding for Occupation ... 115
A.3 Coding for Birthplace ... 131
A.4 Coding for Place of Death ... 135
A.5 Coding for Residence ... 135
A.6 Coding for Immigrant Status ... 135
A.7 Coding for Race ... 136
Appendix B: Immigrant Proportions ... 137
List of Tables
Table 1. Overview of Canadian provincial and city mortality rates from the literature. ... 15
Table 2. Primary cause of death in 1918-19 and 1921-22. ... 36
Table 3. Secondary cause of death for 1918-19 and 1921-22. ... 37
Table 4. Influenza versus non influenza deaths by time period. ... 38
Table 5. Influenza versus non influenza deaths by residence and time period. ... 38
Table 6. Influenza versus non influenza deaths by gender and time period. ... 39
Table 7. Influenza versus non influenza deaths by age category and time period. ... 40
Table 8. Influenza versus non influenza deaths by place of death and time period... 41
Table 9. Influenza versus non influenza deaths by birthplace and time period. ... 42
Table 10. Influenza versus non influenza deaths by employment status and time period. 42 Table 11. Influenza versus non influenza deaths by employed occupation and time period. ... 43
Table 12. Influenza versus non influenza deaths by marital status and time period... 44
Table 13. Influenza versus non influenza deaths by race and time period. ... 44
Table 14. Influenza versus non influenza deaths by simple immigration status and time period. ... 45
Table 15. Influenza versus non influenza deaths by length of residence in Canada and time period. ... 45
Table 16. Mortality Rates for Gender, Age Group, Immigrants, and the Employed, 1918-1919. ... 50
Table 17. Chi-square test for gender and influenza deaths in 1918-1919. ... 51
Table 18. Chi-square test for gender and influenza deaths in 1921-1922. ... 51
Table 19. Chi-square test for age category and influenza for 1918-19 and 1921-22. ... 52
Table 20. Chi-square test for place of death and influenza in 1918-19 and 1921-22. ... 53
Table 21. Chi-square test for birthplace and influenza in 1918-19 and 1921-22. ... 53
Table 22. Chi-square test for employment status and influenza in 1918-19. ... 54
Table 23. Chi-square test for employment status and influenza 1921-22. ... 54
Table 24. Chi-square test for occupation and influenza in 1918-19 and 1921-22. ... 55
Table 25. Chi-square test for marital status and influenza in 1918-19 and 1921-22. ... 55
Table 26. Chi-square test for immigration status and influenza in 1918-19. ... 56
Table 27. Chi-square test for immigration status and influenza in 1921-22. ... 56
Table 28. Chi-square test for race and influenza in 1918-19. ... 57
Table 29. Chi-square test for race and influenza in 1921-22. ... 57
Table 30. Chi-square test for length of residence in Canada and influenza in 1918-19 and 1921-22. ... 58
Table 31. Correlation between age, gender, birthplace, immigration status, length of residence in Canada, and employment status for Vancouver residents in 1918-19. ... 59
Table 32. Multivariate analysis showing odds ratio‟s for age category, marital status, gender, and employment status in 1918-19. ... 60
List of Figures
Figure 1. Crowd outside of Hotel Vancouver, November 1918. ... 7
Figure 2. The SEF leaving the Port of Vancouver. ... 14
Figure 3. Vancouver, 1915. ... 22
Figure 4. City of Vancouver boundary in 1915... 35
Figure 5. Influenza versus non influenza deaths by time period.. ... 47 Figure 6. Location of residence for Vancouver residents dying of influenza in 1918-19. 49
Acknowledgments
First and foremost, I would like to extend my heartfelt thanks to my supervisor Aleck
Ostry for his guidance and encouragement throughout this project. To my committee
members, Denise Cloutier and Mary-Ellen Kelm, thank you for your time, feedback, and
support toward this work; I have truly appreciated your perspectives. Additionally, a big
thank you to Ashley LeBourveau who so carefully transcribed the death certificates from
microfiche into Excel for me. And finally, Patrick Lucey and Cori Barraclough, thank
you for your total support and keen interest in all that I do and for teaching me to see the
Dedication
I dedicate this thesis to my family for their never-ending love, support, encouragement,
Chapter 1: Introduction
From the spring of 1918 until the spring of 1919 the world was under siege from a deadly
influenza virus known as Spanish Influenza. As an H1N1 influenza A strain, the virus caused “severe infection in the upper and lower respiratory tract, resulting in fatal respiratory
complications and bacterial pneumonia” (Pica et al., 2010, p. 125). During this period, the
pandemic killed an estimated 50 to 100 million people worldwide (Johnson and Mueller, 2002).
In Canada, approximately 50,000 people died from Spanish Influenza resulting in a mortality
rate of 6.1 per 1,000 people (Fahrni, 2004; Jenkins, 2007; Johnson and Mueller, 2002; Jones,
2005; Pettigrew, 1983; Patterson and Pyle, 1991). The social consequences of the pandemic were
severe and altered the way in which public health was understood.
While there is a plethora of historical research on the Spanish Influenza around the world,
there is little research that takes a geographical and epidemiological approach, especially in Canada. Furthermore, very little research has been conducted on the influenza‟s effect on cities
in British Columbia. This research sought to fill this knowledge gap by taking a medical
geography approach in conjunction with epidemiological methods to characterize the timing,
extent, and socio-demographic determinants of influenza mortality in the City of Vancouver,
British Columbia, Canada.
1.1 Purpose/Objectives
The purpose of this research was to conduct an in-depth geographical and
epidemiological analysis of Spanish Influenza in Vancouver, British Columbia, during the
pandemic of 1918-1919 and identify its demographic and geographical characteristics. One
investigation of Spanish Influenza in Vancouver, British Columbia, estimated, using newspaper
than the Canadian average of 6.1 per 1,000 (Andrews, 1977). This same rate is quoted by Kelm (1999) and O‟Keefe and MacDonald (2004). However, in all these publications, it is unclear
whether this is a maximum or an average rate and the methods are not detailed enough to
determine how these rates were calculated. Additionally, there is no indication as to which
populations were most affected or whether socio-economic status or location (separately or
interactively) played a role in either contracting the disease or dying from this virulent strain of
influenza. This research was based on detailed and comprehensive analysis of all Vancouver
deaths from influenza during the epidemic and two years after the epidemic. The study sought to
determine the timing of the epidemic, clearly define mortality rates and demographic/population
risk factors for contracting Spanish Influenza. As well, adoption of a “case” and a “control”
period permitted a comparison of the characteristics of the pandemic influenza with the more
normal influenza season, which occurred in 1921-1922.
Based on the literature, six categories of questions were used to guide the analysis of Spanish
Influenza in Vancouver for 1918-1919. These questions are:
1. Did the first wave of Spanish Influenza in Vancouver occur in the Fall of 1918 and how many waves of the infection occurred?
2. Was the Fall 1918 wave of influenza more deadly than the other waves?
3. Were the effects of influenza consistent across all areas of Vancouver or did some locations experience a more severe mortality rate?
4. Were employed people more likely to die of Spanish Influenza? Were there specific occupations that experienced higher numbers of deaths?
5. Were young adults in Vancouver more likely to die of influenza than other age groups? 6. Did foreign born populations experience higher mortality rates than Canadian born
citizens?
In order to answer these questions, the thesis is organized into five sections. Chapter two, the
literature review, grounds this thesis in medical geography and highlights the story of the
of influenza in the United States and Canada. Chapter three explains the methods used to answer
the research questions while chapter four reveals the results of the analysis. Finally, chapter five
(Discussion) and chapter six (Conclusions) link the results back to the specific research questions
Chapter 2: Literature Review
2.1 Understanding Spanish Influenza through the Lens of Medical Geography Medical geography is defined by the Encyclopedia of Geography as the “application of
geographical methods and techniques to the study of health” (Jordan, 2012, p.1). While the term
medical geography suggests a relationship to medicine, Mayer (2010) points out that this may be
a misnomer due to the fact that the research has little to do with diagnoses and treatment but is
more of an examination of the geography of disease and health service delivery with an
epidemiological view point. Medical geography first arose in 1952 with the production of a
report by the Commission on Medical Geography (Ecology) of Health and Disease to the
International Geographic Union (Meade and Earickson, 2000), with a focus on disease ecology,
disease distribution, and health care service delivery and provision. However, as Meade and
Emch (2010) indicate, medical geography goes beyond just the simple presence of disease and
also allows for the inclusion of social, political, and economic factors that influence the
susceptibility of a population to disease and overall population health in general.
One of the key elements of medical geography is its biomedical perspective, a way of
viewing issues of health using biological mechanisms of illness and disease (Moon, 2009). The
idea of a biomedical gaze is attributed to the works of Michael Foucault, a philosopher of history
(or historical philosopher) who has written extensively on theoretical/methodological approaches
related to the medical field (Bishop, 2009). The biomedical gaze involves
examining/understanding a person based on their illness and the way that illness is biologically
manifested within their body (Frank and Jones, 2003; Hick, 1999).
Consequently, the major critique of medical geography is not the research focus itself but
Elliot, 2009). What is interesting to note is that Meade and Earickson (2000) in the preface to
their medical geography textbook specifically state that medical geography has become “less
concerned with the optimization of health service delivery or a dichotomy between health service
and disease ecology” in favour of becoming more “concerned with health geography as a
behavioral and social construction and disease ecology as an interface between the natural
(physical world) and cultural dimensions of existence” (Meade and Earickson, 2000, p.v). This
advancement in the field of medical geography has allowed, as part of this research, the inclusion
of socio-demographic characteristics that help contextualize the quantitative results.
Furthermore, in their identification of questions of medical geography, Meade and Earickson
(2000) include a question regarding place which is typically associated with health geography as one of that discipline‟s central themes rather than a key element of medical geography. This
suggests that medical geographers are distinctly aware of the prominent critique in their field of
being dehumanizing and reductionist and that this is an evolving methodology moving toward a
more broadly conceptualized health geography.
The framework of medical geography forms the basis of this research because it focuses
on disease and illness (factors of ill-health) rather than health promotion and incorporates the use
of quantitative analysis of disease. Due to the nature of the data being used to undertake this
research, quantitative analysis is necessary as death certificates do not provide the opportunity
for interviews nor qualitative analysis. Consequently, a medical geography approach was taken
as opposed to a health geography approach.
2.2 Spanish Influenza around the World
During the last year of World War One (1918), the world experienced a horrendous death
referred to as the Spanish Flu. While the Spanish Influenza did not originate in Spain, it is
believed by many authors that neutral Spain was the first to publicize its emergence because of
less wartime censorship (Canadian Medical Association Journal, 1918; Rogers, 1968; Dickin
McGinnis, 1981; Kreiser, 2006).
Between the spring of 1918 and the following spring of 1919, Spanish Influenza spread
around the world killing millions of people. First estimates of the worldwide death toll in the
1920s suggested that 20 million people died (Jordan, 1927). However, this mortality estimate has
increased over time. For example, Patterson and Pyle (1991) estimated 24.7-39.3 million people
died and, more recently, Johnson and Mueller (2002) estimated somewhere from 50 to 100
million may have died in the pandemic. Regardless of which estimate is deemed reliable, as
Tuckel et al. (2006) point out, this worldwide pandemic was unique as it killed millions of
people in a very short time altering the way in which public health was viewed and understood.
For the purposes of this literature review, only material from the United States of
America and Canada were referenced in support of specific research to determine the effect of
Spanish Influenza on people living in Vancouver, British Columbia. The global mortality
highlights the effect Spanish Influenza had during the period between the spring of 1918 and the
spring of 1919. While the focus of this literature review was Canada, and to a lesser extent, the
United States, Spanish Influenza was not unique to North America, nor were mortality rates the
highest in this part of the world (Canada 6.1 per 1,000; USA 6.5 per 1,000; Mexico 20.6 per
1,000) (Johnson and Mueller, 2002). In fact, the most deaths attributed to this disease occurred in
India with a recorded estimate of 18 million deaths (6.1 per 1,000) while the highest morality
rates (between 10.7 and 445 per 1,000) were likely experienced in Africa (Johnson and Mueller,
2.3 Public Health and the Response to Spanish Influenza
Beyond discussions of mortality rates and death tolls, specific themes relating to this
epidemic appear and reappear throughout the Canadian and US literature, regardless of the
location being examined. These themes include greater emphasis on waging war than on dealing
with the public health problems created by the pandemic; the ineffectiveness of quarantine; a
lack of hospital beds, equipment, and personnel; the use of influenza fears to market so-called
curative or preventative treatments; and, finally, the mobilization of ordinary citizens into
volunteer organizations to take care of the sick.
Andrews (1977), Byerly (2005),
and Kreiser (2006) point out that despite
warnings about public gatherings, people
still congregated in city streets to welcome
soldiers home from the war. These authors
claim that these public gatherings resulted
in a spike in influenza cases in the days
following. Furthermore, Kreiser (2006)
calls attention to the fact that the desire to
provide people and supplies to the war
effort precluded precautions, such as
quarantine, from being implemented in the early days of the epidemic. Hence, activities
specifically related to the war effort were not hampered by quarantine protocols leading to the
conclusion that quarantine measures were not implemented effectively. The inability to
effectively implement quarantine measures is also illustrated by the Toronto experience, whereby
Figure 1. Crowd outside of Hotel Vancouver, November 1918. Thomson, Stuart. (Photographer). Armistice Day crowd outside Hotel Vancouver, Georgia Street. [Online image]. Retrieved from City of Vancouver Archives, http://searcharchives.vancouver.ca/armistice-day-crowd-outside-hotel-vancouver-georgia-street;rad.
the disease spread so quickly to so many people that quarantine and isolation became impractical
almost immediately (MacDougall, 2007).
Additionally, hospitals were badly understaffed during the war and communities had to
compensate by opening other buildings such as schools and hotels to care for the sick (Boucher,
1918; Dickin McGinnis, 1976; Andrews, 1977; Jones, 2005; MacDougall, 2007). As well,
personnel were difficult to find with large numbers of citizens overseas, and provisioning was
more difficult because the military had first access to materials. Finally, many medical
professionals became ill with the flu further straining resources for caring for the sick (Dickin
McGinnis, 1981).
As a result of the widespread concern about contracting Spanish Influenza, many
businesses and other organizations cashed in on the panic to promote and market products as
essential flu prevention and protection techniques (Andrews, 1977; Jenkins, 2007). Pettigrew
(1983) spends a whole chapter describing these different products. Some of the more interesting
remedies included: lard mixed with turpentine, gin pills, goose grease, camphor, sulphur, and oil
of cinnamon (Pettigrew, 1983; Jones, 2007). However, none of these products proved to be
consistently effective remedies.
One of the most impressive feats throughout the epidemic was the mobilization of
volunteers and organizations to care for the sick. The lack of medical personnel resulting from
the war effort was well recognized and entire communities responded to supply the labour and
care necessary to help those who were ill in the community. Dickin McGinnis (1976, p. 7)
undertook a study of Calgary, Alberta, and noted that volunteering to help with influenza was viewed as an opportunity for “women to do their bit for war and for civilization”. These women
also reflected by Jones‟ work on influenza in Winnipeg (Jones, 2002, 2005, 2006, 2007). For
more information see Jones (2007) who examines in her book the role of women and
volunteering in aiding in the care of those ill with influenza.
So what made Spanish Influenza so deadly and unique? Researchers such as Herring
(1993) and Kelm (1999) hypothesize that part of its virulence was because it was a virgin soil
epidemic so no one had immunity to it. Others postulate that it was not necessarily the flu itself
that was deadly but the secondary pneumonia infection that caused fatality (Canadian Medical
Association Journal, 1918; MacDougall, 2007; Robertson, 1919). Perhaps, as Dickin McGinnis
(1981) suggests, the conditions created by the war, the transportation of large numbers of people
in a short amount of time, and the dismal conditions of the war trenches produced an
environment ripe for the influenza virus to thrive. Additionally, it could have been that the
medical world was not clear on what the cause of influenza was. The Spanish Influenza strain
was not identified until 1933 by British scientists (MacDougall, 2007; Noymer and Garenne,
2000) and is still being analyzed genetically by researchers such as Taubenberger today. The
genetic analysis of the influenza virus gathered from the frozen lung tissue of victims of
influenza indicates that the 1918 virus was an H1N1-subtype influenza A virus (Taubenberger,
2003). Taubenberger (2003 and 2006) further classifies this Spanish Influenza virus as
containing genes from avian-influenza strains and suggests that pigs may have acted as the
intermediary host between birds and humans. Furthermore, the flu is an illness that consistently
infects populations, often with high morbidity but low mortality so it may not have been taken
seriously at the outset; i.e.an attitude of “after all, it‟s only the flu” may have set in (Kreiser,
2.4 Origins, Timing, and Mortality from the Epidemic
There is much debate within the literature regarding where Spanish Influenza originated.
Some researchers suggest that the flu began in China in February 1918 and was brought to
Western Europe by Chinese labourers before being disseminated by troops fighting throughout
Europe and, from there, back to soldiers‟ countries of origin (Dickin McGinnis, 1981; Pettigrew,
1983). The virus requires an intermediate host, such as pigs, to be transmitted from the avian
population to the human population. The close proximity between the natural influenza reservoir
of birds, the intermediate hosts (pigs), and humans in Asia has resulted in that area being
considered an influenza epicentre (Shortridge and Stuart-Harris, 1982; Webby and Webster,
2001). Consequently, China as the origin makes sense biologically.
Now consider how the virus could have been transmitted from China to the western front.
A recently published book by Guoqi (2011) documents the journey of upwards of 80,000
Chinese labourers across the Pacific Ocean to Vancouver. From Vancouver they boarded onto
railway cars being transported across Canada to the eastern ports for passage to the Western
front. This transportation of Chinese labourers occurred between March 1917 and March of 1918
and was considered to be of utmost secrecy with a complete blackout of the media. At the time, a
$500 head tax was required from any Chinese person entering Canada; however, because these
labourers were enroute to Europe, the Government of Canada waived the tax on the condition the
labourers did not get off the trains (Guoqi, 2011). Not all of the 140,000 – 200,000 Chinese
labourers, sought to address manpower shortages in Western Europe, travelled through Canada
on their way to France. King (1918) indicates that some of them were sent through the
Mediterranean although the majority are said to have passed through Canada and the United
States. Whatever the route, perhaps some of these labourers harboured a precursor to the deadly
Another hypothesis of the origin of Spanish Influenza comes from Oxford (2001) who
suggests that an acute respiratory infection at Etaples, France, in the winter of 1916 was an
influenza virus of a type similar to the 1918 flu. However, most authors suggest that this strain of
influenza actually originated in the United States in Haskell County and spread to the Camp
Funston army base in north central Kansas where the illness was first reported on March 5, 1918
(Vaughan, 1921; Patterson and Pyle, 1991; Crosby 2003; Barry, 2004). From there it was carried
by troops transported by ships to serve on the European battlefields and then to the rest of the
world. However, Erkoreka (2009, p.192) cautions that it is “problematic to assign such a specific
date to the beginnings of the pandemic, since its origins are likely to be much more complex and varied”. This statement leaves room for continued debate on where the Spanish Influenza strain
first developed.
Whether Spanish Influenza originated in Kansas or not, the infection‟s presence there
was a factor in the dissemination of the virus and kicked off the first of three influenza waves
beginning in the spring of 1918, the fall of 1918, and the winter of 1918-1919 (Crosby, 2003).
By June of 1919, 675,000 people had perished in the United States as a result of Spanish
Influenza that, according to Kreiser (2006), is more than the total number of American military
deaths incurred during World War One, World War Two, the Korean War, and the Vietnam War
combined.
Part of the reason for this high death toll, and the speed with which the epidemic spread
throughout the United States, is attributed to the efforts of supplying as many soldiers as possible
to the battlefields in Europe, which ruled out a public health approach of complete quarantine.
Crosby (2003, p. 56) indicates that the American armed services (Army and Navy) were affected “earlier and more severely than the civilian population and to a certain extent were the foci from
which the civilian population received the disease”. Furthermore, Byerly (2005) and Kreiser
(2006) propose that quarantine (the only public health measure that could be used to effectively
combat the epidemic) would have counteracted efforts to rapidly deploy US forces in Europe.
The war was deemed a more important task and troops were dispersed despite illness among the
ranks.
Overall, the US mortality rate directly attributable to Spanish Influenza between 1918
and 1919 is 6.5 deaths per 1,000 people (Johnson and Mueller, 2002). However, as we will see
for Canada, this rate was not consistent across the country. Rogers (1968) identifies mortality
rates for five cities: Philadelphia (158 per 1,000), Baltimore (109 per 1,000), Washington (109
per 1,000), Boston (100 per 1,000), and New York (60 per 1,000). These rates suggest that while
the overall mortality rate for the United States may have been relatively low, certain centers were
more severely impacted. This is also reflected in a study conducted by Noymer and Garenne
(2000) who reveal that in the United States the life expectancy at birth dropped by 11.8 years in
1918 showing the disproportionate impact of the epidemic on young people.
Influenza did not reach Canada until late summer/early fall of 1918; although Dickin
McGinnis suggests it could have arrived earlier in June or July 1918, her hypothesis is refuted by
Humphries (2005). Therefore, it appears that the first wave of the pandemic did not involve
Canada directly. Despite this, the second wave was still destructive to Canadian populations
when it did arrive. The most commonly cited estimate for the Canadian death toll caused by
Spanish Influenza and the pneumonia that accompanied it is 50,000 people out of a total
population of about eight million, a mortality rate of approximately 6.1 deaths per 1,000 people
(Fahrni, 2004; Jenkins, 2007; Johnson and Mueller, 2002; Jones, 2005; Pettigrew, 1983;
While this mortality rate of 6.1 per 1,000 is consistently documented, it is unclear when
Spanish Influenza first arrived in Canada. This is because of an unresolved debate about how the
epidemic arrived and spread throughout the country. Early researchers such as Dickin McGinnis
(1976, 1981), Pettigrew (1983), and early works of Jones (2002), support the belief that the
epidemic originated with Canadian soldiers returning from Europe on troop ships. The broader
implication of this origin is that American troops infected in the first wave went to Europe where
they infected Canadian troops who then brought the illness to Canada.
However, Humphries (2005) makes a strong case, based on military archives and ship
records, that ships thought to have brought ill soldiers back from Europe were actually sitting
empty in quarantine waiting to be sent to Europe. Thus, these ships could not have brought
Spanish Influenza to Canada. Instead, he presents a new hypothesis that the “physical path taken
by pandemic influenza in Canada was determined by the intensification of the war effort, not by its waning” (Humphries, 2005, p. 241). In short, he theorizes that Spanish Influenza entered
Canada in late August 1918 with newly trained American recruits who were to be sent overseas
via Canada. The “greatest single factor in the diffusion of the disease” according to Humphries
(2005, p.252) was the creation of the Siberian Expedition Force, which was to travel to Russia
from Vancouver and included both American and Canadian troops. This expeditionary force was
recruited during late August and early September of 1918, around the same time that the first
influenza cases appeared in Ontario, Quebec, and Nova Scotia. This force was assembled in
Eastern Canada and transported via rail to Vancouver on October 2, 1918 (Humphries, 2005).
However, according to Humphries (2005), and corroborated by Jones (2007), the train
drop off military members sick with influenza to the local hospitals. Accordingly, this railway
expeditionary force seeded the epidemic across Western Canada in the early fall of 1918.
The theory of an August 1918 arrival of influenza in Canada is in conflict with research
conducted by Palmer et al. (2007) who show that influenza was present in Newfoundland as
early as May 1918. Thus, it is possible that influenza entered Atlantic and Eastern Canada a few
months prior to August 1918; however, the train
journey of September 1918 was the primary route that
seeded the influenza virus into the west.
As influenza made its way across Canada,
different localities had dissimilar experiences with the
illness. Some areas such as Newfoundland (Palmer et
al., 2007) experienced two long waves of the illness (May – July 1918, September 1918 – June 1919) while
other locations like Winnipeg experienced two shorter
waves (October – December 1918 and February –
April 1919) (Jones, 2005). This discrepancy implies a
geographical determination of the character and timing of the illness as it could have been
shaped by different regional attributes such as the make-up of the social organization in the
cities, towns, and communities‟ Spanish Influenza affected.
Such local or regional heterogeneity appears to also be reflected in mortality rates across
Canada. As shown in Table 1, rates varied from a low of 3.6 per 1,000 (in Ontario) to a high of
780 per 1,000 in one small town in Newfoundland). Why certain regions experienced higher
mortality is unclear as the methods of prevention were similar across Canada with quarantine,
Figure 2. The SEF leaving the Port of
Vancouver. Thomson, Stuart. (Photographer). “Monteagle” leaving Vancouver with Siberian Expeditionary Force. [Online image]. Retrieved from City of Vancouver Archives,
http://searcharchives.vancouver.ca/monteagle- leaving-vancouver-with-siberian-expeditionary-forces;rad.
reduction of business hours, closure of theatres, picture shows, and schools, sanitation agendas,
and provision of vaccines being the common medical and public health practices (Andrews,
1977; Jenkins, 2007; Jones, 2006; McCullough, 1918; Dickin McGinnis, 1976).
Table 1. Overview of Canadian provincial and city mortality rates from the literature.
City Mortality Rate (per 1,000) Source
Newfoundland (Nfld) 5.0 Palmer et al., 2007
New Brunswick 4.0 Jenkins, 2007
Ontario (ON) 3.6 Jenkins, 2007
Quebec 7.0 Jenkins, 2007
Winnipeg, Manitoba 6.0 to 6.5 Jones, 2006
South Winnipeg 4.0 Jones, 2005
North Winnipeg 6.7 Jones, 2005
Saskatchewan cities 6.0 Lux, 1997
Rural Saskatchewan > 10 Lux, 1997
Alberta 11.5 Jenkins, 2007
British Columbia: non-native population
6.2 Kelm, 1999
Vancouver, British Columbia 23.3 Andrews, 1977
Aboriginal Communities
Hebron, Nfld 680 Palmer et al., 2007
Okak, Nfld 780 Palmer et al., 2007
Norway House, Manitoba 183 Herring and Sattenspiel, 2003
British Columbia: native population
One of the difficulties in understanding these marked differences in rates is that the
methods used to obtain them are not clearly stated. Most of these rates appear to have been
derived from the media (especially newspapers). It is therefore impossible to determine their and
to assess the suitability of the methods used to obtain them.
For example, Jenkins (2007), while providing mortality rates for New Brunswick,
Ontario, Quebec, Saskatchewan and Alberta, is unclear which populations were used as the
denominator in the rate calculation. Jenkins extracts the total number of deaths from influenza
from various other authors, and then (it appears) calculates rates using the quoted numbers as
numerators. For example, the total number of influenza deaths for Quebec was extracted from
Patterson and Pyle (1991). Jenkins then calculated a rate but does not say which denominator
was utilized; therefore, this calculation cannot be repeated and its accuracy is unknown.
Lux (1997) provides an overview of the impact of Spanish Influenza on Saskatchewan
cities and rural regions. Influenza rates were lower in urban than in rural Saskatchewan,
however, as with many of these historical accounts of influenza, the methods used to calculate
rates are not explicit. The only clue provided by Lux is that the mortality data were based on
information from the Saskatchewan Board of Public Health.
A similar conundrum is presented with the work of Jones (2005 and 2006). Rates for
Winnipeg are extracted from City Health Department records. While these may be good
secondary sources, it is not indicated whether the mortality data obtained was raw, with rates
calculated by Jones, or whether the rates were obtained directly from the City Hall records.
Either way, it is unclear how the rates were determined.
While methods were not explicit for Jones and Lux, they were based on “official” health
than 10 in rural Saskatchewan. These studies have the advantage of being based on official data
rather, than in some of the other studies quoted in Table 1, on media reports and appear to put a
range on rates in the prairies in 1918-19 of 4 to 10 deaths per 1,000 population, much lower than
the extremely high rates of 780 per 1,000 observed in one Newfoundland study.
There are very few studies that undertake an examination of the impact of Spanish
Influenza in British Columbia. Two commonly cited papers discussing the influenza epidemic in
British Columbia were written by Andrews (1977) and Kelm (1999). Andrews (1977),
specifically examines influenza in the City of Vancouver. While she specifically acknowledges “precise morbidity and mortality figures are not available for Vancouver for the six-month
epidemic period” she does provide a mortality rate of 23.3 per 1,000 (Andrews, 1977, p. 27).
This rate is based upon information provided by the Vancouver Daily Province newspaper and
the 1918 annual report for the City of Vancouver. The period of the epidemic this rate represents
is from the beginning of the epidemic to the end of January 1919. Unfortunately, it is not clear
when the beginning of the epidemic is, as numerous cities are included within the same
timeframe but not all cities were infected at the same time. Nor is it clear whether Andrews
herself calculated the rate or if it was provided by the newspaper or annual report. While the
mortality rate of 23.3 per 1,000 has been quoted in other historical discussions of Spanish
Influenza, its accuracy is unclear. My research will, by using death certificates and clear
denominators, be able to calculate the mortality rate associated with Spanish Influenza for the
City of Vancouver with greater accuracy.
Kelm (1999) quotes an overall B.C. mortality rate of 6.21 per 1,000 (based on Vital
Statistics Sessional papers for May of 1919) and the same Vancouver mortality rate of 23.3 per
population was 46 per 1,000 implying that Aboriginals suffered much more severely than the
non-Aboriginal population. The Aboriginal mortality rate is based on Department of Indian
Affairs and Vital Statistics population information at the band level and mortality from
flu-related illness (Kelm, (1999); Kelm, 2011, pers. comm.). It is not clear from Kelm‟s article what
year the population totals represent or what mortality figures were used (Vital Statistics versus
Department of Indian Affairs) as part of the rate calculation.
One of Kelm‟s key findings is that, among Aboriginals in BC, young children and older
adults were most affected whereas mortality from Spanish flu appears to have been greater
among young adults in the non-Aboriginal populations. One reason for these differences may be that at this time “one-third of the total population of British Columbia‟s First Nations were under
the age of fifteen” (Kelm, 1999, p. 30). However, that does not explain the greater impact among
older Aboriginal adults versus older non-Aboriginal adults she observed.
The studies that do clearly identify how the rates were calculated are Palmer et al. (2007)
and Herring and Sattenspiel (2003). Palmer et al. (2007) used 1911 census population data to
calculate death rates per 10,000 for districts across Newfoundland by wave of influenza
infection. Sources of information to determine mortality rates included death records, interviews
with island residents, census data, vital statistics, records of international shipping, newspapers,
local museums, and government correspondence about the influenza epidemic. Accordingly, the
overall mortality rate for Newfoundland came from a fairly rigorous methodological
investigation. However, the rates provided for Hebron and Okak (small towns in Newfoundland)
were not calculations conducted by the authors but sourced elsewhere. Consequently, the
Herring and Sattenspiel (2003) undertook a slightly different approach to determining
mortality rates for the isolated, primarily Aboriginal community of Norway House in Manitoba.
The authors used parish records to estimate the mortality rate of Norway House (183 per 1,000)
and performed a more detailed analysis using Treaty Annuity Pay Lists for population estimates
to examine family distributions of mortality. The analysis of family distribution reveals that
while there is a high community mortality rate, the impact of influenza was concentrated within seven nuclear families (25% of the deaths) and within “families related through the male line”
(56% of the deaths) (Herring and Sattenspiel, 2003, p. 169). Therefore, the high mortality rate is
more of a result of the impact on these particular families than on the impact of influenza on the
community as a whole.
They also conducted a travel-based analysis using Hudson‟s Bay Company journals to
establish the travel patterns between Norway House and other trading posts where influenza was
contracted. These journals indicate that travel occurred more frequently during the warmer
months when waterways were open, to accommodate travel by water, while in winter months
travel was less frequent, and took longer, which delayed the dissemination of influenza into the
community. However, after running computer simulations on the movement of influenza
between communities, in the vicinity of Norway House, the author‟s note that while travel (i.e.
rate of travel and patterns of travel) influenced the timing of influenza infection it did not affect
the number of cases experienced by the community. The study did not provide seasonal mortality
rates so there is no discussion on how mortality rates were influenced by travel patterns.
My research follows a similar method, in determining mortality rates, to that of Palmer et
al. (2007) by also utilizing census populations, census information, and death records. Death certificates were used as the primary source of information to identify the number of deaths
caused by Spanish Influenza in Vancouver. The key difference is that this study will utilize the
census population from 1921, rather than from 1911, as it is closer in time to the period of the
epidemic (1918-1919) so the population numbers should be more reflective of 1918 – 1919 than
the 1911 census numbers.
2.5 Groups Most Vulnerable to the Epidemic
Perhaps the most commonly cited unique feature of this strain of influenza is that it
resulted in high mortality rates among young adults. Noymer and Garenne (2000) and Palmer et
al. (2007) discuss the W-shaped age-specific mortality curve where high mortality rates are experienced in the youngest and oldest populations, as well as a peak between the ages of 15 to
45. Additionally, virtually all authors cited in this literature review indicate at some point within
their research that young adults were most affected. The use of the term „young adult‟ in these
studies is not consistent. For example, Jones (2006) identifies that 60% of all deaths in the city of
Winnipeg occurred in people between the ages of 20 to 39. On the other hand, Humphries (2005)
suggests that the majority of deaths in Canada were healthy males and females between the ages
of 15 and 35. Tuckel et al. (2006) in their analysis of Hartford, Connecticut indicate that there
was a higher mortality rate among males and for young adults between the ages of 25 to 35
years.
In addition to the age-specific population effects, there is some discussion within the
literature that socio-economic status may have played a role. This suggestion is primarily based
on observed differences in mortality rates between immigrant populations and non-immigrant
populations in American cities. The most detailed examination of this population difference is
conducted by Tuckel et al. (2006) in their study of Hartford, Connecticut. They demonstrated
nativity and higher among those with Canadian, Russian, Austrian, and Polish maternal nativity.
Those people with Italian maternal nativity experienced the highest mortality rates in Hartford.
Tuckel et al. (2006) concluded that a higher proportion of immigrants, renters, and the less
educated tended to have higher mortality rates in Hartford. Accordingly, this suggests that
socio-economic status may indeed have been a risk factor for death caused by influenza in 1918-19
with those of a lower socio-economic status more at risk. Moreover, a similar immigrant effect is
proposed by Jones (2002, 2005, 2006, and 2007) in her analysis of Winnipeg, where the
mortality rate for northern Winnipeg, with a population primarily composed of immigrants and
the poor, was higher than southern Winnipeg where the wealthier, Canadian-born citizens lived
(see Table 1). However, it is important to keep in mind that the way that Jones calculated the
mortality rates is unclear so the accuracy of these rates is undeterminable.
2.6 Spanish Influenza in British Columbia and Vancouver
Very little can be found in the literature specifically related to British Columbia; the most
widely quoted papers are written by Andrews (1977) and Kelm (1999) and a book written by O‟Keefe and MacDonald (2004). Andrews (1977), and O‟Keefe and MacDonald (2004),
concentrate on the City of Vancouver, while Kelm (1999) conducts her analysis on the First
Nations populations of British Columbia.
Andrews (1977) investigated the influenza epidemic in the city of Vancouver using
newspaper articles published in 1918 and 1919 as her primary references. From these, she quotes
the mortality rate as 23.3 per 1,000 (a total of 795 deaths) and suggests miners and pregnant
women suffered the highest mortality rates. The same mortality rate is also referenced in
O‟Keefe and MacDonald (2004) but does not explain where this figure originated. Additionally,
America but, as suggested by Keiser (2006, p.
130) in her review of this book, these rates “beg analysis”.
O‟Keefe and MacDonald (2004) in their
book exploring the actions of Vancouver‟s
medical health officer, Dr. Underhill, reveal that
his assessments of mortality rates show that the
white population of Vancouver was most
affected in contrast to other ethnicities such as
the Chinese, Japanese, and Hindu populations. If this is correct, then Vancouver‟s ethnic
mortalities are in opposition to other locations in Canada, and also the United States, where
studies indicate that the highest mortality rates occurred among immigrant populations (Tuckel et
al., 2006; Jones, 2002, 2005, 2006, and 2007). However, O‟Keefe and MacDonald were not clear whether the white population was composed of Canadians only or if it also includes white
immigrants. Consequently, analysis of race (based on immigration status and Caucasian versus
non-Caucasian) in order to clarify this point will be part of the forthcoming research.
Furthermore, Andrews indicates that there were three waves of Spanish Influenza in
Vancouver, the first in October of 1918, the second in January of 1919, and the third in March of
1919. This observation is consistent with Humphries‟ hypothesis of spread of influenza from the
American military to a location in Eastern Canada and subsequent early fall spread, via train,
across to Vancouver (Humphries, 2005). However, without clear methods and a more precise
investigation of mortality in Vancouver this cannot be ascertained with any reliability.
Figure 3. Vancouver, 1915. Calder, Walter H. (Photographer). View of north west area of downtown Vancouver, showing parts of Stanley Park, Burrard Inlet, and the North Shore. [Online image]. Retrieved from City of Vancouver Archives,
http://searcharchives.vancouver.ca/view-of- north-west-area-of-downtown-vancouver- showing-parts-of-stanley-park-burrard-inlet-and-north-shore;rad.
Kelm (1999), in her examination of the 1918-1919 influenza epidemic in the Aboriginal
populations of British Columbia, highlights the varied mortality rates stated for these populations
depending on who is doing the reporting; for example, the Department of Indian Affairs (600
deaths) versus the Vital Statistics Branch (700 deaths) versus agent reports (over 1,000 deaths),
and suggests that under reporting is likely. Additionally, Kelm indicates that First Nations
populations experienced higher relative mortality rates (46 per 1,000) as compared to non-Native
populations (6.21 per 1,000). However, it is not clear on the exact method with which these rates
were calculated other than that native mortality rate is based on Department of Indian Affairs and
Vital Statistics population information at the band level and mortality information from
flu-related illness (Kelm, (1999); Kelm, 2011, pers. comm.).
Similar to the rest of the literature, transportation routes such as railways and shipping
lines are highlighted as key influenza dissemination vectors in British Columbia with Kelm
(1999) also citing the closure of canneries and the exodus of ill workers back to the reserves as
an example.
2.7 Summary
In conclusion, the Spanish Influenza pandemic of 1918-1919 circumnavigated the globe
in a very short time frame and killed millions of people worldwide. The Spanish flu targeted
young adult populations, a fact uncommon to the more typical influenza virus strains, and
highlighted the importance of transportation routes as dissemination corridors. What remains
fascinating about this pandemic is the speed at which infectious disease can circle the globe,
which is potentially much greater today with the speed and accessibility of world travel. By
understanding the elements of historical epidemics and identifying susceptible populations,
As expressed previously, there is still little known about how Spanish Influenza affected
the population of British Columbia and, according to Kelm (pers. comm., 2011), there is
uncertainty about how socio-economic factors may have played a role in the experience of the
disease during and after the epidemic. The purpose of this research was, therefore, to re-examine
Vancouver death certificates from 1918 and 1919 as a case study and determine what
populations in Vancouver were affected the most and whether or not this can be tied to
socio-economic differences at that time.
There are major limitations in the literature on Spanish Influenza. The biggest limitation
is the difficulty in determining the exact number of deaths because of incomplete reporting,
misdiagnosis, and the potential for confusing diagnoses between this particular influenza strain
and pneumonia (Johnson and Mueller, 2001; Palmer et al., 2007; Patterson and Pyle, 1991).
Specifically, when official records indicate the cause of death as pneumonia during 1918-1919 it
is difficult to know whether the individual had Spanish Influenza and then contracted pneumonia
or whether they only had pneumonia. Overall, these limitations culminate in a commonly
suggested underestimate in both the number of cases and number of deaths from Spanish
Chapter 3: Methods
With the limitation of misdiagnosis in mind, the following methods, grounded in medical
geography, were employed to create as accurate a representation as possible of the impact of
Spanish Influenza in Vancouver.
3.1 Data Collection and Determining Population Size
Death certificates from the City of Vancouver in 1918-1919 and 1921-1922 were
collected from the British Columbia Vital Statistics Agency. The hard copies of these records
were obtained from the microfiche vaults at Simon Fraser University library and the information
on them transcribed to create the excel database used for analysis. Transcription of the records
was conducted by Ashlee LeBourveau at Simon Fraser University. The transcription was verified
by reviewing a random sampling of the original death certificates and comparing it to the created
database. The information provided by these death certificates included some or all of the
following: name of deceased, registered number of the certificate, date of birth of deceased, date
of death, place of death, how long deceased was at place of death, former or usual residence,
cause of death, duration of cause of death, secondary cause of death, duration of secondary cause
of death, sex, age, occupation, parent occupation, race, marital status, birthplace, length of residence in Canada, father, father‟s birthplace, mother, mother‟s birthplace, informant‟s name,
and informant‟s address. In short, the data used for this thesis were Vancouver mortality data for
the period of June 1918 to June 1919 and June 1921 to June 1922.
Mortality rates were calculated for influenza deaths for both of these time periods in
order to determine if Spanish Influenza had a larger affect during the 1918-1919 flu season. The
total population size used for the calculations was based on Government of Canada 1921 census
completed in 1921. However, the population size for 1918-1919 is also based upon the 1921
population size. Considering population size can change substantially over a short timeframe, it
was necessary to examine what was happening in Vancouver in terms of demographics between
1918-1919 and 1921-1922. As an example Macdonald (1992) states that between the census years of 1911 and 1921, Vancouver‟s population increased by a third to 163,000 people.
One way to do this was to see how fast the city grew or shrank and whether migration
patterns changed over this time period. According to Macdonald (1992), Vancouver experienced
a large population drop between 1910 and 1920 especially around the years of 1913 and 1914 as
a result of the depression and the beginning of World War I when military personnel went
overseas. However, Barman (1986, p. 98) states that by World War I the “socio-demographic
framework of Vancouver was in place and the growth was moderated”. During the 1920s growth
in Vancouver occurred again and between the 1921 and 1931 censuses the population grew from
175,000 to 246,000 (Macdonald, 1992). Because 1921-1922 is still the early part of the 1920s
decade in which the major growth began, the 1921 census should still be a good estimate in
terms of population for 1918-1919. Military personnel would have begun returning home during
this period as well increasing population levels to better match those assessed in 1921. During
this same period, populations in Point Grey and South Vancouver more than doubled to 13,000
and 32,000 respectively (Macdonald, 1992) which becomes important when determining the
boundary of analysis. Or in other words, how should Vancouver be physically defined for the
purposes of this study?
3.2 Boundary of Analysis
Identifying the boundary of Vancouver to use for the analysis of Spanish Influenza
the City of Vancouver prior to the amalgamation of Point Grey and South Vancouver in 1929
could be used. This boundary, as represented by historical maps in Hayes (2005), would include
Alma Street to the west, Boundary Street to the east, 16th Avenue to the south, and Burrard Inlet
to the north. In this case, only death certificates with place of residence within this geographical
area would be included for analysis. The second option would be to include both South
Vancouver and Point Grey along with the City of Vancouver, which represents the boundary of
present day Vancouver.
Due to the fact that death certificates were only collected for people who died in
Vancouver and excluded certificates that stated death occurred in Point Grey or South
Vancouver, the boundary of analysis is the City of Vancouver proper, prior to amalgamation (i.e.
option one).
3.3 Calculating Mortality Rates
Mortality rates have been selected as a method of analysis as they provide an “index of
the severity of the disease from both clinical and public health standpoints” (Gordis, 2009, p.
67). This provides an indication of how negatively the population of the City of Vancouver was
affected by Spanish Influenza. Three sets of mortality rates, most conservative, moderately
conservative, and least conservative were used to analyze the affect of Spanish Influenza in
Vancouver, British Columbia. All mortality calculations followed the basic epidemiological
calculation of number of resident deaths from Spanish Influenza divided by number of residents
in the population multiplied by 1,000 as shown in Equation 1 (Gordis, 2009). The denominator
used for this calculation will be 117, 217, the population of the City of Vancouver (excluding
South Vancouver and Point Grey) according to the 1921 Canadian census (Government of
Equation 1. Mortality Rate Formula
The most conservative mortality rate was calculated using only those records that indicated
influenza as the primary cause of death. The moderately conservative mortality rate used only
those records with influenza listed as either the primary cause of death or secondary cause of
death. Finally, the least conservative mortality rate included records having influenza as the
primary or secondary cause of death as well as all pneumonia deaths. Pneumonia deaths were
included in this mortality rate to acknowledge the close association between pneumonia and
Spanish Influenza.
3.4 Managing the Association between Pneumonia and Influenza
In order to manage the close association with Spanish Influenza and pneumonia, the
initial analysis was conducted with the two causes of death categories reviewed separately. By
treating influenza and pneumonia separately and using the 1921-1922 data set as a comparison
any discrepancy in pneumonia deaths can be observed. If pneumonia deaths for 1918-1919 far
exceed the deaths for 1921-1922, it is likely that Spanish Influenza played a role in this
difference. As this was observed, cases having the primary or secondary cause of death as
influenza or primary cause of death as pneumonia are included in the analysis. The
amalgamation of pneumonia and influenza deaths follows methods used by Tuckel et al. (2006)
and Palmer et al. (2007).
3.5 Analysis using the Statistics Package for Social Sciences (SPSS)
Prior to conducting the analysis, the data from the death certificates were coded to collect
similar variables into larger categories (Appendix 1). Cross-tabulations, univariate and
version 20. Once coded, cross-tabulations were run for each year to assess the number and
percent of people in each category versus influenza or non-influenza deaths for the whole data
set. Variables with high proportions of influenza deaths are carried over into univariate analysis
focusing only on residents of Vancouver. Univariate statistical analysis uses chi-square tests to
determine if there is any association between influenza deaths and the following
social/geographical characteristics: gender, age, place of death, birthplace, employment status,
type of employed occupation, marital status, race, simple immigration status, and length of
residence in Canada. The Fisher‟s Exact test for statistical significance was used for two-by-two
cross-tabulations and the Pearson Chi-square test for significance was used for cross-tabulations
larger than two-by-two (Field, 2009). Chi-squares were selected as they test whether there is an
association between two categorical variables, the type of variables present in this study. While
one limitation to chi-square tests is that the “sampling distribution of the test statistic has an
approximate chi-square distribution” because there was a large sample size and the expected
frequencies for the analysis run here were above five in each cell, this limitation was overcome
(Field, 2009, p. 690).
Prior to running the multivariate model, a correlation test using the Cramer‟s V statistic
determined whether any of the variables are correlated to each other or acting as confounders
(Field, 2009). Based on the output of the correlation matrix along with the information provided
by the univariate analysis, age, gender, marital status, immigrant status, birthplace, length of
residence in Canada, and employment status were entered into the multivariate model.
The multivariate analysis used logistic regression to determine if the significant
relationships between influenza and the socio-demographic characteristics presented in the
for age, gender, and marital status were forced into the multivariate model followed by the
introduction of immigrant status, birthplace, and length of residence in a forward stepwise
fashion. This means that:
“[T]he current model is compared to the model when that predictor is removed. If the removal of that predictor makes a significant difference to how well the model fits the observed data, then the computer retains that predictor (because the model is better if the predictor is included). If, however, the removal of the predictor makes little difference to the model then the computer rejects the predictor.” (Field, 2009, p.272)
The SPSS model rejected all three of these variables indicating that neither immigrant status,
length of residence or birthplace could be considered to influence death by influenza when age,
gender, and marital status are controlled.
Employment status was then added to the model, again using the forward stepwise
method, to determine if employment impacted influenza deaths when controlling for age, gender,
and marital status. This variable was retained in the model implying that employment could be
considered to influence death by influenza when age, gender, and marital status are controlled.
3.6 Visual Representation
Finally, maps of Spanish Influenza were created for the City of Vancouver using ArcGIS
9.3. As detailed neighbourhood classification cannot be undertaken, maps pinpointed locations of
residence rather than providing amalgamated mortality rates for communities. An Address
Locator created in ArcGIS used present day street configuration information from the open
source database at the City of Vancouver. The addresses for only the residents of the City of
Vancouver in 1918-1919 were uploaded into the GIS program and the address locator geocoded
them to the current street configuration. Any addresses that showed up outside of the boundaries
showing the density of deaths in the City of Vancouver during 1918-1919. The mapping was
used as a visual aid due to the limited ability to match the historical addresses to present day
configuration with a high percentage of accuracy.
3.7 Limitations
While death certificates, census data, and mortality rates provide useful information
regarding the affect of Spanish Influenza in Vancouver, it is important to recognize and
acknowledge the limitations of using these sources and methods.
The death certificates collected for the analysis are from 1918-1922 and are hand-written.
While they represent a good primary source of data, in some cases the information provided was
illegible or did not have all sections completed resulting in missing information. If the missing
information excluded important elements such as age or location of death, or the cause of death
was illegible then this record was removed from the analysis when those variables were used.
The higher the number of records that are removed from analysis the less accurately the results
will represent what actually occurred during the 1918-1919 Spanish Influenza epidemic in
Vancouver. For this reason, missing information could be a severe limitation to this study.
Additionally, all death certificates that were from Vancouver were selected; however, it is not
clear what area this actually constitutes, i.e. does it include Point Grey and South Vancouver? As
such, residence information provided by the death certificate was relied upon for the majority of
the study but this information was not consistently recorded. The address locator was also
intended to reduce error in terms of the boundary of analysis. Lastly, at that time, death
certificates were generally not recorded for the Aboriginal population meaning that there is no
analysis of this population component included in this study (Belshaw, 2009). To conclude, the
factors therefore restricting the understanding of these elements to indirect measures such as
occupation.
The census provides population data for amalgamated areas, for example, British
Columbia, and Vancouver but does not provide information on smaller neighbourhood areas
within Vancouver such as Shaughnessy, Kitsilano, or other smaller neighbourhood units.
Because the census data is limited in what it can provide on the small scale, a neighbourhood
analysis is impossible to undertake. Neighbourhood analysis would have been useful in teasing
out underlying socio-economic risk factors such as poor versus wealthy as some neighbourhoods
are known for the financial attributes of their residents.
Other limitations to the census data, as pointed out by the census itself include error
sources such as the uncertainty of exact age, intentional misstatements, missing information
(missing data/lost data/no response), or collection of information via third parties (i.e. not the
individual themselves) (Census, 1921). Furthermore, within the census itself there is little, to no,
analytical or interpretive discussion of what the data provided represents or what the margin of
error may be. Lastly, the census did not occur during the period of 1918-1919 so analysis for
both sets of death certificate data is conducted using population sizes from 1921. However,
census data are useful as they follow a consistent method of data collection.
As mentioned previously, mortality rates are a good indication of the severity of a disease
in a population; however, these rates do not give any indication of how many people were
affected by the disease, as they only consider those who have died as a result of the disease.
Cases of influenza cannot be analyzed using this method and, as such, the total impact of Spanish
Influenza cannot be fully understood. Furthermore, death certificates do not include any
out the ability to examine more social impacts or responses to the epidemic. Finally, mortality
rates are restricted by a reliance on population data. As this is a historical analysis, specific
elements such as a neighbourhood analysis had to be left out due to the lack of appropriate
available population information.
Mapping the residence locations also proved to be a challenge and the resulting map does
not capture all the deaths as not all the addresses could be matched to present day street
configurations in Vancouver.
Lastly, the choice to solely use the quantitative data provided by the death certificates and
census population information limited my ability to utilize all the facets of a medical geography
approach which also supports understanding the spatial distributions and patterns of movement
of people within the city, to and from the city, and other more social facets of life in Vancouver