See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/12198231

Heat-stress-related mortality in five cities in Southern Ontario: 1980-1996

Article  in  International Journal of Biometeorology · December 2000

DOI: 10.1007/s004840000070 · Source: PubMed

CITATIONS READS

164 193

3 authors, including:

Karen Smoyer Tomic Daniel Gareth Charles Rainham

Oxford Pharmagenesis Ltd. Dalhousie University
13 PUBLICATIONS   897 CITATIONS    77 PUBLICATIONS   1,518 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Identifying urban thermal hotspots using space time cubes View project

UrbanSense View project

All content following this page was uploaded by Karen Smoyer Tomic on 16 April 2014.

The user has requested enhancement of the downloaded file.

Int J Biometeorol (2000) 44:190–197 © ISB 2000

O R I G I N A L A RT I C L E

Karen E. Smoyer · Daniel G.C. Rainham

Jared N. Hewko

Heat-stress-related mortality
in five cities in Southern Ontario: 1980–1996

Received: 9 August 1999 / Revised: 9 March 2000 / Accepted: 1 May 2000

Abstract The Toronto–Windsor corridor of Southern Introduction

Ontario, Canada, experiences hot and humid weather
conditions in summer, thus exposing the population to Heat waves are responsible for a large number of deaths
heat stress and a greater risk of mortality. In the event of in North America. For example, in the United States, the
a climate change, heat-stress conditions may become heat wave of 1980 resulted in approximately 10 000 ex-
more frequent and severe in Southern Ontario. To assess cess deaths, and the heat wave of 1988 resulted in an es-
the impact of summer weather on health, we analyzed timated 5000–10 000 deaths. In comparison, the number
heat-related mortality in the elderly (older than 64 years) of deaths resulting from each of the major floods, hurri-
in the metropolitan areas of Windsor, London, Kitch- canes, and blizzards in the United States between 1980
ener-Waterloo-Cambridge, Hamilton, and Toronto for a and 1999 ranged from zero to 270, with an average of 41
17-year period. Demographic, socioeconomic, and hous- deaths per event (National Climatic Data Center 1999).
ing factors were also evaluated to assess their effect on Clearly, the health risks that heat waves pose to the pub-
the potential of the population to adapt and their vulnera- lic warrant major concern.
bility to heat stress. Heat-stress days were defined as Past research has observed heat-related mortality in
those with an apparent temperature (heat index) above Toronto and Montreal for the period 1958–1988, al-
32°C. Mortality among the elderly was significantly though this is of lesser magnitude than in many United
higher on heat-stress days than on non-heat-stress days States cities (Kalkstein and Smoyer 1993a,b). The re-
in all cities except Windsor. The strongest relationships sults of these previous studies indicate that a climate
occurred in Toronto and London, followed by Hamilton. change stemming from increasing greenhouse gas con-
Cities with the greatest heat-related mortality have rela- centrations would elevate the risk of heat-related mor-
tively high levels of urbanization and high costs of liv- tality for the population of Toronto, Montreal, and the
ing. Even without the warming induced by a climate surrounding areas, because of increased exposure to
change, (1) vulnerability is likely to increase as the pop- heat stress through more frequent and severe heat waves
ulation ages, and (2) ongoing urban development and in these regions.
sprawl are expected to intensify heat-stress conditions in This study focuses on heat-stress/mortality responses
Southern Ontario. Actions should be taken to reduce vul- in five cities in the Toronto–Windsor corridor of South-
nerability to heat stress conditions, and to develop a ern Ontario, an area that experiences hot and humid
comprehensive hot weather watch/warning system for weather conditions in summer. With regard to human
the region. health, vulnerability and adaptation to climate change in
the Toronto–Windsor corridor are likely to be influ-
Keywords Heat-related mortality · Apparent enced by a combination of demographic composition,
temperature · Heat stress · Southern Ontario · Canada urban morphology, and socioeconomic conditions. By
comparing heat-stress/mortality relationships within ar-
eas experiencing similar climatic conditions, this study
begins to elucidate factors affecting vulnerability and
K.E. Smoyer (✉) · D.G.C. Rainham · J.N. Hewko adaptation to climate variability as expressed through
Climate and Health Research Program, heat-stress episodes.
Department of Earth and Atmospheric Sciences, The primary objectives of this research are (1) to
1–26 Earth Sciences Building, University of Alberta,
Edmonton, Alberta T6G 2E3, Canada identify the relationship between conditions expected to
e-mail: Karen.Smoyer@ualberta.ca lead to heat stress and mortality for the largest cities in
Tel.: +780-492-3287 the Toronto–Windsor corridor, and (2) to identify poten-
 191

tial demographic and socioeconomic risk factors in heat- tic effects of high temperature and humidity and their ability to
stress mortality that may impede adaptation in the event stress the body’s thermal regulatory systems (Steadman 1984).
In previous research in St. Louis, Missouri, we have used a heat
of climate change. index of 40.6°C (105°F) as an indicator of heat-stress days
(Smoyer 1998a), but the heat index in the Toronto–Windsor corri-
dor rarely exceeds this value. Populations tend to respond to weath-
Materials and methods er in a relative, rather than an absolute fashion (Kalkstein and Da-
vis 1989), and thus we expect that heat-stress will occur under less
The research spans 1 June—31 August for each year between 1980 severe conditions in this area. Therefore, we have evaluated the in-
and 1996. The study includes the five largest cities (or amalgama- crease in mortality when the maximum heat index exceeds 32°C.
tions of adjacent cities) of the Toronto–Windsor corridor in South- This value (approximately 90°F) corresponds with the potential im-
ern Ontario: Windsor, Kitchener-Waterloo-Cambridge (KWC), Lon- pacts on health of sunstroke, heat cramps, and heat exhaustion
don, Hamilton, and the newly amalgamated city of Toronto, referred (Steadman 1979), and has been adopted by the United States Na-
to as Metropolitan Toronto (Fig. 1). Metropolitan Toronto is the tional Weather Service for conveying heat-stress warnings.
largest of the cities studied, with over 2×106 people, while Windsor We calculated apparent temperature (the heat index) as fol-
is the smallest at just under 200 000 people as of 1991. The other lows. First, we identified the maximum air (dry bulb) temperature
three cities had 1991 populations ranging from 303 165 in London for each day and selected the dew point temperature that occurred
to 332 235 in KWC. Persons over 64 years of age comprised be- at the same hour as the maximum air temperature (which was not
tween 10.2% of the total population in KWC and 14.4% of the pop- necessarily the maximum dew point temperature). The maximum
ulation in Hamilton (Table 1). Because the elderly population, rep- daily air temperature and the corresponding dew point temperature
resented here by persons over 64 years old, is more susceptible to were obtained from Environment Canada data. For 1 June–31 Au-
weather-related mortality than younger populations, the results pre- gust for the 17-year period between 1980 and 1996, we used 24-h
sented in this paper are for the elderly only. weather data from the Toronto Pearson, Windsor, and London air-
An apparent temperature (Tapp) approach, also known as the port stations; daytime data for Hamilton and Waterloo Wellington
heat index, was adopted to identify heat-stress days. This measure Airport were used since 24-h data were not available.
takes into account human physiology and is based on the synergis- We calculated the apparent temperature from the following
equation, which we derived from Steadman’s table of air tempera-
ture, dew point temperature, and apparent temperature, assuming
Table 1 Demographic characteristics of cities in the Toronto- no or light winds (Steadman 1979):
Windsor corridor, 1991. Elderly population over 64 years old;
KWC Kitchener/Waterloo/Cambridge Tapp=-2.719+0.994Tair+0.016(Tdewpt)2 (1)

City Total population Elderly population where Tair=air temperature (°C) and Tdewpt=dew point temperature
(% of total) (°C). Days with Tapp above 32°C are considered heat-stress days
for each of the weather stations used in the analysis.
Windsor 191435 27355 The term “heat-stress,” rather than “heat wave” was used to de-
(14.3) note hot and humid conditions that include those that are less ex-
London 303165 35900 treme (e.g., a heat index above 32°C) than what would typically be
(11.8) associated with heat-wave conditions in Mid-Western and Mid-At-
KWC 332235 33865 lantic cities in the United States. This lower cut-off value was chosen
(10.2) for this study because the Toronto–Windsor corridor does not reach
Hamilton 318499 46020 the temperature extremes found in regions of the United States.
(14.4) In addition to apparent temperature, we created variables to
Metropolitan 2275771 291095 measure the onset and duration of heat-stress episodes. Past re-
Toronto (12.8) search has shown that heat waves occurring early in the summer
season have a greater impact on mortality than those of similar in-

Fig. 1 The Toronto–Windsor

corridor study area and weather
stations
192
tensity occurring later in the season (Greenberg et al. 1983; Kalk- ing the role of possible non-climatic factors in heat-stress mortali-
stein 1991; Kalkstein and Smoyer 1993b). To investigate the effect ty in the Toronto–Windsor corridor, each city was described in
of heat-stress onset, a dummy variable was created, based on the terms of its social and urban character. We characterized housing,
number of each day of the summer season from 1 (for 1 June) to urban morphology, demographic composition, and socioeconomic
92 (for 31 August). conditions, using several indicators from the 1981, 1986, 1991 and
To measure the duration of heat-stress episodes, we used two 1996 Statistics Canada censuses. These profiles contextualize the
variables. One measures the position of a given day within a consec- results of the heat-stress/mortality analyses.
utive string of days with Tapp above 32°C, and the other measures the
number of hours per day with an Tapp above 32°C (HSH). The HSH
variable requires 24-h data and thus was only used for cities repre-
sented by the Toronto, London, or Windsor weather stations. Results
Daily mortality data were obtained from Statistics Canada.
Mortality from many different causes increases during heat-stress Toronto Pearson station recorded the greatest range of
periods, and heatstroke is often under-reported as a cause of death
(Kilbourne 1989). Therefore, all non-accidental causes of death mean temperatures over the study period. All stations
were used, rather than only those that are directly related to heat. reported their highest mean maximum air temperature
The daily number of elderly deaths was standardized by the devel- during the summer of 1988. Mean apparent tempera-
opment of a between-years standardization method to account for tures were relatively similar across all weather stations
declining mortality rates over the study period. Statistics Canada
census data, available every 5 years (1976, 1981, 1986, 1991, and in the study area. In Windsor, London, and Toronto, the
1996) were used to establish the baseline elderly populations for highest heat index (Tapp) occurred in 1995, rather than
calculating rates. For each year in the study period, we used the 1988 (Table 2). For Waterloo and Hamilton, however,
same population denominator. In intervening (non-census) years, the summer of 1988 had the most severe apparent tem-
we estimated population size on the basis of an exponential perature as well as the highest mean maximum tempera-
growth function following Palmore and Gardner (1983):
ture. Windsor station recorded the highest percentage of
( population at year t ) heat-stress days and the greatest average number of
e rt = (2)
( population in initial year) heat-stress days per summer, more than twice the num-
where r=rate of change and t=number of years. ber found at Waterloo. The number of hours with a tem-
The equation was solved for r using consecutive census popu- perature above 32°C (HSH) was more than two times
lations (e.g., 1976 and 1981, and then 1981 and 1986, and so on). greater for Windsor than for London or Toronto. Over-
This method interpolates inter-census populations using growth
rates specific to each inter-census period. Population-standardized all, the Windsor station recorded the highest maximum,
mortality rates, which are commonly employed in many other minimum, and apparent temperatures. In addition, the
studies, do not take into account changes in mortality rates over Windsor station had the longest consecutive run of heat-
time that reflect non-climatic health trends. As a result, a between- stress days, double the number recorded at the London
years standardization procedure was used to remove inter-annual
trends from mortality rates by fitting a regression line through the or Toronto stations.
mean annual summer mortality rates for each of the 17 years in Mean daily elderly death rates ranged from a low of
the study period. Regression lines were fit for each city separately. 11.4/100 000 seniors for Metropolitan Toronto to a high
On the basis of regression-fitted values, we used the difference of 13.35/100 000 in Windsor (Table 3). Cities with high
between observed and expected mortality rates for each day (i.e., mean mortality rates for the elderly do not necessarily
daily residuals) instead of mortality rates. Between-years stan-
dardization was calculated as follows: exhibit heightened sensitivity to heat-stress. In fact, dur-
ing heat-stress days, Windsor had the highest elderly
Between-years standardized mortality (i,j,k) = mortality rates of the five cities, yet the rates on these
mortality(i,j,k) – predicted mortality (j,k) (3) days were not statistically significantly higher than those
where i=day (1 June–31 August), j=summer season (1980–1996), on non-heat-stress days (results not shown). These find-
and k=city or census subdivision (CSD). This method compen- ings illustrate the importance of using excess rather than
sates for the decline in mortality rates that was observed over the total mortality to identify sensitivity to heat-stress.
period in all cities except London and KWC. Heat-stress/mortality relationships among the elderly,
An excess mortality method was used to investigate heat-
stress/mortality relationships on the basis of the between-years based on the significance level of the difference between
standardized mortality for each city. Excess mortality attributed to mean daily heat-stress mortality and mean daily non-
heat-stress for each city was derived by taking the mean of daily heat-stress mortality, adjusted using the between-years
between-years standardized mortality for days classified as heat- standardization method, were strongest in Metropolitan
stress days over the whole period and comparing this value to the Toronto, Hamilton and London, followed by KWC and
mean daily mortality on non-heat-stress days. Because the study
hypothesis was that mortality was greater (rather than merely dif- Windsor (Table 3). The difference of means tests reveal
ferent) on heat-stress days than on non-heat-stress days, a one- that mean mortality was higher on heat-stress days than
tailed, unpaired t-test was used to evaluate statistical significance. on non-heat-stress days for all cities, but this relationship
The strength of heat-stress/mortality relationships in the Toronto– was not statistically significant (at the 0.05 level) for
Windsor corridor can be assessed by comparing the value of t for
each city. However, because the between-years standardized mor- Windsor.
tality treatment is specific to each place, neither the magnitude of Daily time series of deaths and apparent temperature
mortality on heat-stress days nor the difference in means between illustrate the association between the onset of heat-stress
heat-stress and non-heat-stress days should be compared among and its intensity and duration, and mortality. For exam-
places.
Past research has implicated factors such as population density ple, in the summer of 1988, in London the correlation
and low income levels in mortality risk during heat waves between deaths per day among the elderly and Tapp is
(Buechley et al. 1972; Smoyer 1998a,b). As a means of investigat- significant at the 0.05 level (r=0.23), but the relationship
 193
Table 2 Description of weather
in the Toronto–Windsor corri- Parameter Windsor London Waterloo Hamilton Toronto
dor. The stations at Waterloo 42°16′N, 42°59′N, 42°27′N, 43°16′N, 43°40′N,
and Hamilton do not have com- 82°58′W 81°13′W 80°23′W 79°54′W 79°38′W
plete 24-h weather data. Tmin
mean minimum air temperature, Mean minimum air 16.9 14.3 − − 14.3
Tmax mean maximum air temper- temperature (Tmin, °C)
ature, Tapp apparent temperature, Mean maximum air 26.2 24.6 23.6 24.6 24.8
HSH heat stress hours number temperature (Tmax, °C)
of hours with Tapp>32 °C, heat Year with highest mean (Tmax) 1988 1988 1988 1988 1988
stress number of days with
Tapp>32 °C Mean Tapp (°C) 27.5 25.6 24.6 25.6 25.4
[% days Tapp>32°C] [20.4] [11.6] [9.5] [11.3] [11.7]
Year with highest 1995 1995 1988 1988 1995
mean Tapp
Mean occurrence of 18.9 10.7 8.7 10.4 10.8
heat-stress days/year [136.3] [63.5] [−] [−] [68.1]
[mean HSH/year]
Longest consecutive 14 8 7 7 7
period of Tapp (days) (1995) (1987) (1987) (1987) (1987)
[year of occurrence]

Table 3 Heat-stress mortality relationships for the elderly (over 64 years old) in the Toronto-Windsor corridor

City Mean daily Mean daily Mean standardized Mean standardized MHS–MNHS
deaths death ratea daily heat-stress daily non-heat-stress t
mortality mortality (p-value)
(MHS) (MNHS)

Windsor 3.44 13.35 0.51 –0.13 0.65

t=1.35
(0.09)

London 3.97 12.23 1.58 –0.21 1.79

t=3.76
(0.00)

KWC 3.72 12.22 1.11 –0.10 1.22

t=2.02
(0.02)

Hamilton 5.29 12.50 1.20 –0.15 1.35

t=2.92
(0.00)

Metropolitan 30.43 11.40 0.86 –0.11 0.97

Toronto t=5.22
(0.00)
a Daily deaths per 100000 elderly people

between the two variables is not consistent throughout Deaths in London increased on 9 July 1988, 2 days after
the summer (Fig. 2, top). Both onset and duration appear the 7 July onset of the first prolonged heat-stress episode
particularly important. The spike in mortality on 22 June of the season. Mortality levels decreased below the base-
1988 occurred on a heat-stress day (Tapp=33°C), yet mor- line of 4.07 deaths/day after this period, suggesting some
tality did not increase dramatically with the very high pre-shifted mortality associated with the 7–11 July epi-
heat index (Tapp=40°C) that occurred shortly thereafter, sode. Another heat-stress period occurred on 2–6 Au-
on 25 June. This observation may be the result of “pre- gust. Deaths among the elderly increased, but not to the
shifted mortality” or “harvesting”, wherein high mortali- levels of the first heat-stress period. Thus mortality in
ty among the most vulnerable on 22 June would result in London in 1988 appears to be associated with a combi-
a smaller pool of susceptible persons remaining who nation of early onset and prolonged heat-stress condi-
could have died after exposure to the heat-stress condi- tions.
tions on 25 June. The temporal correlation between apparent tempera-
Under more prolonged heat-stress conditions, howev- ture and mortality in Windsor was lower than that in
er, Tapp and mortality appear to be linked more closely. London (r=0.11) and not significant at the 0.05 level.
194

Fig. 2 Daily time-series of summer mortality and apparent tem- small number of daily deaths (approximately 3) in Lon-
perature in London (top) and Windsor (bottom) for 1988 don, KWC and Windsor (Table 3) may have influenced
the results in these smaller cities, as minor variations in
The correlation analysis of the entire study period, how- mortality can significantly modify the observed relation-
ever, showed that the time of the summer when heat- ship. In Toronto, heat-stress duration [as measured in
stress occurred was significantly associated with mortali- days and hours (HSH)] had a relatively strong correla-
ty in Windsor (Table 4). Deaths of the elderly in Windsor tion with mortality (Table 4). The date of heat-stress on-
were quite high on several days with high apparent tem- set was important in London and Hamilton, yet was not
peratures earlier in the season, such as 22 June. Howev- significant at the 0.05 level in Toronto or KWC. For all
er, the association was inconsistent, weakening later in cities but Windsor, the heat-stress days (Tapp>32°C), ap-
the summer season (Fig. 2, bottom). parent temperature, and maximum temperature were sig-
The weather variables having the strongest associa- nificantly correlated with mortality. Minimum tempera-
tion with mortality were not consistent for each city. The ture had a lower correlation with mortality than maxi-
 195
Table 4 Correlation of selected daily weather variables and mortality

City Heat-stress Tapp Tmax Date of Heat stress Tmind HSHe

daya onsetb durationc

Windsor 0.035 0.029 0.035 –0.134* 0.038 0.028 0.037

London 0.091* 0.085* 0.080* –0.147* 0.103* 0.055* 0.093*
KWC 0.056* 0.070* 0.064* –0.095 0.039 − −
Hamilton 0.077* 0.098* 0.103* –0.173* 0.077* − −
MetropolitanToronto 0.146* 0.136* 0.130* –0.115 0.151* 0.128* 0.151*
*Significant at P <0.05 cNumber of consecutive heat-stress days
aDichotomous variable where 0 = Tapp≤32 °C, and 1 = Tapp>32 °C dData were unavailable for KWC and Hamilton
bTime of the summer (1=1 June, 92=31 August) when heat-stress eHours with Tapp>32 °C in a heat-stress day. Data were unavail-
day occurred (this variable has been correlated only with daily able for KWC and Hamilton
mortality on heat-stress days)

Table 5 Summary of heat-stress mortality relationships and census characteristics. The categorization is based on each city relative to
all other cities in the Toronto–Windsor corridor study area

City Strength of Degree of Percentage Incidence of Cost Percentage of

heat-stress/mortality urbanization of seniors low incomeb of living older housingc
relationshipa

Windsor Not significant Low High Moderate Low High

London Very strong Moderate Low Low Moderate Low
KWC Moderate Low Low Low Low Low
Hamilton Strong High High High High High
Metropolitan Toronto Very strong High Moderate High High Moderate
a Strength of relationship based on t value where t≤1.65=not signif- b Incidence of low income was unavailable for 1981
icant; 1.65<t≤1.96=weak; 1.96<t=2.56=moderate; 2.56<t=3.29= c Housing built before 1946
strong; and t >3.29=very strong

mum temperature or apparent temperature in both Toron- high cost of living (Table 5), while the incidence of low
to and London, and was not significant in Windsor. income was low in London and high in Hamilton and
Variation in the urban, socioeconomic, and demo- Toronto. We also noted different heat-stress-related
graphic structure of the cities in the Toronto–Windsor mortality levels in cities with relatively similar housing
corridor is considerable. Toronto and Hamilton are the and economic histories. For example, Windsor has the
most urbanized, Windsor and KWC are the least urban- greatest percentage of dwellings built before World War
ized, and London falls in the middle. Windsor and II, followed by Hamilton, and manufacturing is the
Hamilton have the highest concentration of older dwell- leading source of employment in both cities (Statistics
ings, KWC and London have the lowest, and Metropoli- Canada 1994; 1998). The two cities, nevertheless, had
tan Toronto has a moderate level of older housing (Ta- markedly different levels of excess mortality. The inci-
ble 5). Windsor and Hamilton also have the highest per- dence of low income and the percentage of elderly resi-
centage of seniors, London and KWC have the lowest dents also did not appear to be related to heat-stress
percentage, and Toronto has a moderate percentage of mortality.
people 65 years and older. The cost of living has fluc-
tuated over the study period, but in 1991 and 1996 Me-
tropolitan Toronto and Hamilton had the highest costs Discussion
of living, Windsor and KWC had the lowest, and Lon-
don had a moderate cost of living. On the basis of The results of this research reveal that heat-stress mortal-
1991 and 1996 data, Hamilton and Toronto had the ity varies among places with relatively similar weather
highest incidence of low income, London and KWC the conditions. These findings illustrate the importance of
lowest, and Windsor a moderate incidence of low in- non-climatic factors in mediating the effect of summer
come (Table 5). weather on human health, and provide information about
The difference of means tests and Pearson’s correla- the ability of the Toronto–Windsor corridor’s population
tion coefficients consistently identified the strongest to adapt to heat-stress conditions. The largest uncertainty
heat-stress/mortality relationships in Metropolitan To- lies in identifying which specific non-climatic factors are
ronto and London, followed by Hamilton. Census data the most important. The results of the analysis implicate
on population, housing, and socioeconomic conditions urbanization as a potential risk factor in heat-stress mor-
reveal some similarities among these cities, but differ- tality. This finding corresponds with the results of stud-
ences as well. For example, all three cities have rela- ies of heat-related mortality in the United States (Bu-
tively high levels of urbanization and a moderate to echley et al. 1972; Smoyer 1998a,b). Urban areas experi-
196

ence higher temperatures than less built-up areas, partic- nomic characteristics, and urban structure may also be
ularly at night when buildings and paved areas continue confounded by unmeasured variables such as air pollu-
to emit long-wave radiation. The lack of a respite from tion. A large body of literature has illustrated the harmful
heat-stress conditions at night has been thought to be effects of air pollution on health (e.g., Choi et al. 1997;
particularly dangerous to human health (Kilbourne et al. Katsouyanni et al. 1993; Burnett et al. 1998a,b; Smoyer
1982). As urban development continues in the Toron- et al. 2000). Further analysis of heat-stress in the Toron-
to–Windsor corridor, population vulnerability to heat- to–Windsor corridor would benefit from the addition of
stress is likely to increase. air pollution data.
The effect of demographic composition and socioeco- An important question arising from our research is
nomic conditions on vulnerability to heat-stress is less how population vulnerability and adaptation to climate
clear than that of urbanization. Demographic and socio- change may vary within the Toronto–Windsor corridor.
economic characteristics did not appear to be consistent- The results suggest that vulnerability to increases in
ly associated with the strength of the heat-stress/mortali- heat-stress induced by climate change will be greatest
ty relationship. Among the cities with very strong rela- among elderly urban dwellers, particularly in cities with
tionships, London had a low percentage of seniors and a high cost of living. Changes in demographic composi-
incidence of low income and a moderate cost of living, tion must be considered along with changes in the cli-
while Metropolitan Toronto had a moderate percentage mate of this area. As the Canadian population ages, a
of seniors along with a high cost of living and incidence greater number of people will enter the high-risk group,
of low income. Hamilton ranked high on all three of and thus we expect to see increases in the number of
these variables. KWC and Windsor, which had the weak- deaths during heat-stress events. Therefore, high-risk cit-
est heat-stress/mortality relationships, had the lowest ies with larger numbers of elderly residents, such as
costs of living. Thus, it appears that heat-stress mortality Hamilton and Metropolitan Toronto, are most in need of
is not a function of the age structure of the population, policies that reduce vulnerability and increase the poten-
but that socioeconomic factors are important. These find- tial for adaptation.
ings correspond with the results of studies in the United In summary, heat-stress mortality among the elderly
States that have demonstrated a relationship between varied within the Toronto–Windsor corridor despite rela-
poverty and mortality risk during heat waves (Buechely tively similar weather within the region. The more
et al. 1972; Kilbourne et al. 1982; Smoyer 1998b). It is densely populated cities of Toronto, London, and Hamil-
notable that cost of living has a stronger association with ton had the strongest relationships, while the relationship
heat-stress mortality than the incidence of low income in in lower-density Windsor was not statistically signifi-
the Toronto–Windsor corridor. Elderly people residing in cant. Although minor differences in summer weather
cities with lower costs of living are likely to have larger variability may account for the variation in the strength
disposable incomes that can provide them with more op- of the relationships identified, non-climatic factors such
tions for avoiding heat-stress, such as air conditioning. as urbanization and cost of living also appear to be im-
Given the more frequent exposure to heat-stress con- portant.
ditions in Windsor, we initially expected to see a strong- Heat-stress conditions are predictable, and heat-stress
er heat-stress/mortality relationship in this city. There are mortality is preventable. A combination of hot weather
several potential explanations for this not being record- watch/warning systems, well-organized plans for re-
ed. The small population and relatively low number of sponding to heat emergencies, and ongoing education
elderly deaths per day (approximately 3.4) in Windsor about precautions against heat-stress is likely to be use-
may have affected our ability to isolate a statistically sig- ful in reducing the harmful impacts of heat-stress condi-
nificant heat-stress/mortality relationship. However, we tions on health. Toronto and Hamilton have begun to put
hypothesize that non-climatic variables, such as protec- such measures in place, but their effectiveness has not
tive behaviors, population density, and air conditioning, yet been formally evaluated. These public health mea-
rather than an insufficiently robust statistical procedure, sures are also important early steps for adaptating to cli-
account for the lack of a heat-stress/mortality effect in mate change. We emphasize, however, that such efforts
Windsor. For example, in 1981, the population density of to promote awareness of heat stress and prevent mortali-
Windsor’s urban core was the lowest among 27 Canadi- ty, which require behavioral changes at the individual
an cities (Bourne 1989). Windsor also consistently had level, are insufficient without public health measures
the greatest percentage of single-family detached dwell- aimed at reducing vulnerability and increasing the adapt-
ings during the study period (Statistics Canada 1983, ability and adaptive capacity of the population as a
1987, 1994, 1998). Unfortunately, air conditioning data whole. We recommend that attention be given to reduc-
for the five cities in this study are not readily available, ing vulnerability to heat-stress and enhancing adaptabili-
and we have not yet been able to investigate differences ty in the Toronto–Windsor corridor. For example, poli-
in air conditioning availability among the cities studied. cies should focus on (1) promoting wise land use so as to
This, however, is an important area for future study. minimize the effects of urban heat islands and provide
In addition to the role of air conditioning, the associa- vegetated and ventilated areas to facilitate cooling, and
tion between heat-stress mortality in the Toronto–Wind- (2) ensuring that health and social services are readily
sor corridor and demographic composition, socioeco- available to populations in need.
 197
Acknowledgements We would like to thank Brian Mills and Ab- D (1993) Evidence for interaction between air pollution and
del Maarouf of Environment Canada, as well as the anonymous high temperature in the causation of excess mortality. Arch
reviewers, for their helpful comments on this manuscript. This re- Environ Health 48:235–242
search was generously funded by the Adaptation and Impacts Re- Kilbourne EM (1989) Heat waves. In: Gregg MB (ed) The public
search Group of Environment Canada, under contract KM170–8- health consequences of disasters 1989. CDC, US Department
6640. We also wish to acknowledge the University of Alberta for of Health and Human Services, Atlanta, pp 51–61
providing start-up funds for mortality data, computer hardware, Kilbourne EM, Choi K, Jones TS, Thacker SB, Field Investigation
and software. Team (1982) Risk factors in heat stroke: a case-control study.
JAMA 247:3332–3336
National Climatic Data Center (1999) Billion dollar US weather
disasters 1980–1999. Website http://wwwncdcnoaagov/ol/re-
References ports/billionzhtml
Palmore JA, Gardner RW (1983) Measuring mortality fertility and
Bourne LS (1989) On the spatial structure of metropolitan areas in natural increase: a self-teaching guide to elementary measures.
Canada: a discriminant analysis. University of Toronto: Centre East-West Center, Honolulu
of Urban and Community Studies. Research paper no 172 Smoyer KE (1998a) A comparative analysis of heat waves and as-
Buechley RW, Van Bruggen J, Truppi LE (1972) Heat island sociated mortality in St Louis, Missouri – 1980 and 1995. Int J
equals death island? Environ Res 5:58–92 Biometeorol 42:44–50
Burnett RT, Çakmak S, Brook JR (1998a) The effect of the urban Smoyer KE (1998b) Putting risk in its place: methodological con-
ambient air pollution mix on daily mortality rates in 11 Cana- siderations for investigating extreme event health risk. Soc Sci
dian cities. Can J Public Health 89: 152–156 Med 47:1809–1824
Burnett RT, Çakmak S, Raizenne ME, Steib D, Vincent R, Krew- Smoyer KE, Kalkstein LS, Greene JS, Ye H (2000) The impacts of
ski D, Brook JR, Philips O, Ozkaynak H (1998b) The associa- weather and pollution on human mortality in Birmingham, Ala-
tion between ambient carbon monoxide levels and daily bama and Philadelphia, Pennsylvania. Int J Climatol (in press)
mortality in Toronto, Canada. J Air Waste Manage Assoc 48: Statistics Canada (1983) Census divisions and subdivisions: popu-
689–700 lation occupied private dwellings private households and cen-
Choi K, Inoue S, Shinozaki R (1997) Air pollution, temperature sus and economic families in private households – Selected
and regional differences in lung cancer mortality in Japan. social and economic characteristics Ontario. 1981 Census of
Arch Environ Health 52: 160–168 Canada, catalogue number E-576, Table 1. Ministry of Supply
Greenberg JH, Bromberg J, Reed CM, Gustafson TL, Beauchamp and Services Canada, Ottawa
RA (1983) The epidemiology of heat-related deaths, Texas – Statistics Canada (1987) Profiles: census divisions and subdivi-
1950, 1970–79, and 1980. Am J Public Health 73:805–807 sions Ontario. Part 2. 1986 Census of Canada, catalogue num-
Kalkstein LS (1991) A new approach to evaluate the impact of cli- ber 94–112. Ministry of Supply and Services Canada, Ottawa
mate on human mortality. Environ Health Perspect 96:145–150 Statistics Canada (1994) Profile of census divisions and subdivi-
Kalkstein LS, Davis RE (1989) Weather and human mortality: an sions in Ontario. Part B. 1991 Census of Canada, catalogue
evaluation of demographic and interregional responses in the number 95–338, Table 1. Industry Science and Technology
United States. Ann Assoc Am Geog 79:44–64 Canada, Ottawa
Kalkstein LS, Smoyer KE (1993a) The impact of climate change Statistics Canada (1998) Profile of census divisions and subdivi-
on human health: some international implications. Experientia sions in Ontario. 1996 Census of Canada, catalogue number
49:969–979 95–187-XPB. Ministry of Industry Canada, Ottawa
Kalkstein LS, Smoyer KE (1993b) The impact of climate on Cana- Steadman RG (1979) The assessment of sultriness part I. A tem-
dian mortality: present relationships and future scenarios. Envi- perature-humidity index based on human physiology and
ronment Canada/Canadian Climate Center Report no 93–7 50 clothing science. J Appl Meteorol 18:861–873
Katsouyanni K, Pantazopoulou A, Touloumi G, Tselepidaki I, Steadman RG (1984) A universal scale of apparent temperature.
Moustris K, Asimakopolous D, Poulopoulou G, Trichopoulos J Climate Applied Meteorol 23:1674–1687

View publication stats