Health & Place 41 (2016) 89–99

Contents lists available at ScienceDirect

Health & Place
journal homepage: www.elsevier.com/locate/healthplace

Heat Death Associations with the built environment, social
vulnerability and their interactions with rising temperature
David P. Eisenman a,b,n, Holly Wilhalme a, Chi-Hong Tseng a, Mikhail Chester c,
Paul English d, Stephanie Pincetl e, Andrew Fraser c, Sitaram Vangala a,
Satvinder K. Dhaliwal a
a
Division of General Internal Medicine and Health Services Research, David Geffen School of Medicine at UCLA, 911 Broxton Plaza, Los Angeles,
CA 90095-1736, USA
b
Center for Public Health and Disasters, UCLA-Fielding School of Public Health, 650 Young Drive South, Los Angeles, CA 90095-1772, USA
c
Civil, Environmental, and Sustainable Engineering, 660 S College Ave, Tempe, AZ 85287, USA
d
California Department of Public Health, Environmental Health Investigations Branch, 580 Marina Bay Parkway, Building P-3, Richmond, CA 94804-6403,
USA
e
Institute of the Environment, UCLA, Life Sciences 5362, Los Angeles, CA 90095, USA

art ic l e i nf o

a b s t r a c t

Article history:
Received 8 March 2016
Received in revised form
30 June 2016
Accepted 3 August 2016
Available online 30 August 2016

In an extreme heat event, people can go to air-conditioned public facilities if residential air-conditioning
is not available. Residences that heat slowly may also mitigate health effects, particularly in neighborhoods with social vulnerability. We explored the contributions of social vulnerability and these infrastructures to heat mortality in Maricopa County and whether these relationships are sensitive to temperature. Using Poisson regression modeling with heat-related mortality as the outcome, we assessed the
interaction of increasing temperature with social vulnerability, access to publicly available air conditioned space, home air conditioning and the thermal properties of residences. As temperatures increase,
mortality from heat-related illness increases less in census tracts with more publicly accessible cooled
spaces. Mortality from all internal causes of death did not have this association. Building thermal protection was not associated with mortality. Social vulnerability was still associated with mortality after
adjusting for the infrastructure variables. To reduce heat-related mortality, the use of public cooled
spaces might be expanded to target the most vulnerable.
& 2016 Elsevier Ltd. All rights reserved.

1. Introduction
In the face of climate change and more frequent and severe
extreme heat events, it has become more important to understand
which neighborhoods and populations are at risk of death from
extreme heat and how properties of the urban environment contribute to or mitigate the risk of heat related death (United Nations, 2014). Heat related deaths can be prevented if individuals
can cool themselves. Air conditioning is the most important form
of cooling available in most cities and its absence from homes is a
factor in heat mortality and morbidity (Blum et al., 1998; Semenza
et al., 1996). In its absence, the U.S. Centers for Disease Control and
Prevention recommend people “reduce their risk for heat-related
illness by spending time in public facilities that are airn
Corresponding author at: Division of General Internal Medicine and Health
Services Research, 911 Broxton Plaza, Los Angeles, CA 90095-1736, USA.
E-mail address: deisenman@mednet.ucla.edu (D.P. Eisenman).

http://dx.doi.org/10.1016/j.healthplace.2016.08.007
1353-8292/& 2016 Elsevier Ltd. All rights reserved.

conditioned”, specifically suggesting shopping malls and public
libraries as public venues for accessing cooled space (U.S. Centers
for Disease Control and Prevention, 2015). Local public health
departments and emergency managers open and provide access to
air-conditioned, cooling shelters in extreme heat events, as this is
commonly believed to be helpful in reducing heat-related deaths,
and neighbors are asked to check on elderly neighbors (Mees,
2015). There is indirect evidence that heat-health warning systems
which couple early warnings with a broad array of emergency
response measures save lives, though it is unknown which of the
measures contribute to any reduction (Ebi et al., 2003).
A community's vulnerability to extreme heat can be understood as a function of its heat exposure, population characteristics
and adaptive capacity. Exposure, in the form of increasing temperature, is related to mortality: in one study for every 10 °F increase in ambient temperature, there was a 2.6% increase in cardiovascular mortality (Basu, 2009). The temperature-mortality
relationship may be non-linear so that longer duration heat events

90

D.P. Eisenman et al. / Health & Place 41 (2016) 89–99

may have more population effects (Anderson and Bell, 2011). Population characteristics including socio-economic and health factors contribute to heat-mortality relationships. Age is commonly
cited as a risk factor with some studies finding that populations
are at higher risk of mortality as well as persons under age five
(Luber and McGeehin, 2008). Minority groups are frequently at
greater risk of heat mortality too, though studies are not consistent in this finding (Hondula et al., 2012; Uejio et al., 2011;
Golden et al., 2008). Poverty, chronic health conditions and social
isolation can be risk factors (Harlan et al., 2013; Kovats and Hajat,
2008, McGeehin and Mirabelli, 2001, Le Tertre et al., 2006; Klinenberg, 2002). Which populations are at risk varies between and
within cities likely due to differences in environment, climate,
culture, demography and adaptations (Harlan et al., 2013; Reid
et al., 2012; Hondula et al., 2015; Davis et al., 2003b).
Less well studied are adaptations to the effects of climate
change that might reduce heat wave mortality. Adaptation is
“adjustment in natural or human systems in response to actual or
expected stimuli or their effects, which moderates harm or exploits beneficial opportunities” (Huang et al., 2011; Hess et al.,
2012). Adaptive capacity refers to the resources available for
adaptation and the ability to use them effectively and efficiently.
These resources can come in many forms including social, educational, physical or financial and together form the inputs for
adaptation interventions, programs and community actions
(White-Newsome et al., 2014; Huang et al., 2011). Access to cooled
space may be the most important physical resource imparting
adaptive capacity and reducing the risk of heat-related health effects (Harlan et al., 2013; Reid et al., 2012; Ostro et al., 2010; O'neill
et al., 2005b). It can come from several sources including airconditioning at work and home and from publicly accessible
cooled spaces such as libraries, commercial venues and public
transportation. But simply getting more people to use home airconditioning is not a panacea as estimates of increasing heat and
water shortages may strain the electricity grid's ability to provide
power during extreme heat events (Sathaye et al., 2012). Buildings
with thermal properties that allow it to heat slower in the sun and
maintain cool air within may, besides contributing to environmental sustainability and reducing strain on the power grid, provide further adaptive capacity in the face of extreme heat. Similar
to social vulnerability, adaptive capacities and their potential for
reducing mortality could vary between regions. For instance,
cooling centers that are integrated into a community may be more
broadly useful in an extreme heat event than cooling centers that
carry the stigma attached to primarily serving the homeless or
senior population (Hayden et al., 2011; White-Newsome et al.,
2014).
This study focuses on the relationship between social vulnerability and adaptive capacity in Maricopa County, Arizona. Maricopa County includes Phoenix, the second most populous city in
the southwestern United States (Los Angeles is the most populous), where summertime temperatures frequently exceed 105
(40.5 °C) F, as well as Mesa, Glendale, Chandler, Scottsdale and
Tempe. The daily mean summer temperature in Phoenix, 33 °C
(91.4 F), is the highest of all major metropolitan areas in the United
States (Petitti et al., 2016). Several studies have addressed aspects
of adaptive capacity in Phoenix and Maricopa. One study of three
socioeconomically vulnerable Phoenix neighborhoods at potential
high risk from extreme heat found access to cooled space was
more nuanced than simply having an air conditioner in the home
(Hayden et al., 2011). The cost of electricity prevented over 36% of
survey respondents from using their air conditioner and an additional 6% reported having a nonfunctional air conditioner. 38% of
participants endorsed feeling too hot inside their homes and home
renters and Hispanic respondents were significantly more likely to
experience this. The authors observed that air conditioner use was

“often limited to simply reducing the extreme heat but not necessarily providing relief” [p275]. Hydration and cooling centers
were not well used by this sample: Only 9 of 359 respondents had
ever used a heat refuge station. Other Phoenix-based surveys, and
multi-city studies report that residential air-conditioning may be
installed but residents may not turn them on due to the costs of
electricity or disliking the feel (Sheridan, 2007; Lane et al., 2014).
Thus, while characteristics of the urban infrastructure such as
the availability of residential air conditioners can contribute to
heat's impact on health it is necessary to examine other sources of
adaptation such as building thermal properties that maintain cool
air and retard heating and access to public air-conditioned spaces
where one may cool down. For instance, residents living in communities with air-conditioned spaces and buildings that heat
slowly might experience less risk as temperatures increase than
residents without these protections. Though the thermal properties of buildings, which in this context means the ability of
buildings to keep the indoor space cool, has also been offered as a
factor in heat mortality we know of little research that has tested
this connection (Kovats and Hajat, 2008; Mavrogianni et al., 2012;
Anderson et al., 2013). Similarly, individuals cope with extreme
heat in a variety of ways. Kalkstein and Sheridan reported that 72%
of respondents in their Phoenix survey “went to an air-conditioned
location or stayed indoors” on excessively hot days (Kalkstein and
Sheridan, 2007). This suggests that one way individuals may cope
is by going to publicly available cooled spaces, though the compound nature of their survey question limits any estimation of
frequency. On the one hand, the senior centers, libraries, community centers and other community sites that are designated as
cooling centers may be distributed sparsely, unevenly and loosely
tied to the neighborhoods most in need. Fraser reported that only
2% of Maricopa County, Arizona residents are within walking distance of an official cooling center (Fraser et al., 2016). On the other
hand, individuals may seek non-residential, publicly available air
conditioned spaces such as shopping malls and libraries. Still, we
know of no research examining how differential access to air
conditioned public spaces relates to health outcomes, a question
we ask in this study. Examining the spatial variation of social
vulnerability to heat, a common approach in heat-vulnerability
studies, should examine these adaptive capacities that may mitigate vulnerability.
This paper explores the contributions of exposure, susceptibility and adaptive capacity to extreme heat mortality. Specifically, it investigates how susceptibility and adaptive capacity, operationalized as social vulnerability and infrastructure factors respectively, contribute to mortality from extreme heat and whether
the relationships between mortality and the social vulnerability
and infrastructure factors are sensitive to increasing temperature
in Maricopa County, Arizona. The infrastructure factors examined
are access to publicly available air conditioned space, home air
conditioning and the thermal protective properties of residential
buildings. We hypothesized that census tracts with increased social vulnerability, less home air conditioning, less access to publicly available cooled space and lower scores on an index of
building thermal protection would have excess heat mortality.

2. Methods
Two outcomes were evaluated: mortality from all internal
causes and mortality from heat-related illnesses. All internal causes of mortality is the most frequent overall mortality outcome
studied because it is broad enough to capture the full range of
causes of death from heat (excessive heat can exacerbate chronic
medical conditions thereby leading to death), it provides sufficient
power to detect associations and it allows comparison to a wider

D.P. Eisenman et al. / Health & Place 41 (2016) 89–99

range of health risk factors. These other causes of death may be
more consistently applied by coroners than heat-related illnesses,
especially if coroners are more likely to consider heat-related
causes during official heat waves than on non-heat wave days
(Shen et al., 1998). Studying heat-related illness mortality may
capture a different health risk profile than does all-internal causes
of mortality since it captures deaths directly attributed to the acute
effects of heat exposure or to specific demographic profiles including outdoor workers.
Mortality data came from the Arizona Department of Health
Services for 2005–2010. Each record included date of death, the
underlying cause and the secondary causes of death. We excluded
external causes of death, such as homicide or suicide, except heat
related causes. Heat-related mortality was defined as any death
which had code X30, exposure to excessive natural heat, and T67,
effects of heat and light, listed as either the underlying cause of
death or other causes of death (up to 20) on the death certificate.
All internal causes of death were those who died of heat-related
causes or any of the following causes of death listed as either the
underlying cause of death or other causes of death (up to 20):
cardiovascular (I00–I99), respiratory (J00–J99), acute kidney failure
and chronic kidney disease (N17–N19), disorders of fluid, electrolyte, and acid-base balance (E87), dehydration (E86) and diabetes
(E08–E13). These causes increase during heat waves (Basu, 2009;
Knowlton et al., 2009; Reid et al., 2012). Frequencies of death from
all internal causes and heat-related illness were generated for each
census tract per day on the heat days and control days which are
defined below (see Fig. 1).
Thirteen census variables previously shown to contribute to
social vulnerability during heat waves were selected as the socioeconomic predictor variables (Knowlton et al., 2009; Basu, 2009;
Reid et al., 2009; Semenza et al., 1996; Harlan et al., 2006; Petitti
et al., 2013; Cooley et al., 2012; Uejio et al., 2011; Chow et al.,
2012). The 2012 American Community Survey's 5-year estimates
were used to extract census-level data for the 911 census tracts in
Maricopa County. The following socioeconomic variables were
considered for analysis: percent female householder (no husband
present), percent householder living alone, percent householder
living alone (4 65 years old), percent foreign born, percent who
work in construction, percent who work in agriculture, forestry,
fishing and hunting, and mining, percent of people whose income
in last 12 months is below poverty level, percent of households
with no vehicles available, percent 65 years or older, percent
Hispanic or Latino, percent uninsured, percent female and percent
renter households.
Three infrastructural variables were studied as predictor variables: access to public cooled spaces, home air-conditioning and
thermal protection of residential buildings. Fraser et al. (2016)
developed an index which describes neighborhood access to
publically available cooled space in Maricopa County. This access
to cooled spaces index, developed for individual households,
considers walkable access to different types of spaces which could
provide air-conditioning and heat relief to large numbers of people
including official cooling centers, public libraries, shopping malls,
museums and restaurants (American Library Association, 2012).
The index was based on pedestrian access in order to address the
potential challenges those without motorized transportation, a
particularly vulnerable population subset, may face in finding airconditioned relief. In brief, the accessibility metric is based on the
distance between any pair of publicly accessible cooled space and
individual residential parcels and is defined by three parameters:
walking time, walking speed and the existing street network. The
75th percentile for walking duration for non-leisure trips based on
data from the National Household Travel Survey, 15 min, was selected as the maximum time for this analysis along with an
average walking speed of 3.5 km/h, representing the average

91

Fig. 1. Mortality rate from heath related causes of death and all internal causes of
death by census tract in Maricopa county, 2005–2010.

walking speed of an elderly adult. Thus, a distance of 0.89 km
assessed within a regional road network established the catchment area for each publicly accessible cooled space. ArcGIS's
Network Analyst tool generated the catchment areas for every
publicly accessible cooling location assuming individuals would
select the shortest path. Household-level accessibility measurements were aggregated at the census tract and the mean of
household accessibility scores characterizes neighborhood level
accessibility. Higher numbers are associated with greater access.
Aggregated to the census tract scale, the index demonstrates wide
variation in access to cooled spaces with accessibility greater in
the older, inner city core of Phoenix and lower in the surrounding
suburbs (see Fig. 2). Household central air conditioning was determined using the Maricopa Assessor's database.
The thermal protection of residential buildings was measured
using a building thermal index of Nahlik et al. The building thermal index can be described as the number of hours it takes for the
indoor temperature to reach an unsafe temperature in response to
outside air temperature, measured as degrees Celsius per hour.
The thermal properties of 39 housing prototypes were assessed

92

D.P. Eisenman et al. / Health & Place 41 (2016) 89–99

Fig. 2. Accessibility index and building vulnerability index scores by census tract,
Maricopa county.

using simulation modeling that measured how building internal
temperatures increase in response to outside heat, when air conditioning is not used, in an extreme heat event. A numerical index
of building vulnerability was computed and assigned to each of
the building prototypes based on the hours elapsed for each
structure to increase from 25 to 32 °C (78–90 °F). A greater index
value indicates buildings that take longer to warm up during
periods of extreme heat and are therefore more protective against
extreme heat. Wall materials (concrete block, wood frame), insulation thickness, window panes, attic insulation, roofing material and size of home contribute to its thermal characteristics. The
building prototypes were then matched to every equivalent existing building based on building type and vintage to assign a
building vulnerability index. Finally, results were aggregated to the
census-tract level. Census tracts with the highest building thermal
index score are mostly characterized by single family homes, with
the census tracts with the highest score, on average, located in the
older areas of Maricopa county (see Fig. 2).
Heat events were defined consistent with the simplest definition of extreme heat, which is summertime temperatures that are

substantially hotter and/or more humid than the historic average
for that location at that time of year (U.S. Centers for Disease
Control and Prevention, 2013). We used maximum temperatures
since they are well associated with a heat-mortality response,
though investigators have also found heat effects with mean daily
temperature so it is also likely that the various temperature metrics produce similar results (Chow et al., 2012; Basu, 2009; Anderson and Bell, 2009; Baccini et al., 2008; Michelozzi et al., 2006;
Davis et al., 2016). Gridded historical daily temperature data reported at 1/8° (12 km) published by Maurer et al. (2002) were
used. Maurer et al. (2002) provide the gridded data through 2010
on a project website. They estimated the gridded temperatures
from readings at NOAA Cooperative Observer stations. The station
readings were gridded to 12 km cells using the synergraphic
mapping system (SYMAP) algorithm. They fit the data using an
asymmetric spline to estimate daily maximum and minimum
temperatures. Extreme heat events were identified based on the
methodology by Meehl and Tebaldi (2004): (a) the 97.5th historical percentile of daily maximum temperature for summertime
months (hereafter referred to as T1), and (b) the 81st historical
percentile of daily maximum temperature for summertime
months (hereafter referred to as T2). Summertime months are
June through August. For each grid cell, extreme heat events are
identified based on: 1) Maximum daily temperature must be
above T1 for at least three consecutive days; 2) The average
maximum daily temperature for the period must be greater than
or equal to T2; and 3) The maximum daily temperature must be
above T2 for every day in the period.
As studies find that heat-health associations diminish after a
lag period of three days, we added a lag three days to the end of
each heat event to capture any deaths which may have been related to the heat event but did not occur during the heat event
(Baccini et al., 2008; Curriero et al., 2002; O'neill et al., 2005a;
Davis et al., 2016). If the time between two heat events was three
days or less, the lag days would therefore join the two heat events,
and it was treated as one single event. For each heat event, a
corresponding control period was defined as the period of time in
the weeks prior to the heat event of the same number of days on
the same days of the week. In the analysis we used the maximum
temperature of the day to evaluate the association between outcomes and heat exposure. Only data from heat events and control
periods were used in the analysis.
Principal components analysis with varimax rotation was used
to construct composite variables from the 13 socio-economic
variables. This reduces collinearity and helps avoid over-fitting the
regression models. Components were retained based on the Kaiser
criterion (eigenvalue 41).
First, we examined the relationships between the principal
components, the three technical infrastructure predictor variables
and the two mortality outcomes. Second, we performed regression
modeling of mortality from heat-related illnesses and mortality
from all internal causes. Because the outcomes are non-negative
count variables, Poisson log-linear models based on generalized
estimating equations method were used to model the number of
deaths per day per census tract in each outcome. No formal spatial
correlation was considered in this analysis, however, within census tract correlation was included to account for the repeated
measures in the data from the same census tract, similar to previous heat-mortality studies (Ostro et al., 2009; Medina-Ramon
and Schwartz, 2007; Uejio et al., 2011). To determine if the relationship between the independent variable and mortality is
modified by maximum temperature, the models included the
variable, the maximum temperature on that day in that census
tract and an interaction between the independent variables and
the maximum temperature:

D.P. Eisenman et al. / Health & Place 41 (2016) 89–99

log (μ ij)=α + β1*tmax + β2*compx + β3*tmax*compx
where μ ij is the expected number of heat related mortalities in
tract i on day j , tmax is the maximum temperature in tract i on day
j and compx is the calculated component score (x ∈ 1,2,3,4).
Using the maximum temperature in the models, instead of an
indicator of extreme heat day versus control day, increases the
power of detecting the association between outcomes and heat
exposure. All models were offset by population size of the census
tract to convert to per capita rate. The estimates from the main
effect–only models predict the estimated change in expected
mortality for a unit risk factor change. For example, a predictor
with an estimated IRR of 1.5 indicates that for each one unit increase in the predictor (e.g. Building Thermal Property Index,
percent of households with air conditioning, etc.), expected mortality increases by 50%. We standardized the components such
that a one unit change in the component is the change of one
standard deviation for the component in the study population. For
the models considering the interaction of temperature, a positive
IRR indicates that the mortality rate increases faster as the maximum temperature increases in census tracts with a higher value
of the predictor. For example, an estimated IRR of 1.5 resulting
from the interaction between a variable and temperature indicates
that as maximum temperature increases by one degree, the rate of
mortality increases 50% in census tracts with higher scores for that
variable. All results presented are interaction effects, except in the
Poisson regressions where all variables are presented.
Stepwise regression of the component scores was used to
create the model predicting the relationship of the principal
components of the social vulnerability variables to the number of
deaths per day. Technical infrastructure variables were added into
the model one at a time before building a full model. QIC (quasilikelihood under the independence model criterion) was calculated to facilitate model comparison. Models with lower QIC are
considered better models (Pan, 2001).

93

Table 2
Incident Rate Ratios (IRR) with 95% confidence intervals for interaction of variable
with maximum temperature,a by cause of mortality, Maricopa county, 2005–2010.
Variable

Socioeconomic vulnerability (factor 1)
Renters living alone, without vehicle (factor 2)
Older aged, female, living
alone (factor 3)
Agriculture, extraction industry (factor 4)
No air conditioning
Public cooled space
accessibility
Building thermal
protection

Mortality from all
internal causes

Mortality from heat
related illness

IRR (95%C.
I.)

P-value

IRR (95% C. P-value
I.)

1.03 (1.00,
1.06)
0.99 (0.97,
1.01)
0.99 (0.97,
1.01)
0.97 (0.94,
1.00)
1.02 (1.00,
1.03)
1.00 (0.99,
1.01)
1.00 (0.99,
1.00)

0.03

1.17 (1.00,
1.38)
0.91 (0.80,
1.02)
0.83 (0.75,
0.92)
1.3 (1.01,
1.69)
1.02 (0.99,
1.04)
0.99 (0.98,
1.00)
1.02 (0.99,
1.04)

0.54
0.29
0.08
0.01
0.34
0.97

0.05
0.11
0.0004
0.04
0.16
0.03
0.25

a
Maximum temperature centered at 40 degrees; IRR¼ incidence rate ratio; 95%
C.I ¼95% Confidence interval. Each model included predictor variable, maximum
temperature, and interaction of variable with maximum temperature. Only the
results of the interaction term are presented.

variables and together they explained 77% of the variability in the
dataset (Table 1). Factor One can be understood as a socioeconomic vulnerability component with the addition of lacking
health insurance, working outdoors in construction and female
headed households. Factor Two contains persons living alone,
renters, and those without vehicles. The commonality among
these variables is that they indicate persons at risk for social isolation due to living alone with less connection to their community
as renters and potentially less mobile transportation. The first two
components explained 60% of the variance. Factor Three is an
older aged, female living alone component and Factor Four is an
agricultural and extraction industry workers component.

3. Results
3.1. All-internal causes of mortality
The final data set of 49,394 census-tract/days (24, 697 censustract/heat days and an equal number of census-tract/control days)
yielded1584 deaths from all-internal causes (814 on heat days and
770 on control days) and 40 deaths from heat-related illness (35
on heat days and 5 on control days).
Four factors were extracted from the social vulnerability

As shown in Table 2, cooled space accessibility did not modify
the effect of temperature on all internal causes of mortality (IRR
1.00 95% CI 0.99, 1.01; p ¼0.34) nor did the building thermal
protection index modify the effect of temperature on all internal
causes of mortality (IRR 1.00 95% CI 0.99, 1.00; p ¼0.97). The

Table 1
Principal components from socioeconomic variables.

Percent Hispanic/Latino
Percent foreign born
Percent uninsured
Income below poverty level
Work in construction
Female householder, no husband
present
Householder living alone
Percent renter households
No vehicles available
Householder living alone 4 65 years
old
Percent 465 years old
Percent female
Works in agriculture, forestry, fishing,
hunting, mining
Percent of variance explained

Factor 1: socioeconomic
vulnerability

Factor 2: renters living
alone without vehicle

Factor 3: older age/
female/living alone

Factor 4: agriculture or
extraction industry

0.90
0.84
0.82
0.76
0.74
0.70

0.09
0.17
0.31
0.50
 0.09
0.10

 0.18
 0.12
 0.20
 0.10
 0.06
 0.23

0.10
0.07
0.09
0.07
0.13
 0.27

 0.17
0.40
0.42
 0.16

0.84
0.78
0.76
0.28

0.33
 0.25
0.05
0.90

0.01
 0.08
0.04
0.08

 0.28
 0.05
0.11

0.02
 0.24
 0.03

0.88
0.62
0.00

0.10
 0.37
0.88

41%

19%

9%

8%

0.002
0.895

13,700.1

1.01 (1.00, 1.02)

0.90 (0.81, 1.00)
1.40 (1.29, 1.52)
1.71 (1.61, 1.82)
1.04 (0.93, 1.16)
1.00 (0.98, 1.02)
1.03 (1.00, 1.06)
0.97 (0.94, 1.00)
0.96 (0.91, 1.02)
1.01 (1.00, 1.03)
0.024
o 0.001
o 0.001
0.613
0.854
0.012
0.034

Missing components and variables were not chosen by model building procedure.
Maximum temperature (Tmax) centered at 40°.

13,794.0

0.349
0.752
1.01 (0.99, 1.05)
1.00 (0.99, 1.01)

1.01 (1.00, 1.02)
0.99 (0.99, 1.00)
13,730.1

0.89 (0.80, 0.98)
1.40 (1.29, 1.52)
1.71 (1.60, 1.82)
1.03 (0.92, 1.15)
1.01 (0.92, 1.11)
1.03 (1.01, 1.06)
0.97 (0.94, 1.00)
0.043
o 0.001
o 0.001
0.420
0.975
0.024
0.005
0.90 (0.82, 1.00)
1.31 (1.19, 1.44)
1.75 (1.64, 1.86)
1.05 (0.94, 1.18)
1.00 (0.98, 1.02)
1.03 (1.00, 1.06)
0.96 (0.93, 0.99)

IRR (95% CI)
P-value

0.054
o 0.001
o 0.001
0.542
0.971
0.040
0.020
0.202
0.118
–
–
0.001
–

b

When the outcome was mortality from heat-related illness,
there was a significant negative interaction between cooled space
accessibility and maximum temperature (IRR ¼ 0.99, 95%CI 0.98,
1.00; p ¼0.03) indicating that mortality from heat related illness
increases faster as the maximum temperature increases in census
tracts with less cooled space accessibility (Table 2). This is illustrated in Fig. 3.
In census tracts at the 25th percentile of cooled space accessibility, that is census tracts at the lowest end of the distribution,
the IRR is 1.38 (95% CI ¼1.12, 1.69; p¼ 0.002), indicating that for
every increase in maximum temperature by one degree Celsius,
there was a 38% increase in deaths from heat-related illness. In
census tracts at the 75th percentile, that is census tracts with

a

3.2. Heat-related illness mortality

0.106
o 0.001
o 0.001
0.334
0.831
0.047
0.003
0.103
0.093

interaction between maximum temperature and absence of home
air conditioning was significantly associated with risk of death
from all internal causes (IRR 1.02 95% CI 1.00, 1.03; p ¼0.01). This
means that a one percent rise in the proportion without air-conditioning increases the incidence rate ratio of deaths from all internal causes due to increased maximum temperature by two
percent. Socioeconomic vulnerability (factor 1) was also positively
associated with all internal causes of mortality in the interaction
model (IRR 1.03 95% CI 1.00, 1.06; p ¼0.03). Table 3 provides the
IRR at the 25th, 75th, and 90th percentiles for the factors that have
a significant interaction with temperature. As shown in Table 3,
the increase is most apparent in census tracts at the 90th percentile of the distribution of factor 1 scores (IRR 1.05 95% CI 1.00,
1.09; p ¼0.04).
In the multivariable analyses for all internal causes of mortality
(Table 4), Model One shows that the interaction between the socioeconomic vulnerability factor and temperature is significantly
associated with an increased mortality (IRR ¼1.02, 95% CI 1.00,
1.05; p ¼0.047). In other words, as maximum temperature increases, deaths from all internal causes increase more in census
tracts with higher index of socioeconomic vulnerability (factor 1).
In the full model, as the building thermal protection index increases mortality increases (main effect IRR ¼1.01, 95% CI 1.00–
1.02; p ¼0.001) but there is no association with the interaction
between maximum temperature and building thermal protection.
The QIC indicates that the full model has the best fit.

Socioeconomic vulnerability (factor 1)
0.92 (0.82, 1.02)
Renters living alone, without vehicle (factor 2) 1.35 (1.25, 1.47)
Older aged, female, living alone (factor 3)
1.74 (1.63, 1.85)
Agriculture, extraction industry (factor 4)
1.06 (0.94, 1.19)
Tmaxb
1.00 (0.98, 1.02)
Factor 1  Tmax
1.02 (1.00, 1.05)
Factor 4  Tmax
0.96 (0.93, 0.99)
No air conditioning
0.95 (0.89, 1.01)
No air conditioning  Tmax
1.01 (1.00, 1.03)
Cooled space accessibility
Cooled space accessibility  Tmax
Building thermal protection
Building thermal protection  Tmax
QIC
13,760.3

Maximum temperature centered at 40°.

IRR (95% CI)

0.01
o 0.001
o 0.001

P-value

o 0.001
0.002
0.06

IRR (95% CI)

0.24
0.79
0.27

P-value

a

0.43
0.08
0.04

IRR (95% CI)

All internal causes

Table 4
Mortality from all internal causes: Incident Rate Ratios (IRR) with 95% confidence intervals from Poisson multivariate regression, Maricopa county, 2005–2010.

Socioeconomic vulnerability (factor 1)
(25th percentile)
0.99 (0.97, 1.01)
(75th percentile)
1.02 (0.99, 1.05)
(90th percentile)
1.05 (1.00, 1.09)
Heat-related illness Socioeconomic vulnerability (factor 1)
(25th percentile)
1.16 (0.90, 1.49)
(75th percentile)
1.38 (1.15, 1.66)
(90th percentile)
1.58 (1.25, 1.99)
Older aged, female, living alone (factor 3)
(25th percentile)
1.51 (1.24, 1.83)
(75th percentile)
1.36 (1.12, 1.63)
(90th percentile)
1.20 (0.97, 1.48)
Agriculture, extraction industry (factor 4)
(25th percentile)
1.30 (1.06, 1.60)
(75th percentile)
1.43 (1.18, 1.75)
(90th percentile)
1.73 (1.31, 2.29)

P-value

Variablea

IRR (95% CI)

Model 3: socio-vulnerability þ building thermal
protection

Component with
percentiles

Model 2: socio-vulnerability þcool space
access

Mortality cause

Model 4: full model

Table 3
Incident Rate Ratios (IRR) with 95% confidence intervals for interaction of significant social vulnerability components at given percentiles with maximum
temperature.a

P-value

D.P. Eisenman et al. / Health & Place 41 (2016) 89–99

Model 1: socio-vulnerability þ air
conditioning

94

D.P. Eisenman et al. / Health & Place 41 (2016) 89–99

Fig. 3. Effect of cooled space accessibility and temperature on heat-related deaths
in Maricopa county.

relatively greater accessibility, the IRR was 1.36 (95% CI ¼1.11, 1.66;
p ¼0.003) and at the 90th percentile of cooled space accessibility,
the IRR was 1.31 (95% CI ¼ 1.07, 1.60; p ¼ 0.007). In contrast, we
found no association between the interaction of maximum temperature with households without air-conditioning and mortality
from heat-related illness (Table 2).
There was a significant negative interaction between the older
aged female living alone factor, and maximum temperature
(IRR ¼ 0.83; 95%CI 0.75-0.92; P-value o0.001) indicating that the
rate of mortality from heat-related illnesses increases more slowly
as the maximum temperature increases among census tracts with
more older aged females living alone (Table 2). In other words,
census tracts with a larger proportion of older aged females living
alone experienced less of a rise in fatalities as maximum temperatures increased than did tracts with smaller proportions of
this group. As shown in Table 3 and illustrated in Fig. 4, in census

95

tracts at the 25th percentile of the older aged female living alone
factor, the IRR was 1.51 (95% CI ¼1.24, 1.83; p o0.001), indicating
that for every degree Celsius increase in maximum temperature,
there was a 51% increase in deaths from heat-related illness. In
census tracts at the 75th percentile, the IRR was lower, 1.36 (95%
CI ¼1.12, 1.63; p ¼0.002), and in census tracts at the 90th percentile of the older aged female living alone factor—that is census
tracts with the greatest proportion of older aged females living
alone, the IRR was still lower, 1.20 (95% CI ¼0.97, 1.48; p ¼ 0.06).
The interactions between maximum temperature and the socioeconomic vulnerability factor (IRR ¼1.17, 95%CI 1.00, 1.38;
p ¼0.05) and agriculture/extraction industry factor (IRR ¼1.30, 95%
CI 1.01, 1.69; p ¼0.04) were associated with increased mortality
from heat-related illness (Table 3). As shown in Table 3, census
tracts in the 90th percentile of the distribution of the agriculture/
extraction industry factor have significantly greater death from
heat-related illnesses compared to census tracts in the 75th and
25th percentiles.
In the multivariable analysis of mortality from heat related
illnesses (Table 5), Model One shows socioeconomic vulnerability
and advanced aged females living alone are positive main effect
predictors of heat related illness mortality and that the interaction
between agricultural/extraction worker and temperature leads to
increased mortality from heat-related illnesses (IRR ¼1.33, 95%CI
1.05, 1.69; p ¼0.020). As maximum temperature increases mortality from heat related illnesses increases more in census tracts
with higher agricultural/extraction worker indices. In contrast, as
maximum temperature increases, mortality from heat related illnesses increases less in census tracts with higher advanced aged
females living alone indices (IRR ¼0.86, 95% CI 0.79, 0.95;
p ¼0.002). The direction and magnitude of the factors three and
four interaction results do not change as further infrastructure
variables are added into the model. Model Two shows that the
interaction between accessible cooled spaces and temperature
reduces the risk of death from heat-related illnesses (IRR ¼ 0.98,
95%CI 0.96, 1.00; p ¼ 0.024). In other words, as maximum temperature increases, increasing access to public cooled spaces is
associated with slower increases in the mortality rate from heat
related illnesses, controlling for the social vulnerability of a census
tract. The QIC indicates that Model Two has the best fit. Model
Three shows that building thermal protection is not a predictor of
mortality from heat related illness when controlling for social
vulnerability. The importance of all the social vulnerability factors
to heat-related illness mortality did not change across all models.

4. Discussion

Fig. 4. Effect of factor 3 and temperature on heat-related deaths in Maricopa
county.

We found mixed support for the hypothesis that three infrastructure resources that could provide adaptive capacity in Maricopa County reduce mortality in extreme heat events. As maximum temperature increases, mortality from heat-related illness
increased less in census tracts with more publicly accessible
cooled spaces, controlling for social vulnerability (IRR 0.98, 95% CI
0.980, 0.999; p¼ 0.027). This suggests that publicly accessible
cooled space provides adaptive capacity when the metric is mortality from heat-related illness.
However, publicly accessible cooled space is not significantly
associated with all internal causes of mortality as maximum
temperatures increase. Indeed, throughout our analyses, the results differed based on the cause of mortality used as an outcome,
consistent with research showing that the heat-health relationship
depends on the health outcome assessed (Kovats et al., 2004;
Petitti et al., 2016). One explanation for these discrepant findings is
that the all-internal cause grouping may be too non-specific to
sensitively detect heat-related protective factors (Petitti et al.,

96

Table 5
Mortality from heat related illness: Incident Rate Ratios (IRR) from Poisson multivariate regression. Maricopa county, 2005–2010.
Model 2: socio-vulnerability þ cool space
access

Model 3: socio-vulnerability þbuilding thermal
protection

Model 4: full model

Variablea

IRR (95% CI)

P-value

IRR (95% CI)

P-value

IRR (95% CI)

P-value

IRR (95% CI)

P-value

Socioeconomic vulnerability (factor 1)

1.64 (1.17, 2.30)

0.004

1.70 (1.24, 2.31)

0.001

1.71 (1.22, 2.39)

0.002

0.002

Renters living alone, without vehicle (factor
2)
Older aged, female, living alone (factor 3)

1.20 (0.80, 1.80)

0.380

1.12 (0.72, 1.73)

0.627

1.11 (0.75, 1.63)

0.598

2.27 (1.50, 3.44)

0.001

2.40 (1.58, 3.64)

o 0.001

2.53 (1.68, 3.82)

o0.001

Agriculture, extraction industry (factor 4)

0.17 (0.03, 0.97)

0.045

0.17 (0.02, 1.19)

0.073

0.18 (0.03, 1.00)

0.050

Tmaxb

1.42 (1.16, 1.72)

0.001

1.49 (1.22, 1.81)

o 0.001

0.84 (0.43, 1.65)

0.620

Factor 3  Tmax

0.86 (0.79, 0.95)

0.002

0.85 (0.77, 0.94)

0.002

0.85 (0.77, 0.94)

0.001

Factor 4  Tmax

1.33 (1.05, 1.69)

0.020

1.34 (1.03, 1.75)

0.028

1.33 (1.06, 1.67)

0.016

1.73 (1.23,
2.44)
1.05 (0.68,
1.61)
2.52 (1.66,
3.84)
0.16 (0.02.
1.11)
1.50 (1.23.
1.83)
0.85 (0.77,
0.94)
1.34 (1.04,
1.74)

No air conditioning
No air conditioning  Tmax
Cooled space accessibility

1.05 (0.92, 1.21)
1.00 (0.98, 1.03)

0.450
0.792
1.12 (1.04, 1.22)

0.005

0.98 (0.96, 1.00)

0.024

Cooled space accessibility  Tmax
Building thermal protection
Building thermal protection  Tmax
QIC
a
b

603.9

Missing components and variables were not chosen by model building procedure.
Maximum temperature (Tmax) centered at 40°.

593.1

0.92 (0.85, 1.00)

0.052

1.01 (1.00, 1.03)
597.6

0.146

1.11 (1.00,
1.22)
0.98 (0.96,
1.00)
0.97 (0.94,
1.01)
620.2

0.823
o 0.001
0.064
o 0.001
0.002
0.023

0.042
0.030
0.146

D.P. Eisenman et al. / Health & Place 41 (2016) 89–99

Model 1: socio-vulnerability þair
conditioning

D.P. Eisenman et al. / Health & Place 41 (2016) 89–99

2016). Mortality from heat-related illness will be directly dependent on increasing maximum temperature whereas all-internal
causes of mortality may be exacerbated by high heat but may not
be directly dependent on it. Another possibility is that “officially”
designated extreme heat events may prompt urban residents to
use cooled spaces such as libraries and malls at the same time as
official announcements alert coroners to heat-related deaths that
would be otherwise misclassified (Shen et al., 1998). However, one
survey of public responses to extreme heat found only 14% of respondents from Phoenix reported going to a “cooler place” on a hot
day (Sheridan, 2007). Therefore, while use of publicly accessible
cooled spaces cannot be inferred from these results, our results
hint that access to this resource can reduce risk to socially vulnerable populations.
The building thermal protection index was not associated with
mortality in our analyses. In the multivariate models, census tracts
with buildings that are slower to heat up showed a decrease in
heat-related illness mortality (fixed effect IRR 0.92, 95%CI 0.85,
1.00; p ¼ 0.052) and an increase in all internal causes of mortality
(fixed effect IRR 1.01, 95%CI 1.00, 1.02; p ¼0.002) but the interactions with maximum temperature were not significant, thereby
calling into question the meaning of the fixed effect findings. We
would expect that the relationship between building thermal
properties and heat-related mortality would be sensitive to increasing maximum temperature; the environmental benefits of
buildings maintaining cool temperatures longer might translate to
mortality benefits, too. Research has shown a correlation between
older and poorer housing stock and increased mortality from extreme heat (Uejio et al., 2011; Klein Rosenthal et al., 2014). And
Hayden's report that 89% of their survey respondents had air
conditioning but 38% still experienced their homes as too hot inside suggests a protective benefit in improving thermal properties
(Hayden et al., 2011). But studies addressing the relationship of the
thermal envelope of a building to adverse events have not been
conducted until now. The buildings that heat up the quickest in
Maricopa County are the older single family homes that are concentrated in the older, central neighborhoods where poverty is
also concentrated, which may explain the fixed effect relationship
to all-internal causes of mortality. The relationship between this
adaptive functioning of a building's thermal protection and social
vulnerability at the census tract level is likely complex and may
include factors we did not measure such as the willingness to use
air-conditioning in the home. Also, the building index does not
consider the contributions of building orientation or tree cover to
potential cooling (Jenerette et al., 2011; Jenerette et al., 2007).
The interaction of not having air conditioning with increasing
temperature was a significant predictor of all internal causes of
mortality in the bivariate but not the multivariate models. In the
multivariate models predicting mortality from all internal causes,
the protective benefit of air conditioning was moderated by the
social vulnerability factors. One implication is that any effort to
reduce heat-mortality in Maricopa County by increasing air-conditioning availability must focus on the barriers to using air-conditioning. Previous research in Phoenix supports this (Hayden
et al., 2011; Sheridan, 2007). Current programs to provide air
conditioning to residents on fixed incomes are reported to reach
less than half of those requesting service (White-Newsome et al.,
2014).
Factors 1, 3 and 4 were associated with mortality though
sometimes in surprising directions. Socioeconomic vulnerability
(factor 1) was associated with mortality from all internal causes
and heat related illnesses as temperature increases and it was
robust to adjustment in the multivariate models. In contrast, the
interaction of maximum temperature with the older aged/living
alone factor reduces the risk of mortality from heat-related illness,
an interaction effect not previously demonstrated. The effect of

97

increasing maximum temperature on heat-related illness mortality declines as the proportion of a census tract that is female, age
over 65 and living alone increases; census tracts with higher
proportions have less sensitivity to increasing heat than census
tracts with lesser proportions. Few studies have investigated if the
relationship between a predictor variable and mortality is dependent on temperature, which has the advantage of addressing
relationships to continuous temperature that would be missed by
restricting the analysis to the absolute categories of heat days and
control days. This captures how sensitive the variables are to increasing temperature. For example, regarding the fixed effect IRR,
the older aged/living alone factor was positively associated with
increased mortality from all internal causes and heat-related illness in the multivariate analyses. This is consistent with Harlan
and colleagues’ finding that a factor made up of advanced age,
living alone and the interaction of age and living alone predicted
heat-related death in Phoenix (Harlan et al., 2013). Hondula also
reported that a factor comprised of percent living alone, percent
living alone over age 65 and medium intensity development predicted excess mortality in Phoenix (Hondula et al., 2015). Lastly,
Chuang's study in Phoenix also found that a factor containing
proportion of elderly and rate of hospitalization for diabetes predicted heat-related hospitalizations(Chuang and Gober, 2015). But,
our interaction results are also consistent with findings that the
number of heat distress calls was inversely related to the proportion of older residents in a census block group in Phoenix
(Uejio et al., 2011). These results suggest that adaptations in
Phoenix may be having some benefit, though which adaptations
cannot be discerned. For instance, as part of the city's heat health
warning system, emergency managers contact assisted living facilities with warnings. Also, Phoenix has many retirement communities including communities where age specific zoning restrictions mandate every household must include an older adult
(Gober, 2006). The city has a public information campaign each
summer in which “the big thing with the seniors is watch out for
your neighbors so you know if you don’t see your neighbor”
(White-Newsome et al., 2014). It is possible that some combination of protective behaviors (including traveling to cooler climates,
so called “snowbirds”), social networks and social service resources
among Phoenix's seniors, retirement communities and assisted
living facilities may explain these findings. Thus, this study raises
all new questions for research into adaptations for extreme heat.
As for the agricultural/extraction industry worker factor, our
study confirms and extends prior findings of an association between heat-related mortality and outdoor workers (Petitti et al.,
2013; Davis et al., 2003a). Not only was mortality from heat-related illness higher in census tracts with more such workers, but
as maximum temperature increased excess mortality from heatrelated illness increased too. This association remained robust
adjusting for the other social vulnerability components. We were
surprised to find a negative interaction between the agricultural/
extraction industry worker factor and maximum temperature in
the multivariate models of all internal causes of mortality. As with
the results of the older aged/living alone factor, further inquiry is
needed to understand the finding.
There are several limitations to our analysis. We did not assess
the predictive ability of our models, though vulnerability models
have mixed success in predicting heat-related mortality (Reid
et al., 2012). The small number of deaths from heat-illness further
limits our ability to make meaningful predictions. However, the
relationship between heat and health is complex such that temperature-mortality relationships may differ depending on the
health event measured. Petitti and colleagues found that mortality
from the direct effects of heat consistently occurred at lower
temperatures than did all-cause mortality in Maricopa County
(Petitti et al., 2016). Accessible cooled spaces and official cooling

98

D.P. Eisenman et al. / Health & Place 41 (2016) 89–99

centers may be particularly important for the homeless and our
measure of accessibility, based on residential parcels, cannot
capture accessibility for the homeless. It is not uncommon in
Phoenix for homeless persons to ride the air-conditioned public
transit for heat relief (Sanchez, 2011). The census variables used to
define vulnerability are also coarse indicators of vulnerability.
While being older than 65 years may confer statistically significant
risk, clinically meaningful risk may occur mostly to frail elderly
over 75, or older ages, such as 85 and older. Living alone may be an
inadequate proxy for the level of social isolation that elder “shutins” face (Klinenberg, 2002; Bassil and Cole, 2010; Foroni et al.,
2007). Lastly, while this analysis provides census-level associations that may be useful for community planning, it does not necessarily explain individual-level factors. Nesting individual factors
inside group level factors would provide a more nuanced understanding of the interaction between vulnerable residents and
vulnerable neighborhoods.
This study has implications for public health adaptation planning. To reduce heat-related mortality, the use of public cooled
spaces might be expanded to target the most vulnerable. More
research is needed to understand how publicly accessible cooled
spaces might become a feasible and acceptable element of adaptation planning. Community programs that mobilize residents to
public spaces, such as occurs in city parks and during bicycle
events are increasingly used in public health and civic engagement. Such programs might hold the ingredients for designing a
private-public partnership program that gets people from vulnerable neighborhoods into public cooled spaces such as shopping
malls during extreme heat days. Getting more people to use their
home air-conditioning during extreme heat days is a worthy
adaptation for short-term responses to these events but it is
coming under increasing scrutiny as a long-term adaptation (Hess
et al., 2012). Extreme heat of increasing frequency, duration and
severity combined with predictions of Southwest water shortages
raises the danger of grid failure. The risk increases still further
when one considers the hazards posed by wildfires at the urbanwilderness interface occurring during times of drought and extreme heat and threatening electrical grids. The protective effect of
increasing the concentration of older aged adults living alone in a
census tract raises interesting possibilities that adaptive behaviors
are at work that have not previously been captured. Understanding the adaptive behaviors could enhance public health and
safety programs for extreme heat.

5. Conclusions
We studied three elements of the infrastructure of Maricopa
County to examine their potential benefit as adaptive resources to
extreme heat and found mixed results. As temperatures increase,
mortality from heat-related illnesses in a census tract decreases
with increasing accessibility to publicly available cooled spaces.
However, the mortality from all-internal causes was not associated
with improved access. Nor did an index measuring the thermal
ability of a building to remain cool in the heat relate to mortality
with increasing temperature. Mortality rates increase as temperature increases in Maricopa County census tracts that have
higher proportions of Hispanic households living in poverty and
without health insurance. This association is not ameliorated as
the proportion of households with air-conditioning increases,
which further supports the need to remove social and financial
obstacles to air-conditioner use in these neighborhoods. As the
proportion of a census tract that is older aged females living alone
increases, the effect of increasing temperature on mortality rates
from heat-related illness decreases and this relationship is similarly true for agricultural workers and all-internal causes of

mortality. These two surprising findings call for more fine-grained
studies looking for the potential adaptive mechanisms at work.
Overall, this study highlights the value of research on the interplay
between vulnerability and adaptive capacity at the city-level and
the need to uncover the adaptive capacities already in play in our
cities.

Acknowledgments
Funding provided by the National Science Foundation, United
States Award nos. 1335556, 1335640 and 1444755 which did not
otherwise contribute to conducting the research.

References
American Library Association 2012. Extending Our Reach: Reducing Homelessness
through Library Engagement.
Anderson, B.G., Bell, M.L., 2009. Weather-related mortality: how heat, cold, and
heat waves affect mortality in the United States. Epidemiology 20, 205–213.
Anderson, G.B., Bell, M.L., 2011. Heat waves in the United States: mortality risk
during heat waves and effect modification by heat wave characteristics in 43 U.
S. communities. Environ. Health Perspect. 119, 210–218.
Anderson, M., Carmichael, C., Murray, V., Dengel, A., Swainson, M., 2013. Defining
indoor heat thresholds for health in the UK. Perspect. Public Health 133,
158–164.
Baccini, M., Biggeri, A., Accetta, G., Kosatsky, T., Katsouyanni, K., Analitis, A., Anderson, H.R., Bisanti, L., D'ippoliti, D., Danova, J., Forsberg, B., Medina, S., Paldy,
A., Rabczenko, D., Schindler, C., Michelozzi, P., 2008. Heat effects on mortality in
15 European cities. Epidemiology 19, 711–719.
Bassil, K.L., Cole, D.C., 2010. Effectiveness of public health interventions in reducing
morbidity and mortality during heat episodes: a structured review. Int. J. Environ. Res. Public Health 7, 991–1001.
Basu, R., 2009. High ambient temperature and mortality: a review of epidemiologic
studies from 2001 to 2008. Environ. Health 8, 40.
Blum, L.N., Bresolin, L.B., Williams, M.A., 1998. From the AMA council on scientific
affairs. Heat-related illness during extreme weather emergencies. J. Am. Med.
Assoc. 279, 1514.
Chow, W.T.L., Chuang, W.C., Gober, P., 2012. Vulnerability to extreme heat in metropolitan phoenix: spatial, temporal, and demographic dimension. Prof. Geogr.
64, 286–302.
Chuang, W.C., Gober, P., 2015. Predicting hospitalization for heat-related illness at
the census-tract level: accuracy of a generic heat vulnerability index in Phoenix, Arizona (USA). Environ. Health Perspect. 123, 606–612.
Cooley, H.E., Moore, E., Heberger, M., Allen, L., 2012. Social vulnerability to climate
change in california. California. Calif. Energy Comm.
Curriero, F.C., Heiner, K.S., Samet, J.M., Zeger, S.L., Strug, L., Patz, J.A., 2002. Temperature and mortality in 11 cities of the eastern United States. Am. J. Epidemiol. 155, 80–87.
Davis, R.E., Hondula, D.M., Patel, A.P., 2016. Temperature observation time and type
influence estimates of heat-related mortality in seven U.S. cities. Environ.
Health Perspect. 124, 795–804.
Davis, R.E., Knappenberger, P.C., Michaels, P.J., Novicoff, W.M., 2003a. Changing
heat-related mortality in the United States. Environ. Health Perspect. 111,
1712–1718.
Davis, R.E., Knappenberger, P.C., Novicoff, W.M., Michaels, P.J., 2003b. Decadal
changes in summer mortality in U.S. cities. Int. J. Biometeorol. 47, 166–175.
Ebi, K., Teisberg, T., Kalkstein, L., Robinson, L., Weiher, R., 2003. Heat watch/warning
systems save lives: estimated costs and benefits for Philadelphia 1995–1998:
ISEE-165. Epidemiology, 14.
Foroni, M., Salvioli, G., Rielli, R., Goldoni, C.A., Orlandi, G., Zauli Sajani, S., Guerzoni,
A., Maccaferri, C., Daya, G., Mussi, C., 2007. A retrospective study on heat-related mortality in an elderly population during the 2003 heat wave in Modena,
Italy: the Argento project. J. Gerontol. A: Biol. Sci. Med. Sci. 62, 647–651.
Fraser, A., Chester, M., Eisenman, D., Hondula, D.M., Pincetl, S.S., English, P., Bondank, E., 2016. Household accessibility to heat refuges: residential air conditioning, public cooled space and walkability. Environ. Plan., http://dx.doi.org/
10.1177/0265813516657342, In press
Gober, P., 2006. Metropolitan Phoenix: Place Making and Community Building in
the Desert. University of Pennsylvania Press, Philadelphia.
Golden, J.S., Hartz, D., Brazel, A., Luber, G., Phelan, P., 2008. A biometeorology study
of climate and heat-related morbidity in Phoenix from 2001 to 2006. Int. J.
Biometeorol. 52, 471–480.
Harlan, S.L., Declet-Barreto, J.H., Stefanov, W.L., Petitti, D.B., 2013. Neighborhood
effects on heat deaths: social and environmental predictors of vulnerability in
Maricopa County, Arizona. Environ. Health Perspect. 121, 197–204.
Harlan, S.L., Brazel, A.J., Prashad, L., Stefanov, W.L., Larsen, L., 2006. Neighborhood
microclimates and vulnerability to heat stress. Soc. Sci. Med. 63, 2847–2863.
Hayden, M., Brenkert-Smith, H., Wilhelmi, O, H., 2011. Differential adaptive capacity

D.P. Eisenman et al. / Health & Place 41 (2016) 89–99

to extreme heat: a phoenix, arizona, case study. Am. Meteorol. Soc. 3, 269–280.
Hess, J.J., Mcdowell, J.Z., Luber, G., 2012. Integrating climate change adaptation into
public health practice: using adaptive management to increase adaptive capacity and build resilience. Environ. Health Perspect. 120, 171–179.
Hondula, D.M., Davis, R.E., Saha, M.V., Wegner, C.R., Veazey, L.M., 2015. Geographic
dimensions of heat-related mortality in seven U.S. cities. Environ. Res. 138,
439–452.
Hondula, D.M., Davis, R.E., Leisten, M.J., Saha, M.V., Veazey, L.M., Wegner, C.R., 2012.
Fine-scale spatial variability of heat-related mortality in Philadelphia County,
USA, from 1983–2008: a case-series analysis. Environ. Health 11, 16.
Huang, C., Vaneckova, P., Wang, X., Fitzgerald, G., Guo, Y., Tong, S., 2011. Constraints
and barriers to public health adaptation to climate change: a review of the
literature. Am. J. Prev. Med. 40, 183–190.
Jenerette, G.D., Harlan, S.L., Stefanov, W.L., Martin, C.A., 2011. Ecosystem services
and urban heat riskscape moderation: water, green spaces, and social inequality in Phoenix, USA. Ecol. Appl. 21, 2637–2651.
Jenerette, G.D., Harlan, S., Brazel, A., Jones, N., Larsen, L., Stefanov, W., 2007. Regional relationships between surface temperature, vegetation, and human
settlement in a rapidly urbanizing ecosystem. Landsc. Ecol. 22, 353–365.
Kalkstein, A.J., Sheridan, S.C., 2007. The social impacts of the heat-health watch/
warning system in Phoenix, Arizona: assessing the perceived risk and response
of the public. Int. J. Biometeorol. 52, 43–55.
Klein Rosenthal, J., Kinney, P.L., Metzger, K.B., 2014. Intra-urban vulnerability to
heat-related mortality in New York City, 1997–2006. Health Place 30, 45–60.
Klinenberg, E., 2002. Heat Wave: A Social Autopsy of Disaster in Chicago. University
of Chicago Press, Chicago.
Knowlton, K., Rotkin-Ellman, M., King, G., Margolis, H.G., Smith, D., Solomon, G.,
Trent, R., English, P., 2009. The 2006 California heat wave: impacts on hospitalizations and emergency department visits. Environ. Health Perspect. 117,
61–67.
Kovats, R.S., Hajat, S., 2008. Heat stress and public health: a critical review. Annu.
Rev. Public Health 29, 41–55.
Kovats, R.S., Hajat, S., Wilkinson, P., 2004. Contrasting patterns of mortality and
hospital admissions during hot weather and heat waves in Greater London, UK.
Occup. Environ. Med. 61, 893–898.
Lane, K., Wheeler, K., Charles-Guzman, K., Ahmed, M., Blum, M., Gregory, K., Graber,
N., Clark, N., Matte, T., 2014. Extreme heat awareness and protective behaviors
in New York City. J. Urban Health 91, 403–414.
Le Tertre, A., Lefranc, A., Eilstein, D., Declercq, C., Medina, S., Blanchard, M., Chardon,
B., Fabre, P., Filleul, L., Jusot, J.F., Pascal, L., Prouvost, H., Cassadou, S., Ledrans, M.,
2006. Impact of the 2003 heatwave on all-cause mortality in 9 French cities.
Epidemiology 17, 75–79.
Luber, G., Mcgeehin, M., 2008. Climate change and extreme heat events. Am. J. Prev.
Med. 35, 429–435.
Maurer, E.P., Wood, A.W., Adam, J.C., Lettenmaier, D.P., 2002. A long-term hydrologically based dataset of land surface fluxes and states for the conterminous
United States. J. Clim. 15, 3237–3251.
Mavrogianni, A., Wilkinson, P., Davies, M., Biddulph, P., Oikonomou, E., 2012.
Building characteristics as determinants of propensity to high indoor summer
temperatures in London dwellings. Build. Environ. 55, 117–130.
Mcgeehin, M.A., Mirabelli, M., 2001. The potential impacts of climate variability and
change on temperature-related morbidity and mortality in the United States.
Environ. Health Perspect. 109 (Suppl 2), 185–189.
Medina-Ramon, M., Schwartz, J., 2007. Temperature, temperature extremes, and
mortality: a study of acclimatisation and effect modification in 50 US cities.
Occup. Environ. Med. 64, 827–833.
Meehl, G.A., Tebaldi, C., 2004. More intense, more frequent, and longer lasting heat
waves in the 21st century. Science 305, 994–997.
Mees, H.L.P., Driessen, Peter, P.J., Runhaar, Hens, A.C., 2015. "Cool" governance of a
"hot" climate issue: public and private responsibilities for the protection of
vulnerable citizens against extreme heat. Reg. Environ. Change 15, 1065–1079.
Michelozzi, P., DE Sario, M., Accetta, G., De'donato, F., Kirchmayer, U., D'ovidio, M.,
Perucci, C.A., Group, H.C., 2006. Temperature and summer mortality: geographical and temporal variations in four Italian cities. J. Epidemiol. Community

99

Health 60, 417–423.
Nahlik, M., Chester, M.V., Pincetl, S.S., Eisenman, D.P., Sivaraman, D., English, P.
English Building Thermal Performance Varies during Extreme Heat within Cities [Online]. Arizona State University: Arizona State University Digital Repository. Available: 〈http://hdl.handle.net/2286/R.I.34957〉.
O'neill, M.S., Zanobetti, A., Schwartz, J., 2005b. Disparities by race in heat-related
mortality in four US cities: the role of air conditioning prevalence. J. Urban
Health 82, 191–197.
O'neill, M.S., Hajat, S., Zanobetti, A., Ramirez-Aguilar, M., Schwartz, J., 2005a. Impact
of control for air pollution and respiratory epidemics on the estimated associations of temperature and daily mortality. Int. J. Biometeorol. 50, 121–129.
Ostro, B., Rauch, S., Green, R., Malig, B., Basu, R., 2010. The effects of temperature
and use of air conditioning on hospitalizations. Am. J. Epidemiol. 172,
1053–1061.
Ostro, B.D., Roth, L.A., Green, R.S., Basu, R., 2009. Estimating the mortality effect of
the July 2006 California heat wave. Environ. Res. 109, 614–619.
Pan, W., 2001. Akaike's information criterion in generalized estimating equations.
Biometrics 57, 120–125.
Petitti, D.B., Hondula, D.M., Yang, S., Harlan, S.L., Chowell, G., 2016. Multiple trigger
points for quantifying heat-health impacts: new evidence from a hot climate.
Environ. Health Perspect. 124, 176–183.
Petitti, D.B., Harlan, S.L., Chowell-Puente, G., Ruddell, D., 2013. Occupation and
environmental heat-associated deaths in Maricopa county, Arizona: a casecontrol study. PLoS One 8, e62596.
Reid, C.E., O'neill, M.S., Gronlund, C.J., Brines, S.J., Brown, D.G., Diez-Roux, A.V.,
Schwartz, J., 2009. Mapping community determinants of heat vulnerability.
Environ. Health Perspect. 117, 1730–1736.
Reid, C.E., Mann, J.K., Alfasso, R., English, P.B., King, G.C., Lincoln, R.A., Margolis, H.G.,
Rubado, D.J., Sabato, J.E., West, N.L., Woods, B., Navarro, K.M., Balmes, J.R., 2012.
Evaluation of a heat vulnerability index on abnormally hot days: an environmental public health tracking study. Environ. Health Perspect. 120, 715–720.
Sanchez, C., 2011. Tricks of the Shade: Heat Related Coping Strategies of Urban
Homeless Persons in Phoenix. Arizona State University, Arizona.
Sathaye J., Dale L., Larsen P., Fitts G., Koy K., Lewis S., Lucena A., 2012. Estimating
Risk to California Energy Infrastructure From Projected Climate Change. California Energy Commission. Publication Number: CEC‐500‐ 2012‐057.
Semenza, J.C., Rubin, C.H., Falter, K.H., Selanikio, J.D., Flanders, W.D., Howe, H.L.,
Wilhelm, J.L., 1996. Heat-related deaths during the July 1995 heat wave in
Chicago. N. Engl. J. Med. 335, 84–90.
Shen, T., Howe, H.L., Alo, C., Moolenaar, R.L., 1998. Toward a broader definition of
heat-related death: comparison of mortality estimates from medical examiners'
classification with those from total death differentials during the July 1995 heat
wave in Chicago, Illinois. Am. J. Forensic Med. Pathol. 19, 113–118.
Sheridan, S.C., 2007. A survey of public perception and response to heat warnings
across four North American cities: an evaluation of municipal effectiveness. Int.
J. Biometeorol. 52, 3–15.
U.S. Centers for Disease Control and Prevention. 2013. Climate Change and Extreme
Heat Events [Online]. Available: 〈http://emergency.cdc.gov/disasters/ex
tremeheat/heat_guide.asp〉 (accessed 29.04.15).
U.S. Centers for Disease Control and Prevention. 2015. Extreme Heat Prevention
Guide [Online]. Office of Public Health Preparedness and Response (OPHPR)
Available: 〈http://emergency.cdc.gov/disasters/extremeheat/heat_guide.asp〉
(accessed 17.08.15).
Uejio, C.K., Wilhelmi, O.V., Golden, J.S., Mills, D.M., Gulino, S.P., Samenow, J.P., 2011.
Intra-urban societal vulnerability to extreme heat: the role of heat exposure
and the built environment, socioeconomics, and neighborhood stability. Health
Place 17, 498–507.
United Nations 2014. World Urbanization Prospects: The 2014 Revision, Highlights
New York.
White-Newsome, J.L., Mccormick, S., Sampson, N., Buxton, M.A., O'neill, M.S.,
Gronlund, C.J., Catalano, L., Conlon, K.C., Parker, E.A., 2014. Strategies to reduce
the harmful effects of extreme heat events: a four-city study. Int. J. Environ. Res.
Public Health 11, 1960–1988.