RES E AR C H A R T I C L E Open Access
The association between consecutive days’
heat wave and cardiovascular disease
mortality in Beijing, China
Qian Yin and Jinfeng Wang
*
Abstract
Background: Although many studies have examined the effects of heat waves on the excess mortality risk (ER) posed
by cardiovascular disease (CVD), scant attention has been paid to the effects of various combinations of differing heat
wave temperatures and durations. We investigated such effects in Beijing, a city of over 20 million residents.
Methods: A generalized additive model (GAM) was used to analyze the ER of consecutive days’ exposure to extreme
high temperatures.
Results: A key finding was that when extremely high temperatures occur continuously, at varying temperature
thresholds and durations, the adverse effects on CVD mortality vary significantly. The longer the heat wave lasts, the
greater the mortality risk is. When the daily maximum temperature exceeded 35 °C from the fourth day onward, the ER
attributed to consecutive days’ high temperature exposure saw an increase to about 10% (p <0.05),andatthefifth
day, the ER even reached 51%. For the thresholds of 32 °C, 33 °C, and 34 °C, from the fifth day onward, the ER also rose
sharply (16, 29, and 31%, respectively; p < 0.05). In addition, extreme high temperatures appeared to contribute to a
higher proportion of CVD deaths among elderly persons, females and outdoor workers. When the daily maximum
temperature was highe r than 33 °C from the ten th consecu tive day onward, the ER of CVD death among
these groups was 94, 104 and 149%, respectively (p < 0.05), which is considerably higher than the ER for the
overall population (87%; p <0.05).
Conclusions: The results of this study may assist governments in setting standards for heat waves, creating
more accurate heat alerts, and taking measures to prevent or red uce temperature-related de aths, especially
against the b ackd rop of global warmi ng.
Keywords: Heat wave, Cardiovascular diseas e, Con secut ive days’ high temperature
Background
Global climate change has, of course, caused the climate
to become warmer, but it has also increased the fre-
quency, intensity, duration, and spatial extent of certain
extreme weather events (such as heat waves and wind
chill) [1–3]. Heat wave is a prolo nged period of exces-
sively hot weather. The precise definition of a heat wave
varies between countries. The World Meteorological
Organization (WMO) suggests that the key characteris-
tics of a heat wave are the daily maximum temperature
that is higher than 32 °C and duration of more than 3
days. In China, according to the climatic feature, the
China Meteorological Administration states that the
characteristics of a heat wave are daily maximum tem-
peratures higher than or equal to 35 °C that last for
more than 3 days [4]. From 1961 to 2010, the range of
heat waves in China grew increasingly wide, they hap-
pened with ever greater frequency, and lasted exponen-
tially longer [2]. In some areas, there were up to four
heat waves in the summer, with durations of over 30
consecutive days [2].
The associations between heat waves and mortality have
been well studied [5–8]. For example, several recent stud-
ies have reported increased risk of C VD mortality on
* Correspondence: wangjf@lreis.ac.cn
State Key Laboratory of Resources and Environmental Information System
(LREIS), Institute of Geographic Sciences and Nature Resources Research,
Chinese Academy of Sciences, A11, Datun Road, Beijing 100101, China
© The Author(s). 2017 Open Access This article is distributed under the terms of the Creative Commons Attribution 4.0
International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and
reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to
the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver
(http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.
Yin and Wang BMC Public Health (2017) 17:223
DOI 10.1186/s12889-017-4129-7
heat-wave days as compared with non-heat-wave days;
some of these studies have also reported that individual
characteristics may help people to tolerate of prevent
them from tolerating such extreme weather events [9–12].
In addition, several studies have researched the influence
of heat waves based on different definitions (e.g. different
temperature thresholds and durations) on CVD mortality
[6, 13, 14]. Although several previous studies have investi-
gated the CVD mortality risk posed by heat waves with
different temperature thresholds and durations [15–17],
almost all of them estimated the average increased risk of
CVD mortality under different heat wave intensity and
duration, but failed to distinguish that for different heat
wave intensities and durations, the risks of CVD mortality
could vary widely. For example, by analyzing the CVD
mortality risk of heat waves on CVD in 43 U.S. cities
(1987–2005), Anderson et al. (2011) found that the heat
wave mortality risk increased by 2.49 and 0.38% for every
1 °F and 1 day increases in heat wave intensity and dur-
ation, respectively, in the United States. In the present
study, we estimated the CVD mortality effects under vari-
ous combinations of differing temperatures and durations
combinations among a range of ages, sex and occupation
groups, during the summers from 2010 to 2012 in Beijing.
The aim was to enable the provision of a more accurate
heat alert for the Chinese public.
Methods
Study area and meteorological data
Beijing, China’s capital, is a city of over 20 million residents
that is characterized by a typical temperate monsoon cli-
mate: summers are hot and rainy.
We obtained the daily maximum temperature, daily
mean relative humidity, and daily mean atmospheric
pressure from 18 authorized meteorological observation
stations in Beijing in the summers (1 June to 31 August)
from 2010 to 2012; these values were taken from the pub-
lic China Meteorological Data Sharing Service System
website [18]. We estimated the mean values for daily max-
imum temperature, daily mean relative humidity, and
daily mean atmospheric pressure across Beijing, using the
Inverse Distance Weighting interpolation technique in
ArcGIS 10.2 (Environmental Systems Research Institute,
Redlands, CA, USA). For the same period, the daily PM
2.5
concentrations at the US embassy station were obtained
from the website of the Embassy of t he United Sta tes
in China. US embassy station is near the c entre of
the study region.
Mortality data
Cardiovascular death cases among Beijing residents dur-
ing the summers (1 June to 31 Aug ust) from 2010 to
2012 were obtained from the Chinese Center for Disease
Control and Prevention (CDC). These ca ses were all
identified by hospitals, and the death certificates re-
corded the following information: gender, date of death,
age of death and occupation. According to the Inter-
national Classification of Diseases (Version 10), CVD death
cases were coded and classified into: I00-I99. All cases
were classified into two groups, according to age: <65 years
(younger) and ≥65 years (older). Based on the 75 occupa-
tions identified, the cases were divided into the following
four categories: indoor workers, outdoor workers, un-
employed, and unknown.
Statistical methods
A generalized additive model (GAM) was used to analyze
the excess mortality risk (ER) percentage of consecutive
days’ exposure to high temperatures (see Eq. (1)). In this
model, we also considered the effects of mean relative
humidity, mean air pressure, daily temperature range
(DRT) and air pollution [7, 8]:
LogE Y
t
½¼α þ CT M
t
þþns DT R
t
; dfðÞ
þns RH
t
; dfðÞþns P
t
; dfðÞ
þns PM
t
; dfðÞþns Time
t
; dfðÞþηdow
t
ð1Þ
where E(Y
t
) denotes the expected death number on day
t, α is the intercept. The function ns() is a natural spline
function. RH
t
refers to the daily mea n relative humidity
(RH) on day t.Thedf (degree of freedom) for RH wa s 3.
DTR
t
refers to the daily temperature range(DTR) on day
t, with df was 3. PM
t
refers to the daily PM
2.5
concentra-
tion on day t, with a df of 3. P
t
refers to the daily mean
air pressure on day t. The df for P
t
was 3. The smooth
term Time
t
was used to control for secular trends [7];
the df of time was 7 in each year. DOW
t
is the day of the
week on day t, which is a categorical variable. η is a vec-
tor of coefficients. All of the df values for variables in
this paper were chosen based on the Akaike Information
Criterion (AIC) [19]. In contrast to other studi es, we
created a categ orical variable, CTM, that refers to con-
secutive days of high temperature. A previous investiga-
tion had noted that the daily maximum temperature of
minimum mortality risk in Beijing was 30.5 °C [7]. In
this study, we chose the minimum mortality temperature
(MMT, 30.5 °C) as the reference temperature and esti-
mated the excess mortality risks of different temperature
thresholds (32 °C, 33 °C, 34 °C, and 35 °C) and durations
(1–11 days). For the various threshold values, we per-
formed these calculations independently. The MMT was
defined as the specific temperature associated to the
lowest mortality risk. For example, if we chose 32 °C as
the threshold, when the daily maximum temperatures
were lower than or equal to 32 °C on a given day, the
CTM value for this day was denoted as “REF” (refer-
ence). On the first day of exceeding the temperature
Yin and Wang BMC Public Health (2017) 17:223 Page 2 of 9
threshold (32 °C), the CTM value was denoted as “Hot1”
while on the se cond consecutive day, it wa s denoted
as “Hot2” and so on (see the examples shown in
Additional file 1:TableS1).
The ER was determined using the following formula.
RR refers to the relative risk of CVD death at different
temperature and durations, with the minimum mortality
temperature (30.5 °C) used as the reference temperature:
ER ¼ RR−1ðÞ100% ð2Þ
Because the estimation of relative risk in time series
models may change obviously depending on parameter
specifications [20–22], a sensitivity analysis was con-
ducted by adjusting the df from 2 to 7 for DTR , daily
mean air pressure, RH and PM
2.5
concentration, and
from 2 to 10 for each year for time. All of the df values
for variables in this paper were chosen based on the
Akaike Information Criterion (AIC ) [19]. For different
df value, the estimated values of relative risk varied
slightly, but the trends were similar. For all statistical tests,
P values <0.05 were considered statistically significant.
All statistic al analyses were conducted using R statis-
tical software (version 2.11.1; R Development Core Team
2010) and the mgcv package (Version 1.2.4).
Results
Statistical results
This study included 24,169 CVD death cases among
Beijing residents in the summers (1 June to 31 August)
from 2010 to 2012. Table 1 presents the daily mortality
and meteorological data in the summers for Beijing over
this period.
Figure 1 presents the daily maximum temperatures
and daily CVD deaths in the summers from 2010 to
2012 in Beijing. Table 2 shows the summary statistics for
consecutive days on which residents were exposed to
high temperatures; the temperature threshold was 32 °C.
“Ref” denotes the days of daily maximum temperatures
lower than or equal to 32 °C. Hot1, Hot2 … Hot11 stand
for the consecutive days of daily maximum temperatures
higher than 32 °C. The longest period in which there
was a temperature over 32 °C lasted for 11 days, with an
extreme maximum temperature of 40.6 °C.
ER values of consecutive days’ high temperature
Figure 2 presents the ERs of CVD death in terms of high
temperature over consecutive days (with thresholds of
32 °C, 33 °C, 34 °C, and 35 °C, respectively). The hori-
zontal axis stands for the number of consecutive days
with a maximum temperature higher than the threshold,
while the vertical axis represents the ERs caused by vari-
ous durations. Under these four conditions, the referen-
tial temperature of ER is 30.5 °C, which is the minimum
mortality temperature in Beijing.
From Fig. 2, it can be observed that over the first 3
days, the ER of CVD death under the four conditions
was generally consist ent, being lower than 10% for all
four. However, when daily maximum temperature was
higher than 35 °C from the fourth day onward, the ER
attributed to consecutive days’ extreme high temperature
exposure underwent an increase to about 10 and 51% on
the fourth and fifth days, respectively (p < 0.05). For the
thresholds of 32 °C, 33 °C, and 34 °C, over the first 4 days,
the ER of CVD death was generally low (under 10%), but
it rose sharply when high temperatures continued to
Table 1 Summary statistics for daily CVD deaths and meteorological variables in summer in Beijing, from 2010–2012
Category Minimum 25% Median 75% Maximum Mean ± SD
Daily meteorological data
Maximum temperature (°C) 23.2 29.5 31.1 33 40.6 31.2±2.9
Mean relative humidity (%) 21 56.8 66 75 97 64.7±14.6
Mean air pressure (hPa) 990.4 998.1 1001 1004 1014 1001±4.6
Deaths by age
≥65 years of age 39 62 71 79 144 71.9±12.9
0–64 years of age 6 13 15 18 36 15.7±4.3
Deaths by sex
Male 24 38 44 50 88 44.4±7.5
Female 20 36 42 49 92 42.5±8.1
Deaths by occupation
Outdoor 18 29 33 39 78 32.5±8
Indoor 17 33 38 44 67 37.6±7.1
Unemployed 2 7 9 12 19 9.5±4.3
Unknown 0 3 5 7 17 4.7±2.9
Yin and Wang BMC Public Health (2017) 17:223 Page 3 of 9
occur from the fifth day onward, increasing to 16, 29, and
31% as the days wore on (p < 0.05). On the seventh day,
the ER has a clear harvesting effect (mortality displace-
ment) [13, 17]. When the durations with daily maximum
temperatures higher than 32 °C, 33 °C, and 34 °C reached
9, 10, and 11 days, respectively, the ERs of CVD death
soared to 81, 87, and 93%, respectively (p <0.05).
ER values of the different subgrou ps
We estimated the percentages of ER increase, for the en-
tire population as well as for various ages, both sexes
and several occupation groups caused by high tempera-
tures. Figure 3 presents the results. The horizontal axis
represents the consecutive days of daily maximum
temperature that was higher than the threshold, while
the vertical a xis signifies the ER caused by different du-
rations of high temperature. Based on this analysis, we
found that the effects of high temperature on CVD mor-
tality among older people, female and outdoor work ers
were more serious than for the other groups. For ex-
ample, when the daily maximum temperature was higher
than 33 °C (see Fig. 3b) on the tenth consecutive day,
the ERs of CVD death among elderly people, females
and outdoor workers was 94, 104 and 149%, respectively,
which is considerably higher than the ER for the total
population (87%).
Tables 3 and 4 show the percentage increases and con-
fidence interval (95% CI) for CVD mortality in the vari-
ous groups listed in Figs. 2 and 3.
Based on the above results, we offer an innovative pro-
posal for a heat alert (see Table 5), which considers both
the high temperature threshold and duration.
Discussion
Although many studies have examined the effects of
heat wave on the excess mortality risk of CVD, almost
all of them estimated the average increased risk of CVD
mortality under different heat wave intensity and duration,
but failed to distinguish that for the different intensity and
durations, the risks of CVD death can vary widely. In the
present study, we estimated the CVD mortality effects
under various combinations of temperature and duration
combinations. A key finding was that when high tempera-
tures occur continuously, at varying temperature thresh-
olds and different durations, the adverse effects on CVD
mortality vary significantly. The longer the heat wave lasts,
the grea ter the mortality risk is. For example, when
the daily maximum temperature higher than 35 °C
occurred continuously, on the fourth and fifth days, the
ERs attributed to consecutive days’ high temperature ex-
posure underwent an obvious increase to about 10 and
51% (p < 0.05); and for the thresholds of 32 °C, 33 °C, and
34 °C, on the fifth day, the ERs also rose sharply to 16, 29,
and 31%, respectively (p < 0.05). Because very few studies
have specifically focused on the various combinations of
temperature threshold and duration of heat wave on CVD
mortality. So it is difficult to compare our results with
those of previous studies, However, a few studies have
reported that given different temperature thresholds and
durations, heat waves have adverse impacts on non-
accidental and coronary heart disease (CHD) mortality
[16, 17]. Sheridan et al. (2014) found that all non-acciden-
tal mortality ris k increased by 21.2% on the fifth day
of heat wave in New Y ork. By analyzing the CHD
mortality risk cau sed by heat waves in Beijing (2000–
2011), Tian et al. (2013) fou nd that CHD mortality
risk increased by about 60% on the se venth days of
heat wave. Another reason why we could not com-
pare our results with these of previous studies is that
the increased risk of C VD mortality they estimated
for heat-wave days is compared with non-heat-wave
days. Whereas, in the present study, we choose the
Table 2 Consecutive days of exposed to high temperature (≥32 °C) during the summers from 2010 to 2012 in Beijing
Ref Hot1 Hot2 Hot3 Hot4 Hot5 Hot6 Hot7 Hot8 Hot9 Hot10 Hot11
Days 157 43 24 19 11 6 4 3 3 2 2 1
“Ref” denotes the days of daily maximum temperatures lower than or equal to 32 °C
Fig. 1 The daily maximum temperature and daily CVD deaths in summer from 2010 to 2012 in Beijing
Yin and Wang BMC Public Health (2017) 17:223 Page 4 of 9
minimum-mortality temperature (MMT ) as the refer-
ence temperature for relative mortality risk. In my
opinion, using MMT a s the reference temperature is
more rea sonable.
Based on these results, we set forward a new proposal
for a heat alert, which considers both the high temperature
threshold and duration. When the duration of daily
maximum temperatures higher than 35 °C reaches to 4
days or the duration with daily maximum temperatures
higher than 32 °C, 33 °C or 34 °C reach 5 days, govern-
ments should both issue heat alerts, medical institutions
and emergency centers should prepare greater quantities
of medical resources, such as intravenous fluids, oxygen,
and extra beds for rapid treatment [23]. The CVD-
Fig. 3 Percentage increases of CVD mortality in the various groups (female (dashed line), older (dotted line) and outdoor workers (solid line)) due
to consecutive days’ high temperature, for varying threshold values: (a) 32 °C, (b) 33 °C, (c) 34 °C, and (d) 35 °C, with a reference temperature for
ER of 30.5 °C
Fig. 2 Percentage increase in CVD mortality due to consecutive days’ high temperatures for various threshold values, with an ER referential
temperature of 30.5 °C
Yin and Wang BMC Public Health (2017) 17:223 Page 5 of 9