Available online at www.sciencedirect.

com

ScienceDirect
Advances in Climate Change Research 14 (2023) 847e855
www.keaipublishing.com/en/journals/accr/

Temperature change between neighboring days and hospital admissions in

China
Yu-Shu HUANG a,b, He-Jia SONG a,c, Yi-Bin CHENG a, Shilu TONG a,d,e, Rui ZHANG f,
Xiao-Yuan YAO a, Yong-Hong LI a,*
a
China CDC Key Laboratory of Environment and Population Health, National Institute of Environmental Health, Chinese Center for Disease Control and
Prevention, Beijing 100021, China
b
Department of Big Data in Health Science School of Public Health, and Center of Clinical Big Data and Analytics of the Second Affiliated Hospital, Zhejiang
University School of Medicine, Hangzhou 310058, China
c
Jiaxing Center for Disease Control and Prevention, Jiaxing 314050, China
d
Institute of Environment and Population Health, School of Public Health, Anhui Medical University, Hefei 230032, China
e
School of Public Health and Social Work, Queensland University of Technology, Brisbane 4006, Australia
f
Information Center, Chinese Center for Disease Control and Prevention, Beijing 102206, China
Received 29 May 2023; revised 25 October 2023; accepted 28 November 2023
Available online 6 December 2023

Abstract

The short-term temperature fluctuation caused by global climate change is one of the risk factors affecting public health. Exploring the
association between temperature fluctuation and diseases, which has received relatively limited research attention, can contribute to enhancing
awareness of health risks and taking timely measures for health risk alert and management. Therefore, this study aims to investigate the
relationship between temperature change between neighboring days (TCN) and hospitalizations, identify diseases sensitive to extreme TCN, and
evaluate the related disease burden. We collected meteorological and hospitalization data from 2014 to 2019 in 23 sites of China to explore the
impact of TCN on hospitalizations. We first quantified site-specific associations between TCN and hospitalizations and then conducted meta-
analysis to pool the results, to assess the relative risk of extreme TCN for susceptible diseases, and to estimate the related disease burden
attributed to TCN. Stratified analyses were undertaken by age, sex, and disease type. Results showed that all-cause hospital admission was
significantly linked to TCN. A negative TCN (below
1.9  C) in the cool season and a positive TCN (above 1.0  C) in the warm season
increased the risk of hospitalization. People aged 15e64 years, men, and patients with musculoskeletal system or connective tissue diseases were
more sensitive to extremely negative TCN during the cool season. People aged over 65 years, men, and patients with respiratory diseases were
more sensitive to extremely positive TCN during the warm season. The attributable fraction to all-cause hospitalization from negative TCN in
the cool season was 2.05% (95% CI:
0.90%, 4.53%) and from positive TCN in the warm season was 5.79% (95% CI: 2.98%, 8.31%).
Circulatory diseases in the cool season and respiratory diseases in the warm season had the highest disease burden. Our findings indicate that
awareness of TCN and its health risks should be promoted and evidence-informed policies are needed to reduce the risk of TCN.

Keywords: Temperature change between neighboring days; Morbidity; Hospital admissions; Sensitive diseases; Vulnerable populations; Disease burden

1. Introduction

Climate change is causing more frequent and extreme

* Corresponding author. weather events that can result in more severe health problems.
E-mail address: [email protected] (LI Y.-H.). While several studies have shown that daily non-optimum
Peer review under responsibility of National Climate Center (China temperature is closely associated with morbidity and
Meteorological Administration).

https://doi.org/10.1016/j.accre.2023.11.013
1674-9278/© 2023 The Authors. Publishing services by Elsevier B.V. on behalf of KeAi Communications Co. Ltd. This is an open access article under the CC
BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
848 HUANG Y.-S. et al. / Advances in Climate Change Research 14 (2023) 847e855

mortality (Chen et al., 2018; Fu et al., 2018), a growing body foundation for understanding harmful TCN thresholds and the
of evidence suggests that short-term temperature variation, as specific diseases mostly impacted by extreme temperature
another indicator of unstable weather, is also a risk factor of changes.
disease burden (Guo et al., 2021; Lei et al., 2021; Yang et al.,
2021). 2. Materials and methods
The most common indices of short-term temperature
change include diurnal temperature change (Tang et al., 2018), 2.1. Data collection
which reflects intra-day temperature change, temperature
change between neighboring days (TCN) (Zhan et al., 2017), Hospital admission data from 2014 to 2019 were collected
which captures the difference between the mean temperatures from local comprehensive hospitals at 23 sites covering 11
of two neighboring days, and temperature variability (Wu geoclimatic divisions (characterized by geographical, meteo-
et al., 2022), calculated as the standard deviation of the rological, and administrative factors) across China (Fig. 1).
daily minimum and maximum temperatures over a specific These geoclimatic regions were selected based on the National
time period. Although studies have investigated the relation- Project on Scientific Investigation on Regional Climate-
ship between increased disease burden and a wider diurnal sensitive Diseases in China, supported by the National Min-
temperature change or increased temperature variability at istry of Science and Technology. Two or three study sites in
regional, national, and global levels (Byun et al., 2020; Cheng each geoclimatic area were randomly selected to obtain a
et al., 2019; Wu et al., 2022), the current evidence about the representative sample. One to two comprehensive hospitals
health effects of TCN is inconsistent (Ma et al., 2020; Zha were selected at each site. To obtain an adequate sample size,
et al., 2022; Zhan et al., 2017). A nationwide study in the the selected hospitals had to treat at least 200 patients per day
United States suggested that positive TCN (a temperature rise in both the outpatient and emergency departments.
between 2 adjacent days) increased the risk of death from The hospitalization data included patient admission date,
respiratory and cardiovascular diseases in each season (Zhan sex, age, and diagnosis. We performed stratified analyses by
et al., 2017). However, a study in Qingyang city located in sex (men, women), age (14, 15e64, 65 years), and disease
Northwest China found the opposite (Zha et al., 2022). type. Disease was categorized in accordance with the Inter-
Therefore, there is a need to conduct additional multi-area national Classification of Diseases, 10th Revision into the
studies to better understand this issue. following disease types: endocrine and metabolic diseases
Furthermore, while most previous studies have focused on (E00eE90), mental and behavioral disorders (F00eF99),
circulatory or respiratory diseases (Ma et al., 2020; Zha et al., diseases of the nervous system (G00eG99), diseases of the
2022; Zhan et al., 2017), only a few have investigated the eye and adnexa (H00eH59), diseases of the ear and mastoid
effects of TCN on other diseases (e.g., hand, foot, and mouth process (H60eH95), diseases of the circulatory system
disease and tuberculosis) (Cheng et al., 2016; Huang et al., (I00eI99), diseases of the respiratory system (J00eJ99), dis-
2020). Many epidemiological and experimental studies have eases of the digestive system (K00eK93), diseases of the skin
found that genitourinary and endocrine diseases are sensitive and subcutaneous tissue (L00eL99), diseases of the muscu-
to extreme temperatures (Borg et al., 2017; Konkel, 2020), loskeletal system and connective tissue (M00eM99), diseases
implying that they might also be susceptible to TCN. There- of the genitourinary system (N00eN99), and pregnancy and
fore, it is necessary to consider additional disease types to the puerperium (O00eO75, O85e99). Owing to insufficient
obtain a more comprehensive assessment of TCN-sensitive
diseases.
Although some previous studies have calculated the
attributable fraction (AF) and attributable number (Zha et al.,
2022) or have estimated the relative risk (RR) of extreme TCN
(Lei et al., 2021; Zhan et al., 2017), few have identified
dangerous TCN thresholds or ranges (Guo et al., 2011). Un-
derstanding the harmful thresholds of TCN and the extent of
disease burden caused by TCN is essential to inform the
public, especially vulnerable groups, about protective mea-
sures that they can take.
In this study, we explored the associations between TCN
and hospitalizations, determined harmful TCN ranges, and
identified diseases sensitive to extreme TCN. We also calcu-
lated the TCN-related AF to assess the attributable burden of
diseases in different groups (e.g., by age, sex, and disease
type) during the cool season (from November to March) and
warm season (from May to September) across 23 sites of
China. The findings can provide valuable insights into the
relationship between TCN and hospitalizations, offering a Fig. 1. Distribution of 23 sites in China.
 HUANG Y.-S. et al. / Advances in Climate Change Research 14 (2023) 847e855 849

data, we did not perform analyses for diseases of the eye and current day's mean temperature and RH were controlled by
adnexa and diseases of the ear and mastoid process. using an s with 7 df and an s with 3 df, respectively. The effects
The meteorological data, collected from the China Meteo- of long-term trends and the day of the week (dw) were also
rological Administration, included a time-series dataset controlled using an s with 6 df per year and a categorical
comprising the daily mean temperature (Tmean;  C) and rela- variable, respectively. The model was as follows:
tive humidity (HR; %), enabling to use accurate and localized
data for the analysis. TCN (DT ) was calculated by subtracting LogðEðYÞÞ ¼ aþ cbðDTÞþ sðTmean ;7Þ þ sðHR ; 3Þþ sðT;6Þ þ dw
the previous day's mean temperature from the current day's ð1Þ
mean temperature. After accounting for regional heterogene-
ity, the 50th percentile of TCN was set as the reference to where Y is the daily number of hospital admissions, a is the
obtain the RR of TCN variation in the cool and warm seasons intercept, cbðDTÞ is the cross-basis function of TCN to esti-
(Zhan et al., 2017). The 1st and 99th percentile of TCN were mate the exposure and lag effect in two dimensions; sðTmean Þ is
defined as extremely negative TCN and extremely positive the natural cubic spline for the mean temperature, sðHR Þ is the
TCN, respectively (Zhan et al., 2017). natural cubic spline for relative humidity, sðTÞ is the natural
cubic spline for time.
2.2. Statistical analysis In the second stage, a multivariable meta-regression with a
random-effect model was used to pool the site-specific esti-
A two-stage approach was applied to quantify the cumu- mated values obtained in the first stage. The RR of extreme
lative exposureeresponse relationship and the disease burden. TCN and the percentile ranges of TCN for different diseases at
Briefly, we first obtained site-specific associations between the national level were calculated according to the overall
TCN and hospital admissions. We then performed a multi- exposureeresponse relationship. The percentiles were then
variable meta-analysis to pool the regional effects. For this, we converted into actual temperatures to provide the public with a
used two indicators to evaluate the vulnerability of populations more straightforward understanding. Finally, the AFs were
to TCN: the RR of extreme TCN, which was used to identify calculated to estimate the disease burden in different
TCN-sensitive diseases, and the disease burden associated subgroups.
with harmful TCN.
X
L 
In the first stage, we extracted the data to explore the effect AFx;t ¼ 1
exp
bxt ;l ð2Þ
of TCN in the cool season (November to March) and warm l¼0
season (May to September) separately. A distributed lag
nonlinear model (DLNM) with conditional quasi-Poisson where AFx;t is the Pattributable fraction on day t in site x; exp is
L
regression was employed to assess the site-specific associa- power function; l¼0 bxt ;l is the logarithm of the relative
tions between TCN and hospitalizations. In light of variations hazard ratio for l days of cumulative temperature on day t in
across subgroups, we assessed the acting lag days for each site x; l is the maximum lag time of the exposure factor, and b
group by analyzing their respective lageresponse relationships is the effect parameter of the exposure factor.
(Ma et al., 2020; Zha et al., 2022) (Table 1).
We used a natural cubic spline (s) with 3 degrees of 2.3. Sensitivity analysis
freedom (df) to fit the exposureeresponse relationship and the
lageresponse relationship. The confounding effects of the Several sensitivity analyses were conducted to examine the
robustness of our findings. We varied the df of time from 5 to 7
to check whether long-term trends affected the results. We also
Table 1 adjusted the df for confounding factors (Tmean from 6 to 8; HR
Acting lag days (d) in different subgroups during cool and warm season. from 4 to 5). We changed the definition of extreme TCN (5th
Subgroup Cool season Warm season and 95th percentiles) to compare the effect patterns. And we
Total 5e14 0e10 also replaced the daily Tmean and relative humidity with 7-
Age 14 years 0e9 5e11 d moving average temperature and 3-d moving average hu-
15e64 years 2e15 0e9 midity, respectively, to verify the stability of the results.
65 years 5e13 0e8
R software (version 4.0.3, Free Software Foundation) was
Sex Male 5e15 0e9
Female 4e13 0e9 used to perform all data analyses. The ‘gnm’, ‘dlnm’, and
Disease type Endocrine and metabolic 3e12 0e11 ‘mvmeta’ packages were used for the conditional quasi-
Mental and behavioral 0e14 0e14 Poisson regression, DLNM and the meta-regression model,
Nervous 0e14 0e6 respectively.
Circulatory 2e18 0e5
Respiratory 0e10 0e13
Digestive 3e12 3e12 2.4. Role of funding source
Skin 0e14 0e14
Musculoskeletal 0e13 3e6 The funding source for this study had no role in study
Genitourinary 0e11 0e10 design, data collection, statistical analysis, result interpreta-
Pregnancy and puerperium 0e14 0e14
tion, or report writing.
850 HUANG Y.-S. et al. / Advances in Climate Change Research 14 (2023) 847e855

3. Results 70th percentile (1.0  C) in the warm season. Table 3 reveals

the thresholds of TCN for each subgroup.
3.1. Descriptive statistics
3.3. Extreme TCN-sensitive diseases
In 23 sites of China from 2014 to 2019, during the cool
season, the average daily maximum, mean, and minimum Fig. 3 shows the RR of extreme TCN for hospital admis-
temperatures were 9.7, 4.3, and 0.1  C, respectively. The TCN sions. Generally, consistent with the findings on the
ranged from
14.6 to 17.9  C. The median, 1st, and 99th exposureeresponse relationship, extremely negative TCN in
percentiles of TCN were 0.2,
14.3, and 16.1  C, respectively. the cool season and extremely positive TCN in the warm
During the warm season, the average daily maximum, mean, season increased the risk of hospitalizations, with RRs of 1.19
and minimum temperatures were 28.4, 23.2, and 19.0  C, (95% CI: 1.08, 1.31) and 1.15 (95% CI: 1.07, 1.24),
respectively. The TCN ranged from
14.5 to 14.6  C. The respectively.
median, 1st, and 99th percentiles of TCN were 0.2,
14.5, and During the cool season, the RRs of extremely negative
13.7  C, respectively. TCN for people aged 14, 15e64, and 65 years were 0.64
Table 2 shows the total and daily hospitalizations by age, (95% CI: 0.53, 0.77), 1.39 (95% CI: 1.14, 1.69), and 1.16
sex, and disease type. This study included 2,350,223 hospi- (95% CI: 1.05, 1.28), respectively, suggesting that people aged
talized patients during the cool season and 2,415,112 hospi- 15e64 were most sensitive to TCN. Men were found to have a
talized patients during the warm season in 23 sites. People higher risk than women, with a RR of 1.26 (95% CI: 1.12,
aged 15e64 years accounted for the highest proportion of all 1.42) and 1.14 (95% CI: 1.03, 1.25), respectively. Addition-
hospitalizations, followed by those aged 65 years and 14 ally, as shown in Fig. 3, extremely negative TCN in the cool
years. The proportion of women was slightly larger than that season significantly increased the risk of endocrine and
of men. The three leading disease subtypes with the most metabolic diseases, diseases of the musculoskeletal system
hospital admissions were circulatory, respiratory, and digestive and connective tissue, and circulatory, digestive, and genito-
diseases in both the cool and warm seasons. urinary system diseases. However, extremely negative TCN in
the cool season decreased the risk of respiratory diseases, with
3.2. Association between TCN and hospitalizations a RR of 0.62 (95% CI: 0.49, 0.79). The highest RR was
observed for diseases of the musculoskeletal system and
Fig. 2 shows the exposureeresponse relationship between connective tissue (1.91; 95% CI: 1.35, 2.70).
TCN and all-cause hospitalizations. In general, a negative During the warm season, all subgroups of age and sex had
TCN in the cool season and a positive TCN in the warm an increased risk of hospitalization due to extremely positive
season increased the risk of hospitalizations. Statistically TCN. People aged  65 years and men were more sensitive,
significant risk was identified when the TCN was below the with a RR of 1.21 (95% CI: 1.12, 1.31) and 1.19 (95% CI:
20th percentile (
1.9  C) in the cool season and above the 1.11, 1.27), respectively. Genitourinary disease was associated

Table 2
Descriptive characteristics of hospitalizations during the cold and warm seasons, 2014e2019 (by age, gender, and disease type).
Subgroup Cool season Warm season
Total Mean SD Min Median Max Total Mean SD Min Median Max
Total 2,350,223 125.4 125.2 0 91 1455 2,415,112 127.3 124.7 0 90 1065
Age 14 years 134,973 7.2 13.8 0 5 409 114,363 6 5.6 0 5 48
15e64 years 1,439,616 76.8 76.7 0 55 832 1,513,103 79.8 77.7 0 56 688
65 years 778,121 41.5 49.2 0 28 601 786,992 41.5 47.8 0 27 389
Sex Male 1,121,208 59.8 60.6 0 43 718 1,154,525 60.9 59.8 0 43 461
Female 1,229,015 65.6 65.9 0 48 737 1,260,587 66.4 66 0 48 638
Disease type Endocrine and metabolic 110,148 5.9 9 0 3 117 115,386 6.1 9.2 0 3 97
Mental and behavioral 9165 0.5 1 0 0 13 9595 0.5 1 0 0 13
Nervous 72,979 3.9 5.2 0 2 55 78,488 4.1 5.3 0 2 61
Eye 49,375 2.6 5.6 0 0 64 49,821 2.6 5.5 0 1 64
Ear 18,891 1.0 1.7 0 0 18 20,066 1.1 1.7 0 0 18
Circulatory 413,856 22.1 22.2 0 16 230 404,683 21.3 21.1 0 15 163
Respiratory 342,501 18.3 13.8 0 15 145 280,008 14.8 11.7 0 12 84
Digestive 251,709 13.4 15.6 0 9 235 263,214 13.9 15.6 0 10 127
Skin 19,080 1 1.8 0 0 32 22,859 1.2 1.9 0 0 18
Musculoskeletal 94,257 5 7.4 0 2 84 103,920 5.5 7.7 0 3 77
Genitourinary 152,074 8.1 10.7 0 5 123 177,034 9.3 11.7 0 5 100
Pregnancy and puerperium 140,575 7.5 7.9 0 5 66 142,979 7.5 7.7 0 5 48
Note: SD, standard deviation; Min, minimum value; Max, maximum value.
 HUANG Y.-S. et al. / Advances in Climate Change Research 14 (2023) 847e855 851

Fig. 2. Exposureeresponse relationship between TCN and all-cause hospitalizations during the cool and warm seasons.

with a RR of 0.87 (95% CI: 0.76, 0.99). There was a signifi- Altering the df of Tmean (6 or 8 df) and HR (4 or 5 df) also had
cantly increased risk of endocrine and metabolic diseases, little effect on the relationship between TCN and hospital
nervous system, circulatory system, and respiratory diseases, admissions. Changing the definition of extreme TCN (5th and
and musculoskeletal system and connective tissue diseases as 95th percentiles) resulted in slight decreases in the RR of
a result of extremely positive TCN. Respiratory diseases had extremely negative TCN in the cool season and extremely
the highest RR of 1.64 (95% CI: 1.29, 2.07). positive TCN in the warm season, but the results still followed
the same pattern as the main model. Using average windows of
3.4. Attributable disease burden different durations for the daily mean temperature and relative
humidity had no significant impacts on the results.
Table 3 shows the harmful TCN ranges and related AFs.
The AFs owing to negative TCN (below
1.9  C) in the cool 4. Discussion
season and positive TCN (above 1.0  C) in the warm season
were 2.05% (95% CI:
0.90%, 4.53%) and 5.79% (95% CI: We assessed the association between TCN and hospitali-
2.98%, 8.31%), respectively. zations and to evaluate the attributable burden of TCN at the
Regarding the effects of negative TCN during the cool national level in China. The results showed that a negative
season, people aged 15e64 years had the highest AF (3.66%, TCN below
1.9  C during the cool season and a positive
95% CI:
0.88%, 6.94%) among the age groups. Men had a TCN above 1.0  C during the warm season increased the risk
higher attributable burden than women, with AFs of 3.81% of all-cause hospitalizations, resulting in a considerable dis-
(95% CI: 0.52%, 6.52%) and 0.52% (95% CI:
1.89%, ease burden, especially for circulatory and respiratory dis-
2.84%), respectively. Among disease types, the AF of circu- eases. Additionally, patients with diseases of the
latory diseases was the highest at 9.18% (95% CI: 2.03%, musculoskeletal system and connective tissue were most
13.90%), followed by genitourinary diseases and endocrine sensitive to extremely negative TCN during the cool season,
and metabolic diseases, with AFs of 7.46% (95% CI: 1.48%, and patients with respiratory diseases were most sensitive to
10.88%) and 3.22% (95% CI: 0.25%, 5.43%), respectively. extremely positive TCN during the warm season.
Regarding the effects of positive TCN during the warm We found a significant relationship between TCN and all-
season, people  65 years had the highest AF at 5.96% (95% cause hospitalizations, indicating that TCN is an important
CI: 2.18%, 9.26%). Men had a higher AF (4.74%, 95% CI: risk factor of morbidity. Consistently, studies conducted in two
1.60%, 7.20%) than women (3.58%, 95% CI: 1.06%, 5.71%). subtropical cities in China reported that a temperature rise
Among the disease types, respiratory disease had the highest between 2 adjacent days in summer was significantly associ-
AF at 18.08% (95% CI: 11.69%, 22.73%), followed by ated with increased mortality (Lin et al., 2013). The reason for
endocrine and metabolic diseases and circulatory diseases, this association may be that although the human body adapts
with AFs of 6.06% (95% CI: 0.60%, 10.15%) and 3.40% (95% to ambient temperature changes through physiological regu-
CI: 0.97%, 5.44%), respectively. lation and changes in behavior, when the temperature changes
dramatically, the thermoregulatory system is not able to
3.5. Sensitivity analyses respond quickly, especially in young children and older adults
with underlying diseases (Lei et al., 2021; Wang et al., 2021;
Table 4 shows the results of the sensitivity analyses. Zhan et al., 2017). An imbalance in heat exchange between the
Compared with the main model, adjusting the df of time (5 or body and the external environment, immune dysfunction,
7 df per year) had little impact on the effect estimation. pathophysiological changes, and disturbances in hormone
852 HUANG Y.-S. et al. / Advances in Climate Change Research 14 (2023) 847e855

Table 3
Harmful TCN ranges and related attributable fractions for different subgroups.
Subgroup Cool season Warm season

Range ( C) AF (95% CI) (%) Range ( C) AF (95% CI) (%)
Total (e10.6,e1.9) 2.05 (
0.90,4.53) (1.0,6,5) 5.79 (2.98,8.31)*
Age 14 years (e10.6,e1.4)
6.24 (
17.87,1.86) (1.8,6.5) 3.76 (0.18,6.60)*
15e64 years (e10.6,e2.0) 3.66 (
0.88,6.94) (0.6,6.5) 3.73 (0.92,6.30)*
65 years (e10.6,e1.9) 1.94 (
1.47,4.49) (1.0,6.5) 5.96 (2.18,9.26)*
Sex Male (e10.6,e1.8) 3.81 (0.52,6.52)* (1.2,6.5) 4.74 (1.60,7.20)*
Female (e10.6,e2.5) 0.52 (
1.89,2.84) (1.1,6.5) 3.58 (1.06,5.71)*
Disease type Endocrine and metabolic (e10.6,e2.5) 3.22 (0.25,5.43)* (1.5,6.5) 6.06 (0.60,10.15)*
Mental and behavioral / / /
Nervous / / (1.1,6.5) 2.14 (
4.19,7.06)
Circulatory (e10.6,e1.3) 9.18 (2.03,13.90)* (1.5,6.5) 3.40 (0.97,5.44)*
Respiratory (e10.6,e0.5) 3.57 (
4.80,9.74) (0.2,6.5) 18.08 (11.69,22.73)*
Digestive (e10.6,e2.9) 0.91 (
3.07,3.57) /
Skin / / (0.2,1.8)
4.63 (
10.37,0.09)
Musculoskeletal (e10.6,e2.1) 6.24 (
2.28,10.69) (1.8,6.5) 0.92 (
0.80,2.49)
Genitourinary (e10.6,e2.1) 7.46 (1.48,10.88)* (1.6,6.5)
0.91 (
6.62,3.64)
Pregnancy and puerperium / / / /
Note: *p < 0.05.

secretion, which can elevate blood pressure and cholesterol to TCN is primarily associated with compromised thermo-
levels, are all risk factors for insufficient adaption to TCN regulatory capacity, resulting in suboptimal temperature con-
(Cao et al., 2021; Wang et al., 2021). trol. Addition, co-morbidities like cardiovascular diseases and
Our findings show that hospitalizations of almost all studied metabolic disorders will further heighten the vulnerability to
populations were associated with varying degrees of TCN, TCN (Zhan et al., 2017). However, children aged 14 years
especially extreme TCN. People aged 15e64 years and 65 were not apparently harmed by temperature decreases in the
years were sensitive to temperature decreases during the cool cool season, which may be because children are usually well
season and temperature increases during the warm season. For cared for by their parents. In China, children are normally
people aged 15e64 years, they may spend more time engaged asked to wear more clothing to prevent illness and are often
in outdoor physical activities or labor, making them more not permitted to go outside on cold days, which may reduce or
susceptible to the negative impact of TCN (Wang et al., 2022). prevent the impact of TCN. Regarding sex, men were at higher
And especially for individuals aged 65 years, susceptibility risk of TCN, which resulted in a greater health burden than

Fig. 3. Relative risks of extremely TCN on hospitalizations for different subgroups. (a) Negative (1st percentile) TCN on hospitalizations during the cold season,
(b) positive (99th percentile) TCN on hospitalizations during the warm season.
 HUANG Y.-S. et al. / Advances in Climate Change Research 14 (2023) 847e855 853

Table 4 concluded that temperature decreases increased the risk of

Results of sensitivity analysis. asthma exacerbation (Zhu et al., 2022), which is the opposite
Season RR (95%CI) of RR (95%CI) of of what our study found. The differences in the results might
extreme negative TCN extreme positive TCN be because of the disease types that were investigated. Hence,
Cool season additional specific diseases need to be studied separately to
Main model 1.187 (1.076,1.309) 0.955 (0.874,1.044) generate accurate conclusions and to determine the direction
df per year
5 1.154 (1.052,1.266) 0.964 (0.867,1.071)
and extent of TCN impact.
7 1.150 (1.047,1.263) 0.977 (0.887,1.075) The sensitivity of endocrine and metabolic diseases to TCN
df/Tmean should also be noted. We found that the risk of hospitalizations
6 1.189 (1.077,1.312) 0.955 (0.874,1.044) owing to endocrine and metabolic diseases was significantly
8 1.182 (1.073,1.303) 0.954 (0.873,1.043) increased by both negative TCN in the cool season and posi-
df/HR
4 1.186 (1.076,1.308) 0.956 (0.873,1.047)
tive TCN in the warm season. There was no previous study
5 1.184 (1.075,1.303) 0.951 (0.870,1.040) about the relationship between TCN and endocrine and
Different TCN definition metabolic diseases. However, some related information has
5th/95th percentiles 1.150 (1.059,1.248) 0.971 (0.899,1.047) been provided (Li et al., 2014; Yang et al., 2016) that suggests
Moving average ambient temperature can affect glucose tolerance by influ-
7-d temperature 1.242 (1.096,1.407) 0.930 (0.827,1.046)
3-d humidity 1.190 (1.079,1.312) 0.954 (0.873,1.042)
encing the distribution of blood flow (from skin to viscera)
Warm season (Schmidt et al., 1994; Sellers et al., 2021). Low ambient
Main model 0.994 (0.907,1.088) 1.153 (1.074,1.238) temperatures can also lead to a higher risk of blood sugar
df per year disorders and insulin resistance (Valdes et al., 2019). A com-
5 0.978 (0.884,1.082) 1.165 (1.084,1.252) bination of these factors likely resulted in the high incidence
7 1.012 (0.933,1.099) 1.154 (1.068,1.246)
df/Tmean
of endocrine and metabolic diseases associated with TCN.
6 1.015 (0.936,1.102) 1.154 (1.069,1.246) Regarding other types of diseases, TCN may regulate mood
8 1.014 (0.934,1.100) 1.155 (1.069,1.247) through the serotonin system (5-serotonin), which could exert
df/HR antidepressant effects by activating serotonergic neurons in the
4 0.993 (0.907,1.087) 1.153 (1.074,1.237) brain (Zhang et al., 2020). Suboptimal temperature could also
5 0.993 (0.907,1.087) 1.152 (1.073,1.237)
Different TCN definition
cause morphological alterations in human nerve and glial
5th/5th percentiles 0.995 (0.920,1.076) 1.132 (1.063,1.204) cells, and trigger apoptotic pathways and DNA damage (Lin
Moving average et al., 2023). These multiple pathways could contribute to a
7-d temperature 0.996 (0.919,1.079) 1.154 (1.086,1.225) higher prevalence of mental and nervous diseases. Besides, for
3-d humidity 0.984 (0.903,1.073) 1.150 (1.075,1.229) digestive diseases, some potential factors include alterations in
food safety and hygiene, changes in dietary habits, and the
that of women (Cheng et al., 2014; Lin et al., 2013). The influence of temperature on the proliferation of certain path-
greater vulnerability of men to TCN can be explained by a ogens that can affect the gastrointestinal tract (Zhao et al.,
combination of occupational and physiological factors. Men's 2022). Regarding skin diseases, the combined effect of heat
increased involvement in outdoor work exposes them to and humidity heightens the susceptibility to skin diseases, and
temperature variations, and their lower subcutaneous fat and temperature fluctuations can exacerbate bad skin conditions
specific metabolic profiles diminish their adaptability to such as eczema and psoriasis (Hui-Beckman et al., 2023).
extreme TCN (Ma et al., 2021; Tang et al., 2018). When it comes to diseases of the musculoskeletal system and
We found that TCN impacted various types of diseases. connective tissue, temperature variation will affect blood cir-
Overall, circulatory system and respiratory diseases were the culation and collagen in connective tissues, worsen existing
most sensitive to TCN and had the highest attributable burden. conditions and cause inflammation. Additionally, high tem-
Our results were in line with a previous study regarding cir- peratures can lead to dehydration, concentrating urine and
culatory system diseases (Ma et al., 2021). For example, a potentially that may cause kidney stones and affect renal
study that quantified the effects of TCN on cardiovascular function. Hot and humid conditions can create an environment
diseases in cool periods and warm periods in Beijing found a conducive to the development of urinary tract infections (Qu
significant association between TCN and cardiovascular et al., 2023). Furthermore, high temperatures can also lead
morbidity. Experimental studies have indicated that unstable to complications like preterm birth and low blood pressure,
temperatures influence vasoconstriction, blood pressure, heart and elevated body temperature due to extreme heat can
rate, and platelet viscosity (Carder et al., 2005; Garrett et al., potentially harm the fetus and increase the risk of birth defects
2011; Martinez-Nicolas et al., 2015). Additionally, some (Chersich et al., 2020; Zhang et al., 2017).
studies have found that temperature increases in summer in- This study has several major strengths. First, we used a
crease the prevalence of respiratory diseases, pneumonia, and nationwide study sample, covering multiple areas across
chronic obstructive pulmonary disease (Lin et al., 2013) owing different geoclimatic regions, which enhances the general-
to impacts on immune function (Carder et al., 2005). However, izability of our findings. Second, our analysis took into ac-
there is no consensus about the impact of TCN on respiratory count multiple disease types when assessing
diseases in the cool season. A study in 18 Chinese cities TCNehospitalization associations. Thus, we were able to
854 HUANG Y.-S. et al. / Advances in Climate Change Research 14 (2023) 847e855

obtain a more comprehensive picture of TCN-sensitive dis- Acknowledgments

eases. Third, we further evaluated the disease burden attrib-
utable to TCN, which not only adds to the existing literature This work was financially supported by the Science and
but aids in the identification of specific populations that may Technology Fundamental Resources Investigation Program of
benefit from targeted interventions to mitigate the health China (2017FY101201).
impacts of temperature change. Finally, we obtained the TCN
thresholds, below or above which the risk of hospitalization References
would significantly increase, providing an important scien-
tific basis for the health risk early warning system of severe Borg, M., Bi, P., Nitschke, M., et al., 2017. The impact of daily temperature on
weather. renal disease incidence: an ecological study. Environ. Health 16 (1), 114.
https://doi.org/10.1186/s12940-017-0331-4.
This study also had several limitations. First, we did not Byun, S., Myung, W., Kim, H., et al., 2020. Association between diurnal
incorporate air pollutants into the models because of the lack temperature range and emergency department visits for multiple sclerosis:
of data at some sites, this might have resulted in overestimates a time-stratified case-crossover study. Sci. Total Environ. 720, 137565.
of the effect of TCN, to some extent. Second, due to limited https://doi.org/10.1016/j.scitotenv.2020.137565.
sample size for specific diseases, we had to broadly categorize Cao, R., Wang, Y., Huang, J., et al., 2021. The mortality effect of apparent
temperature: a multi-city study in Asia. Int. J. Environ. Res. Publ. Health
disease types, hindering a precise analysis of TCN's effects. 18 (9), 4675. https://doi.org/10.3390/ijerph18094675.
Therefore, for a more detailed evaluation of TCN effects, it is Carder, M., McNamee, R., Beverland, I., et al., 2005. The lagged effect of cold
imperative to subdivide and analyze the disease types. Third, temperature and wind chill on cardiorespiratory mortality in Scotland.
variations in TCN's health impact across different geograph- Occup. Environ. Med. 62 (10), 702e710. https://doi.org/10.1136/
ical regions were influenced by diverse climate patterns, ur- oem.2004.016394.
Chen, R., Yin, P., Wang, L., et al., 2018. Association between ambient
banization levels, adaption disparities, and environmental temperature and mortality risk and burden: time series study in 272 main
factors. However, due to data limitations, regional differences Chinese cities. Br. Med. J. 363, k4306. https://doi.org/10.1136/
remained unexamined. More future researches are needed to bmj.k4306.
explore regional distinctions for a comprehensive under- Cheng, J., Zhu, R., Xu, Z., et al., 2014. Temperature variation between
standing of TCN effects. neighboring days and mortality: a distributed lag non-linear analysis. Int. J.
Publ. Health 59 (6), 923e931. https://doi.org/10.1007/s00038-014-0611-5.
Cheng, J., Zhu, R., Xu, Z., et al., 2016. Impact of temperature variation be-
5. Conclusion tween adjacent days on childhood hand, foot and mouth disease during
April and July in urban and rural Hefei, China. Int. J. Biometeorol. 60 (6),
This study obtained the TCN thresholds both in warm and 883e890. https://doi.org/10.1007/s00484-015-1082-y.
cold seasons, provided information about diseases and pop- Cheng, J., Xu, Z., Bambrick, H., et al., 2019. Impacts of heat, cold, and tem-
perature variability on mortality in Australia, 2000e2009. Sci. Total Envi-
ulations that were sensitive to temperature changes between ron. 651 (Pt 2), 2558e2565. https://doi.org/10.1016/j.scitotenv.2018.10.186.
neighboring days, both in the cool and warm seasons, and Chersich, M.F., Pham, M.D., Areal, A., et al., 2020. Associations between high
evaluated the disease burden attributable to TCN. People over temperatures in pregnancy and risk of preterm birth, low birth weight, and
the age of 15 years and men were most vulnerable to TCN. stillbirths: systematic review and meta-analysis. Br. Med. J. 371, m3811.
TCN increased the risk of hospitalizations from a variety of https://doi.org/10.1136/bmj.m3811.
Fu, S.H., Gasparrini, A., Rodriguez, P.S., et al., 2018. Mortality attributable to
diseases, especially respiratory, circulatory, endocrine and hot and cold ambient temperatures in India: a nationally representative
metabolic, and genitourinary diseases. Thus, awareness about case-crossover study. PLoS Med. 15 (7), e1002619. https://doi.org/
the risk of TCN should be increased to prevent its adverse 10.1371/journal.pmed.1002619.
health outcomes. This matter becomes increasingly important Garrett, A.T., Rehrer, N.J., Patterson, M.J., 2011. Induction and decay of short-
as climate change proceeds. term heat acclimation in moderately and highly trained athletes. Sports
Med. 41 (9), 757e771. https://doi.org/10.2165/11587320-000000000-
00000.
Declaration of competing interest Guo, F., Do, V., Cooper, R., et al., 2021. Trends of temperature variability:
which variability and what health implications? Sci. Total Environ. 768,
The authors declare no conflict of interest. 144487. https://doi.org/10.1016/j.scitotenv.2020.144487.
Guo, Y., Barnett, A.G., Yu, W., et al., 2011. A large change in temperature
between neighbouring days increases the risk of mortality. PLoS One 6 (2),
CRediT authorship contribution statement e16511. https://doi.org/10.1371/journal.pone.0016511.
Huang, K., Yang, X.J., Hu, C.Y., et al., 2020. Short-term effect of ambient
Yi-Bin Cheng: Funding acquisition, Resources, Supervi- temperature change on the risk of tuberculosis admissions: assessments of
sion. Shilu Tong: Writing e review & editing. Yu-Shu Huang: two exposure metrics. Environ. Res. 189, 109900. https://doi.org/10.1016/
Conceptualization, Data curation, Formal analysis, Method- j.envres.2020.109900.
Hui-Beckman, J.W., Goleva, E., Leung, D.Y.M., et al., 2023. The impact of
ology, Software, Visualization, Writing e original draft. temperature on the skin barrier and atopic dermatitis. Ann. Allergy Asthma
He-Jia Song: Methodology, Validation. Xiao-Yuan Yao: Immunol. https://doi.org/10.1016/j.anai.2023.08.007.
Conceptualization, Funding acquisition, Methodology, Super- Konkel, L., 2020. Diabetes in Brazil: the association between extreme heat and
vision. Rui Zhang: Writing e review & editing. Yong-Hong hospitalization. Environ. Health Perspect. 128 (7), 74004. https://doi.org/
Li: Conceptualization, Funding acquisition, Methodology, 10.1289/ehp6853.
Lei, X., Liu, L., Chen, R., et al., 2021. Temperature changes between neigh-
Supervision, Writing e review & editing boring days and childhood asthma: a seasonal analysis in Shanghai, China.
 HUANG Y.-S. et al. / Advances in Climate Change Research 14 (2023) 847e855 855

Int. J. Biometeorol. 65 (6), 827e836. https://doi.org/10.1007/s00484-020- Wang, Y., Chen, Y., Chen, J., et al., 2021. Mortality risk attributable to diurnal
02057-x. temperature range: a multicity study in Yunnan of Southwest China. En-
Li, Y., Cheng, Y., Cui, G., et al., 2014. Association between high temperature viron. Sci. Pollut. Res. Int. 28 (43), 60597e60608. https://doi.org/10.1007/
and mortality in metropolitan areas of four cities in various climatic zones s11356-021-14981-5.
in China: a time-series study. Environ. Health 13, 65. https://doi.org/ Wang, Y., Ye, D., Cheng, Y., et al., 2022. Seasonal variation in association
10.1186/1476-069x-13-65. between temperature change and emergency department visits: a multi-site
Lin, H., Zhang, Y., Xu, Y., et al., 2013. Temperature changes between study in China. Environ. Res. 214 (Pt 3), 113963. https://doi.org/10.1016/
neighboring days and mortality in summer: a distributed lag non-linear j.envres.2022.113963.
time series analysis. PLoS One 8 (6), e66403. https://doi.org/10.1371/ Wu, Y., Li, S., Zhao, Q., et al., 2022. Global, regional, and national burden of
journal.pone.0066403. mortality associated with short-term temperature variability from
Lin, S.Y., Ng, C.F.S., Kim, Y., et al., 2023. Ambient temperature and nervous 2000e19: a three-stage modelling study. Lancet Planet. Health 6 (5),
system diseases-related mortality in Japan from 2010 to 2019: a time- e410ee421. https://doi.org/10.1016/s2542-5196(22)00073-0.
stratified case-crossover analysis. Sci. Total Environ. 867, 161464. Yang, J., Yin, P., Zhou, M., et al., 2016. The effect of ambient temperature on
https://doi.org/10.1016/j.scitotenv.2023.161464. diabetes mortality in China: a multi-city time series study. Sci. Total En-
Ma, P., Zhang, Y., Wang, X., et al., 2021. Effect of diurnal temperature change viron. 543 (Pt A), 75e82. https://doi.org/10.1016/j.scitotenv.2015.11.014.
on cardiovascular risks differed under opposite temperature trends. Envi- Yang, Z., Yang, J., Zhou, M., et al., 2021. Hourly temperature variability and
ron. Sci. Pollut. Res. Int. 28 (29), 39882e39891. https://doi.org/10.1007/ mortality in 31 major Chinese cities: effect modification by individual
s11356-021-13583-5. characteristics, season and temperature zone. Environ. Int. 156, 106746.
Ma, Y., Jiao, H., Zhang, Y., et al., 2020. Impact of temperature changes be- https://doi.org/10.1016/j.envint.2021.106746.
tween neighboring days on COPD in a city in Northeast China. Environ. Zha, Q., Chai, G., Zhang, Z.G., et al., 2022. Impact of temperature changes
Sci. Pollut. Res. Int. 27 (5), 4849e4857. https://doi.org/10.1007/s11356- between neighboring days on cardiovascular disease hospital admissions
019-07313-1. among suburban farmers in Qingyang, Northwest China. Int. J. Bio-
Martinez-Nicolas, A., Meyer, M., Hunkler, S., et al., 2015. Daytime variation meteorol. 66 (6), 1233e1245. https://doi.org/10.1007/s00484-022-02271-9.
in ambient temperature affects skin temperatures and blood pressure: Zhan, Z., Zhao, Y., Pang, S., et al., 2017. Temperature change between
ambulatory winter/summer comparison in healthy young women. Physiol. neighboring days and mortality in United States: a nationwide study. Sci.
Behav. 149, 203e211. https://doi.org/10.1016/j.physbeh.2015.06.014. Total Environ. 584e585, 1152e1161. https://doi.org/10.1016/
Qu, Y., Zhang, W., Boutelle, A.M., et al., 2023. Associations between ambient j.scitotenv.2017.01.177.
extreme heat exposure and emergency department visits related to kidney Zhang, S., Yang, Y., Xie, X., et al., 2020. The effect of temperature on cause-
disease. Am. J. Kidney Dis. 81 (5), 507e516.e501. https://doi.org/10.1053/ specific mental disorders in three subtropical cities: a case-crossover study
j.ajkd.2022.09.005. in China. Environ. Int. 143, 105938. https://doi.org/10.1016/
Schmidt, M.I., Matos, M.C., Branchtein, L., et al., 1994. Variation in glucose j.envint.2020.105938.
tolerance with ambient temperature. Lancet 344 (8929), 1054e1055. Zhang, Y., Yu, C., Wang, L., 2017. Temperature exposure during pregnancy
https://doi.org/10.1016/s0140-6736(94)91711-6. and birth outcomes: an updated systematic review of epidemiological
Sellers, A.J., Pallubinsky, H., Rense, P., et al., 2021. The effect of cold evidence. Environ. Pollut. 225, 700e712. https://doi.org/10.1016/
exposure with shivering on glucose tolerance in healthy men. J. Appl. j.envpol.2017.02.066.
Physiol. 130 (1), 193e205. https://doi.org/10.1152/japplphy- Zhao, Q., Li, S., Coelho, M., et al., 2022. High ambient temperature and risk of
siol.00642.2020 (1985). hospitalization for gastrointestinal infection in Brazil: a nationwide case-
Tang, J., Xiao, C.C., Li, Y.R., et al., 2018. Effects of diurnal temperature range crossover study during 2000e2015. Sci. Total Environ. 849, 157836.
on mortality in Hefei city, China. Int. J. Biometeorol. 62 (5), 851e860. https://doi.org/10.1016/j.scitotenv.2022.157836.
https://doi.org/10.1007/s00484-017-1486-y. Zhu, Y., Yang, T., Huang, S., et al., 2022. Cold temperature and sudden
Valdes, S., Doulatram-Gamgaram, V., Lago, A., et al., 2019. Ambient tem- temperature drop as novel risk factors of asthma exacerbation: a longitu-
perature and prevalence of diabetes and insulin resistance in the Spanish dinal study in 18 Chinese cities. Sci. Total Environ. 814, 151959. https://
population: [email protected] study. Eur. J. Endocrinol. 180 (5), 273e280. doi.org/10.1016/j.scitotenv.2021.151959.
https://doi.org/10.1530/eje-18-0818.