The correlation between high temperature and cognitive function: a CHARLS 2018 cross-sectional study

Abstract

Background

Extreme temperature-related events can have a negative impact on the health of populations, especially the elderly. However, there is a lack of research on the relationship between heat exposure and cognitive function in older adults. We aim to explore the association between temperature and cognitive function through a national study of middle-aged and older adults.

Methods

This study was based on data from the 2018 wave of the China Health and Retirement Longitudinal Study, which screened 6,605 participants who met the inclusion criteria and merged historical meteorological data. We controlled for temperature-related meteorological variables and covariates affecting cognitive function, and used Generalized Linear Model to investigate the correlation between ambient temperature exposure and cognitive function in middle-aged and older Chinese adults. Stepwise regression was used to examine the mediating role of depression levels in the impact of temperature on cognitive function.

Results

After controlling for covariates and meteorological variables, the number of days of high temperature exposure was significantly negatively associated with cognitive function. The number of days of high temperature exposure was positively associated with depression, and depression was negatively associated with cognitive function. Mediation effect analyses indicated that depression playing a partial mediating role in the association between high temperature exposure and cognitive function.

Conclusions

High temperatures exposure was significantly associated with reduced cognitive function, and depression levels may partially mediate this association. This finding has important implications for the design of public policies to reduce the role of high temperatures on mental health.

Text box 1. Contributions to the literature
• This study is the first attempt to reveal the association between heat, cognitive function, and depression.
• This study found that high temperatures exposure was significantly associated with lower cognitive function, and depression levels may partially mediate this association.
• The study could provide scientific evidence for the development of public policies to reduce the impact of heat on mental health.

Background

As one of the greatest challenges of the 2lst century, climate change has been reported to harm human health through multiple pathways [1]. With climate change, extreme weather events are occurring frequently and increasing in intensity, causing serious human and financial losses. Extreme weather events are relevant to both the environment and public health [2, 3]. Extreme weather events are manifestations of climate variability, including heat waves, hurricanes and floods [3]. The relationship between high temperatures/heatwaves and adverse health outcomes has been studied in many diseases [4, 5]. It has been found that high temperatures and heat waves promote the occurrence of adverse health outcomes. Changes in temperature increase morbidity and mortality from diseases such as cardiovascular and respiratory diseases [6–8]. In recent years a growing body of research has focused on the fact that mental health can also be affected by high temperatures. High temperatures can increase the incidence of mental illness or worsen existing conditions. And the effects of climate change on mental illness include anxiety, depression, post-traumatic stress, sleep disorders, and cognitive disorders [9, 10].

Cognitive function encompasses a range of mental processes that are necessary for acquiring knowledge and understanding, including attention, memory, executive function, reasoning, problem-solving, and language [11]. Cognitive impairment is a common psychiatric disorder, one of the pre-existing features of dementia, and an important factor affecting the quality of life of older people in later life. Mild cognitive impairment is prone to evolve into dementia if left unchecked. At present, many scholars have conducted in-depth research on the relationship between environment and cognitive function [12], but they have different views. First, most scholars believe that high temperature will lead to cognitive function decline. For example, Willian’s study took indoor temperature as the exposure temperature and found that heat waves, heat stress and higher indoor temperature were associated with poorer cognitive function [13]. Some scholars have found that high temperatures lead to cognitive decline and that the effects are greater in older age groups [14, 15], especially those with lower levels of physical activity [16]. In addition, Taylor found that high temperatures affect the levels and balance of the neurotransmitters (serotonin and dopamine) in the brain, which play a role in mood, cognitive function, and complex task performance [17]. Second, a small number of scholars have concluded that short-term temperature exposure is not associated with cognitive function [18]. Moreover, “heat therapy” at appropriate temperatures can help reduce systemic insulin resistance, improve vascular function, and prevent and improve dementia [19]. Although the relationship between heat and cognitive function is controversial as mentioned above, most scholars do not completely deny that extreme heat can lead to cognitive decline.

Depression is a common mental disorder, characterized by persistent sadness and a lack of interest or pleasure in previously beneficial or enjoyable activities. It can also disturb sleep and appetite, disturb [20]. Because of its high prevalence, suicide rate, and relapse rate, depression is widely recognized as one of the most pressing mental health problems [21]. Depression is not only the leading cause of mental and physical disability globally, but also a significant contributor to the global burden of disease [22]. The risk of depression among Chinese elderly people is also not optimistic. According to the data from the Seventh National Census in 2020, the size of China’s elderly population at high risk of depression is estimated to be as high as 91,853,300 people. Researchers are now realizing that temperature has an effect on depression as well, with high temperatures leading to a decrease in a person’s mood and an increased risk of depression [23, 24]. Cianconi found that changes in mental health are difficult to detect immediately and are influenced by complex biopsychosocial pathways that are closely related to the local cultural, social, economic, developmental contexts, the type of weather event, and the duration and severity of the event [25]. Aruta found a nonlinear relationship between high temperatures and depression by controlling for meteorological variables, demographic characteristics, place control effects, and birth year control effects [23]. In addition, some demographic characteristics, including age, gender, climate zone, and economic status, contribute to high temperature vulnerability [26, 27]. Moreover, there is a U-type association or J-type association between temperature and depression, and both extreme low temperature and extreme high temperature lead to increased risk of depression [28, 29]. However, scholars do not have the same understanding of the pathways of action by which high temperature affects depression. Some scholars believe that high temperature affects the nervous system through direct action on the body’s biochemistry, while others believe that high temperature indirectly affects mental health by influencing factors such as subjective well-being, negative emotions, cognitive function, and sleep quality [30–33]. In conclusion, most scholars believe that temperature can have an effect on depression. And it can be affected by other factors such as demographic characteristics and meteorological factors, so these confounding variables need to be controlled for in research.

When examining the effects of high temperatures on mental health, depression and cognition are often included as dependent in studies [29, 34]. But we have found that there is also a correlation between cognition and depression. Major depression can cause cognitive impairment such as attention and memory difficulties [35, 36]. Many studies have shown that cognitive impairment can be a valid indicator of depressive tendencies as well as functional recovery in middle-aged and older adults, and that depression can further exacerbate cognitive decline in middle-aged and older adults [37–40]. This suggests that middle-aged and older adults are susceptible to the co-existence of cognitive impairment and depression, and that these two conditions reinforce each other. The relationship between depression and cognitive impairment is intricate, and there is no biological evidence of an underlying mechanism between the two [41]. But there are still many scholars who believe that depression is a risk factor for cognitive impairment. Scholars have found through Meta-analyses and clinical trials that depression is a risk factor for dementia rather than an early manifestation of dementia or a consequence of cerebrovascular insufficiency [42, 43]. Depression is often accompanied by organic damage to the human brain, and cognitive deficits persist even after recovery from mental illness [44]. Moreover, Yuan demonstrated that there is a structurally stable network between depression and cognitive impairment, and that decreased language ability and low mood are important nodes in the interaction between the two [45]. In conclusion, most studies have found a correlation between cognitive impairment and depression. Although there is no biological evidence of a causal relationship between cognitive impairment and depression, most scholars believe that depression is a risk factor for cognitive impairment.

The above literature review provides a theoretical foundation for this paper to study the relationship between temperature, depression and cognitive function, but there is still room for further analysis: (i) Previous studies have predominantly concentrated on the relationship between temperature and depression, temperature and cognitive function, and depression and cognitive function, with limited attention given to the interrelationship among temperature, depression and cognitive function [46]. In addition, scholars have often studied depression as an outcome variable, and less frequently examined the mediating role of depression between temperature and cognitive function. (ii) In the study of temperature on mental illness, previous studies have used pathological experiments to demonstrate that extreme temperatures can be detrimental to human cognition, but the small sample size is not generalizable. Even studies with large sample sizes are mostly conducted using patient admission data and rarely control for individual demographic characteristics of patients, which can easily lead to confounding bias [47, 48]. (iii) Some studies did not consider other meteorological variables when examining the relationship between temperature and cognitive function. Cianconi found that the effects of climate change on mental health were inconsistent and could vary depending on the type of climate to which one was exposed [49]. Also air temperature may be affected by other meteorological variables such as precipitation and air humidity [50–52]. Most of the current studies have been conducted on a small scale and most of the study sites are in economically developed areas and the findings may not be generalizable [53]. In summary, although most scholars believe that high temperatures impair cognitive function, their studies still have limitations such as small sample sizes and failure to control for confounding variables. In this context, this study uses cross-sectional data from the 2018 wave data of the China Health and Retirement Longitudinal Study (CHARLS) database, controlling for demographic characteristics and relevant meteorological variables, to analyse the association between the level of ambient temperature exposure and concurrent cognitive function in middle-aged and older populations.

Methods

Research design

Based on the theoretical analysis described above, we proposed the inter-conceptual relationship shown in Fig. 1, which is the main relationship between temperature, depression and cognitive function.

Fig. 1.

Conceptual relationship between temperature, depression and cognitive function

Data

Study population and covariates Study participants were drawn from the China Health and Retirement Longitudinal Study (CHARLS), a national cohort study of residents aged 45 years and older (https://charls.pku.edu.cn/). CHARLS is a nationally representative survey conducted by the China Center for Social Science Surveys (CCSSS) at Peking University, and supported by a multi-stage probability sampling methodology, the survey covers about 150 counties and cities in 28 provinces in China and aims to collect a wide range of high-quality microdata on middle-aged and elderly people. Moreover, it did not set a strict upper age limit for participants, who were followed from the time they became candidates for the baseline survey until their death. The survey covered demographic background, health status and functioning, socioeconomic status, and retirement information. After a baseline assessment in 2011, CHARLS followed up participants in three waves in 2013, 2015, and 2018. As CHARLS is a dynamic cohort, a small number of new respondents were recruited at follow-up [54]. And CHARLS retains the refreshment sample in the baseline survey. If the selected person is between 40 and 44 years old, he/she will be retained as a refreshment sample for future rounds of the survey [55]. Detailed information on CHARLS has been published previously [56]. In previous studies, more scholars have used this database for research related to mental illness. For example, Luo used CHARLS to study the relationship between obesity and depressive episodes in Chinese middle-aged and elderly people [57]; Hu used CHARLS to demonstrate that sarcopenia in elderly people is associated with more severe cognitive impairment [58]; and Yao used the CHARLS database to demonstrate that exposure to ambient particulate matter for a certain period of time in middle-aged and elderly people in China significantly reduces their cognitive function [59].

Cognitive function data were obtained from the Brief Mental State Evaluation Scale (MMSE) in CHARLS. The MMSE scale is often used to assess dementia status and the degree of cognitive deficits, and generally includes temporal orientation, place orientation, short-term memory, delayed memory, attention, computation, language, and visuospatial [60]. In this study, we referred to the simplified method used by Tan and Yi to calculate cognitive test scores: 30 questions in total, one point for each correct answer, with higher scores representing more sound cognitive function [61, 62]. According to the original scoring scale of the MMSE, it can be divided into three levels by score bands: 24–30 = no cognitive impairment; 18–23 = mild cognitive impairment; and 0–17 = cognitive impairment [63]. However, some studies suggests that socio-cultural, educational and other factors may influence cognitive function scores, and low-educated people may be underestimated due to the cultural and educational bias of the test content [60]. Due to the huge difference in education level among different regions in China, it is more reasonable to classify the critical point of cognitive impairment according to the level of education in this study. Because cognitive impairment is related to education level, when the subject’s education level is illiterate individuals, primary school or higher, and the MMSE score is less than 17, 20 and 24 points respectively, it can be considered that the subject has cognitive impairment and is recorded as 1, otherwise it is 0 [64]. We used the CES-D10 scale in CHARLS to measure self-reported mental health. The CES-D10 consists of 10 questions about depressive symptoms, with answers consisting of 4 options: 0 (rarely), 1 (sometimes; 1–2 days per week), 2 (occasionally; 3–4 days per week), and 3 (most of the time; 5–7 days per week) [65]. The total score ranges from 0 to 30 and reflects the severity of depressive symptoms. Higher scores indicate more severe depressive symptoms [66]. Where a total score of at least 10 is considered to have depressive symptoms [67]. Age, gender, education, and marital status all affect cognitive function, and CHARLS provides this demographic information, and we used the above variables as control variables [68, 69].

Based on the high quality and national representativeness of the CHARLS data and the fact that more scholars have used the database for research related to mental illness, this study obtained data on demographic characteristics, mental health and cognitive function from the 2018 wave of CHARLS. This wave of surveys was conducted intensively in July and August, and the remaining scattered survey respondents were investigated in September and November. Temperatures in several parts of China in July and August 2018 were recorded to have exceeded historical maximums [70]. After excluding participants who were < 45 years old and did not complete cognitive function and depression tests, the number of participants who met the inclusion criteria was 6605 (as shown in Fig. 2). Study participants came from 125 counties in China (as shown in Fig. 3).

Fig. 2.

The flow chart of 2018 wave enrolled participants in the study

Fig. 3.

Baseline interview area

Note: The yellow areas represent the cities covered by the CHARLS data used in this paper

Meteorological data Meteorological information of 2018 was obtained from the National Oceanic and Atmospheric Administration (NOAA), and the relevant meteorological data used include daily average temperature, average wind speed, average relative humidity, precipitation, and hours of sunshine (Daily Observat https://www.ncei.noaa.gov/maps/daily/ ional Data). Most of the existing literature on extreme heat uses meteorological data from surface monitoring weather stations to assess the effects of temperature [2, 71, 72].

First, this study calculates city-level daily meteorological variables based on station-level meteorological variables using the distance inverse weighting method for the month prior to the survey date. Then, the monthly averages of each meteorological variable were calculated. Further, the number of days in which the sample’s daily mean temperature fell into one of eight temperature bins (≥ 30℃, 25–30℃, 20–25℃, 15–20℃, 10–15℃, 5–10℃, 0–5℃, < 0℃) in the month prior to the survey was calculated. The study defines the 20–25℃ bin as the reference temperature bin and omits this bin in subsequent regressions to prevent multicollinearity. In addition, taking into account the vastness of China, its wide latitudinal range, the large differences in distance from the sea, and the differences in altitude, there is a wide variety of combinations of temperature and precipitation, resulting in a wide variety of climates. It is necessary to consider the effects of different temperature zones on the results. Therefore, temperature zones were classified according to Chinese temperature zone division standard, and this variable was included in the statistical analysis in this study [73].

Data analysis

Initially, we explored the relationship between cognitive function and exposure to temperature using Generalized Linear Model (GLM) adjusted for mean wind speed, mean relative humidity, precipitation, and sunshine duration [74]. In addition, we adjusted for the following covariates for each measurement: age, gender, education level, residence and marital status. These covariates were associated with exposure and outcome rather than intermediate variables between exposure and outcome. To examine the nonlinear effects of temperature exposure on cognitive function scores, the authors divided the ambient temperature into eight temperature bins and constructed the basic equation as follows:

1

.

where i represents the individual, c represents the city, and m represent the month of the survey. Y_icm is the cognitive ability score result is the cognitive ability score of individual i in city c, month m. W_cm is a series of other meteorological control variables, including precipitation, relative humidity, wind speed, and sunshine duration. We also control for a vector of individual demographic characteristics X_i, including sex, age, education, marital status, etc., λ_c is a city fixed effect which can control for confounding factors at city level that affect mental health. π_m is month of survey fixed effects. ω_t is temperature zone fixed effect. ε_icm is the error term, which captures time-varying and city-specific unobservable factors affecting mental health. α is the intercept.

Second, we used logistic regression to explore whether temperature exposure was an influential factor in cognitive impairment. In addition, in the sensitivity analysis, this study used a series of methods to test the stability of the results. First, the analysis was performed again after replacing the 2018 data with data from 2011, 2013, and 2015, respectively. Second, the study replaced the dependent variable and re-analysed the natural logarithm of MMSE scores as the dependent variable. Third, the regression test was rerun after excluding the 5% and 95% extremes of the MMSE scores.

Finally, in order to test the influence path between temperature and cognitive function, this paper explored the mechanism of action between high temperature and cognitive function by constructing a mediation effect model. Since stepwise regression method can effectively overcome the problem of multicollinearity and has been applied to mediation effect test, this study referred to most of the studies and constructed the mediation effect model based on stepwise regression method [75]. In this study, based on Eq. (1), stepwise regression test is used to test the mediation effect. First, the estimated coefficient θ in Eq. (1) is tested. If θ is significant, it indicates that temperature is associated with participants’ cognitive function, and the existence of mediating effect is initially established. Second, the estimated coefficient of Eq. (2) is θ’₁ and the estimated coefficient of Eq. (3) is λ₁, which, if significant, indicates the existence of an indirect effect. Finally, the estimated coefficient of Eq. (3) is tested to be θ’’₁, and if θ’’₁ is significant and λ₁ is significant, there is a direct effect, meaning that there is a partial mediation effect; if θ’’₁ is not significant and λ₁ is significant, then the mediating effect is complete.

2

3

Where MDD_icm denotes the mediating variable of depression, and the other variables are described in Eq. (1). In order to verify the robustness of the mediating effect, this study utilized GLM and logistic regression for mediation effect analysis using the scores and categorical variables of depression symptom and cognitive function from the 2018 wave of CHARLS as dependent variables, respectively.

All of the above analyses were performed using Stata16 and the significance level was set at 5%.

Results

Descriptive analysis

Samples missing demographic characteristics were excluded from the analysis. The final sample consists of 6605 data from 125 prefecture-level cities in 25 provinces. We merged personal data and city-level meteorological data based on respondent ID number and city of life. Table 1 provides summary statistics. In Panel A, the mean age of the participants is 63.07 years, with the majority of participants aged 60–69 years, 47.46% of the participants are female, and most of them were educated. In addition, 87.4% of the participants are married and 48.34% lived in the city. The mean of the MMSE scores was 15.40, indicating that participants’ overall perceptions were low. And only 35% of the participants had symptoms of depression, indicating that the majority of participants were psychologically healthy.

Table 1.

Summary characteristics of individual level variables and City level meteorological variables

Variable	±SD/N(%)	Variable	±SD/N(%)
Panel A: Individual level variables(N = 6605)		Cognition	15.40 ± 5.34
Age(y)	63.07 ± 8.18	no cognitive impairment	746(11.29)
<60	2398(36.30)	cognitive impairment	5859(88.71)
60–69	2742(41.51)	Depression	8.09 ± 6.24
70–79	1252(18.96)	no depressive symptoms	4297(65.06)
>=80	213(3.22)	depression symptoms	2308(34.94)
Sex
male	3470(52.54)
female	3135(47.46)	Panel B: City level meteorological variables(N = 200)
Educational level		Number of days (AT≥ 30℃) (d)	0.95 ± 1.56
illiterate individuals	746(11.29)	Number of days (AT 25–30℃) (d)	4.03 ± 2.55
elementary school	2869(43.44)	Number of days (AT 20–25℃) (d)	1.63 ± 2.43
middle school and above education	2990(45.27)	Number of days (AT 15–20℃) (d)	0.37 ± 1.30
Marital status		Number of days (AT 10–15℃) (d)	0.03 ± 0.25
married	5773(87.40)	Precipitation (mm)	151.15 ± 85.54
divorced, widowed, unmarried, etc.	832(12.60)	Relative humidity (%)	72.32 ± 10.16
Area of residence		Wind speed (m/s)	2.14 ± 0.47
unban	3193(48.34)	Sunshine duration (h)	191.78 ± 45.79
rural	3412(51.66)	Air quantity
Self-assessed health		excellent	46(23.00)
very good	734(11.11)	well	116(58.00)
good	793(12.01)	light pollution	35(17.50)
fair	3377(51.13)	medium pollution	3(1.50)
poor	1331(20.15)
very poor	370(5.60)

Note: Our sample consists of 6,605 individuals living in 125 cities in 25 provinces. The number of meteorological variables (N) in Panel B are unique values generated based on the survey city and survey date

The Panel B provides descriptive statistics for meteorological variables at the city level. The mean values of the number of days falling into the five temperature boxes are 0.95 days for ≥ 30℃, 4.03 days for 25 ~ 30℃, 1.63 days for 20 ~ 25℃, 0.37 days for 15 ~ 20℃, and 0.03 days for 10 ~ 15℃. The other meteorological variables were: precipitation 151.15 mm, relative humidity 151.15%, wind speed 2.14 m/s, and sunshine duration 191.78 h.

Relationship between temperature exposure and cognitive function

We first investigated the effect of period exposure to temperature on cognitive function in middle-aged and older adults through GLM. Table 2 provides the main results. The dependent variable is the MMSE score. In column 1, we included only city fixed effects, temperature zone fixed effects and month fixed effects. Estimates showed that the number of days of exposure above 30℃ was negatively associated with cognitive function scores (β₁ = -0.051, P < 0.05). In columns 2 to 3, we further added control variables for individual demographic characteristics and other meteorological control variables. The results show that the estimated coefficients of extreme hot days remain statistically significant after adding demographic characteristics and other meteorological control variables (β₂=-0.041, P < 0.05; β₃=-0.043, P < 0.05). That is, days of exposure above 30℃ negatively associated with cognitive function. Although the estimated coefficients were relatively small in absolute terms, the threat of global warming to mental health cannot be underestimated with the increasing frequency of heat waves. Moreover, all individual demographic characteristics had a significant associated with the cognitive function of middle-aged and older adults. In addition, many studies have found that other meteorological variables may affect cognitive function [52]. The results showed that only the coefficients of sunshine duration and moderate air pollution were significant (β_sun = 0.086, P < 0.05; β_mp=-0.053, P < 0.05), which suggests that longer sunshine duration are associated with higher cognitive function. Whereas, exposure to moderate air pollution for a short period of time was negatively associated with cognitive function scores compared to good quality air.

Table 2.

Effects of temperature exposure on MMSE score

Variable	β
	(1)	(2)	(3)
Number of days (AT ≥ 30℃)	-0.051**	-0.041**	-0.043**
	(0.084)	(0.078)	(0.081)
Number of days (AT25–30℃)	-0.028	-0.011	0.011
	(0.051)	(0.047)	(0.050)
Number of days (AT15–20℃)	0.015	0.030**	0.027*
	(0.099)	(0.094)	(0.097)
Number of days (AT10–15℃)	-0.017	-0.017	-0.013
	(0.886)	(0.818)	(0.825)
Age		-0.269***	-0.271***
		(0.008)	(0.008)
Sex (ref: male)
female		-0.029**	-0.030***
		(0.123)	(0.123)
Educational level (ref: illiterate individuals)
elementary school		0.180***	0.179***
		(0.178)	(0.178)
middle school and above education		0.102***	0.101***
		(0.467)	(0.468)
Marital Status (ref: divorced, etc.)
married		0.064***	0.063***
		(0.189)	(0.189)
Area (ref: rural)
city		0.074***	0.075***
		(0.144)	(0.148)
pre			0.028
			(0.002)
sun			0.086**
			(0.004)
humid			0.052
			(0.019)
wind			-0.013
			(0.219)
Air quantity (ref: excellent)
well			-0.006
			(0.197)
light pollution			-0.030
			(0.309)
medium pollution			-0.053**
			(0.548)
City FE	Yes	Yes	Yes
Temperature zone FE	Yes	Yes	Yes
Month FE	Yes	Yes	Yes

Notes: (1) in column 1, we only include city fixed effects, temperature zone fixed effects and month fixed effects. In columns 2 to 3, we further add individual demographic characteristic control variables and other meteorological control variables. (2) ^*, ^**, and^***, respectively, indicates significance at the 10, 5, and 1% level. Robustness standard error in Parentheses

In order to further verify the impact of extreme heat on cognition, cognitive impairment was taken as the dependent variable (0 = no cognitive impairment, 1 = cognitive impairment), and the influencing factors of cognitive impairment in the elderly were screened by logistic regression. The results are shown in Table 3. The results showed that the number of days of extreme heat exposure and age were risk factors for cognitive impairment with OR of 1.133 (95% CI, 1.022–1.256) and 1.438 (95% CI, 1.285–1.609). And female and married were protective factors for cognition with OR of 0.680 (95% CI, 0.575–0.803) and 0.522 (95% CI, 0.386–0.707). In contrast, influences such as precipitation, relative humidity, wind speed, sunshine, and air quality were not associated with cognitive impairment [76, 77].

Table 3.

Analysis of factors influencing cognitive function

Variable	OR (95%CI)
Number of days (AT ≥ 30℃)	1.133(1.022–1.256)
Number of days (AT25–30℃)	1.051(0.987–1.120)
Number of days (AT15–20℃)	0.992(0.891–1.105)
Age	1.438(1.285–1.609)
Educational level	1.418(1.253–1.604)
Sex (ref: male)
female	1.680(1.575–1.803)
Marital Status (ref: divorced, widowed, unmarried, etc.)
married	0.522(0.386–0.707)
Area (ref: rural)
city	0.882(0.725–1.072)
Precipitation	0.999(0.997–1.001)
Relative humidity	0.995(0.966-1.000)
Wind speed	0.989(0.966–1.014)
Sunshine duration	1.307(0.982–1.738)
air quantity	1.083(0.897–1.307)
City FE	Yes
Temperature zone FE	Yes
Month FE	Yes

Sensitivity analysis

In order to test the stability of the regression results, we conducted a sensitivity analysis by (1) using different years of survey data, (2) changing the dependent variable, (3) excluding extreme values. The results are shown in Table 4.

Table 4.

Sensitivity analysis result

	β(S.E.)
Variable	(1)			(2)	(3)
	2011	2013	2015	lnMMSE	MMSE
Number of days	-0.049**	-0.038**	-0.041***	-0.043**	-0.040**
(AT≥ 30℃)	-0.061	-0.033	-0.053	-0.072	-0.018
Number of days	-0.029	0.007	0.019	-0.005	0.006
(AT25–30℃)	-0.041	-0.023	-0.029	-0.043	-0.013
Number of days	-0.062***	-0.039***	-0.011	0.048***	0.005
(AT15–20℃)	-0.064	-0.054	-0.046	-0.084	-0.026
Number of days	-0.077***	0.01	0.020*	-0.021*	-0.007
(AT10–15℃)	-0.098	-0.073	-0.172	-0.653	-0.196
Number of days	-0.030**	0.002
(AT5–10℃)	-0.22	-0.175
Number of days	0.005	0.006
(AT0–5℃)	-0.195	-0.346
Number of days	-0.022	-0.054*
(AT < 0℃)	-0.62	-3.414
demographic controls	Yes	Yes	Yes	Yes	Yes
Weather controls	Yes	Yes	Yes	Yes	Yes
City FE	Yes	Yes	Yes	Yes	Yes
month FE	Yes	Yes	Yes	Yes	Yes

Notes: ①Model (1) uses survey data from 2011, 2013 and 2015, model (2) uses the natural logarithm of the MMSE as the dependent variable, and model (3) excludes extreme values. ② ^*, ^**, and ^***, respectively, indicates significance at the 10, 5, and 1% level. Robustness standard error in Parentheses

First this study tested the stability of the results by using data from different years. This study used the data of 2011, 2013, and 2015 instead of the data of 2018, respectively. Column (1) in Table 4 shows that the more days of exposure to extreme high temperatures (AT ≥ 30℃), the worse the cognitive function (β₂₀₁₁=-0.049,P < 0.05; β₂₀₁₃=-0.038,P < 0.05; β₂₀₁₅=-0.041,P < 0.01). Second, the natural logarithm of MMSE score was used as the dependent variable to estimate the effect of temperature on cognitive function. Column (2) in Table 4 shows that the number of days of exposure to extreme heat was associated with cognitive function scores ((β=-0.043, p < 0.05), indicating that more days of exposure to extreme heat being associated with poorer cognitive function. Third, this study excluded the extreme values and re-run regression tests after excluding 5% and 95% of the extremes in MMSE scores. Column (3) in Table 4 shows that increased days of exposure to extreme heat associated with lower cognitive function scores (β=-0.040, P < 0.05). The results of the sensitivity analyses all indicated a negative correlation between the number of days of exposure to extreme heat and cognitive function. This result further verified the stability of the results.

Mediating role of depression between temperature effects and cognitive function

To further test the mechanism of the effect of high temperature on cognitive function, we used stepwise regression to verify the mediating effect of depressive symptoms between high temperature and cognitive function. And to verify the robustness of the results, the study used both GLM and logistic regression methods. The results are shown in Table 5, (1)-(3) are GLM results and (4)-(6) are logistic regression results. Column (1) shows that the increase in the number of high temperature days is significantly associated with higher depressive symptoms scores (β = 0.026, P < 0.05), and the coefficients of Number of days (AT ≥ 30℃) (β=-0.046, P < 0.05) and CES-D10 (β=-0.172, P < 0.001) were significant in Column (3), suggesting that depression mediates the association between extreme high temperature and cognitive function. In (4) and (6), the corresponding ORs are 1.062, 1.142 and 1.846, respectively, and these three coefficients are also significant at the 5% level. The main findings of the effect estimations were consistent between the two methods. It can be inferred that exposure to extreme heat is concurrently associated with increased depression levels, which in turn correlate with reduced cognitive function. The specific mechanism of interaction between high temperature, depressive symptoms and cognitive function is shown in Fig. 4.

Table 5.

Mediating role of depression in temperature effects on cognitive function

variable	β(S.E.)			OR (95%CI)
	CES-D10	MMSE	MMSE	Depression	Cognition	Cognition
	(1)	(2)	(3)	(4)	(5)	(6)
Number of days	0.026**	-0.044**	-0.046**	1.062	1.134	1.142
(AT≥ 30℃)	-0.059	-0.08	-0.079	(1.013,1.115)	(1.023,1.258)	(1.029,1.268)
Number of days	0.033**	0.009	0.006	1.029	1.052	1.055
(AT25–30℃)	-0.033	-0.05	-0.05	(1.002,1.056)	(0.988,1.121)	(0.990,1.124)
Number of days	-0.002	0.027*	0.024	1.004	0.988	0.994
(AT15–20℃)	-0.096	-0.097	-0.096	(0.931,1.084)	(0.887,1.101)	(0.892,1.108)
Number of days	0.003	-0.013	-0.011	0.897	1	1
(AT10–15℃)	-0.863	-0.823	-0.811	(0.449,1.795)
CES-D10 / Depression			-0.172***			1.846
			-0.011			(1.526,2.232)
demographic controls	Yes	Yes	Yes	Yes	Yes	Yes
Weather controls	Yes	Yes	Yes	Yes	Yes	Yes
City FE	Yes	Yes	Yes	Yes	Yes	Yes
Month FE	Yes	Yes	Yes	Yes	Yes	Yes

Notes: ^*, ^**, and ^***, respectively, indicates significance at the 10, 5, and 1% level. Robustness standard error in Parentheses

Fig. 4.

Mechanisms of action between extreme heat (AT ≥ 30℃), depressive symptoms, and cognitive function

Discussion

Temperature is negatively correlated with cognitive function

In our study, the scores of the MMSE scale, the main tool for assessing cognitive function, provided us with important information about the cognitive status of the participants. The results of our study showed that after controlling for confounding and meteorological variables, increased temperature was associated with decreased MMSE scores. This is the same as the findings of Gong [50]. Medical evidence suggests that higher ambient temperatures can have an effect on cognitive performance, mainly in the form of decreasing an individual’s attention, memory, and ability to process complex information [78]. Moreover, by setting temperature intervals, we found that the association between temperature and cognitive impairment is nonlinear. When the average temperature is greater than 30℃, increased days of exposure were significantly associated with cognitive function decline. This is in line with the results of a similar study, also conducted in China, which found a negative effect of high temperature on cognitive function even after controlling for demographic characteristics and meteorological factors [79]. Another study also found that short-term indoor environmental temperatures alter cognitive function [80]. In the process of temperature affecting cognitive function the possible etiological pathway can be partially explained by pathological and physiological roles of the nervous system. Dopamine is the neurotransmitter responsible for the performance of complex tasks. High temperature increases plasma 5-hydroxytryptamine which then inhibits dopamine production, leading to reduced memory and inability to concentrate [17, 27]. Moreover, with increasing age, older adults experience a gradual decline in regulatory abilities such as cardiovascular function, which makes them more susceptible to heat stress and reduces cognitive function [81]. It is important to note that due to the older age and low education characteristics of the participants in this study, the mean MMSE score was approximately 15.4. The low MMSE scores may have masked the negative effects of high temperature on cognitive function to some extent. However, our analyses still found a significant negative association between high temperature and cognitive function. And after we classified cognitive impairment according to educational attainment, we still found high temperature exposure remained significantly associated with depression risk. This persistent association suggested that the linkage between high temperature and cognitive function may be underestimated. As climate warming intensifies, implementing targeted interventions to protect older adults from prolonged heat exposure may help mitigate cognitive impairment risks.

Extreme cold temperature is not found to be associated with cognitive function in the nonlinear regression analysis in this study, which is consistent with the results of a study conducted in Shanghai [82]. But according to other scholars, extreme cold temperatures can also have an impact on cognitive function [80]. We speculate that the reason for this result may be that the survey months of CHARLS are mainly concentrated in summer, and the sample size in winter is smaller, which reduces the statistical validity. Moreover, in winter, centralized heating and underfloor heating are common in northern China [83], and some parts of the South also use air conditioner to heat their homes, and people are less likely to be directly exposed to extreme low temperatures, so the association between low temperatures and cognitive function is not significant.

Reading through the literature we learned that cognitive function is related to age, sex, education level and marital status, and our study verified their relationship with cognitive function. First, the study found that the older the age, the worse the cognitive function. This is the same as previous findings, which found that aging is associated with declining cognitive function [68, 84]. Second, we found that women had poorer cognitive performance than men, which is consistent with the findings of Jia and Wang, who found that older women were more likely to have memory impairment than older men [69, 85]. In addition, the higher the level of education, the higher the MMSE scores. It has been found that early higher education can delay cognitive decline due to normal aging [86]. Highly educated individuals, who understand the dangers of high temperatures, actively seek measures to diminish the effects of high temperatures when traveling or choosing housing [87]. After controlling for these confounders, our findings remain significant, suggesting that higher temperatures are indeed associated with a decline in cognitive function.

Moreover, this study was robust by changing three methods: using different years of data (using data from the years 2011, 2013 and 2015), changing the dependent variable (using the logarithm of the MMSE score as the dependent variable), and removing the extremes (re-running the regression test after removing the extremes of 5% and 95%). The main results of all three methods were significant. In summary, high temperatures are negatively associated with cognitive function.

Mediating role of depression in temperature effects on cognitive function

This study verified the mediating role of depression symptoms in the association between high temperature and cognitive function through stepwise regression. This result is the same as the findings of other scholars. Previous studies have suggested that high temperatures affect the body’s biochemistry, which in turn affects the level of depression and cognition: high-temperature environments are associated with altered platelet serotonin levels, which have been linked to the pathogenesis of psychiatric disorders such as depression and schizophrenia [88, 89]. This suggests that heat exposure can have adverse effects on both depression and cognition. Consistent with Zhang’s findings, the present study found a negative correlation between heat exposure on mental health [90]. In terms of physiological mechanisms, a possible explanation for this finding is that the central nervous system is particularly vulnerable to high temperatures, especially in prolonged or extreme situations [91]. In addition, sustained high temperatures over several days or weeks can affect an individual’s mental state and behaviour, leading to irritability and mental health problems [27]. What’s more, the link between depression and cognitive decline in the elderly is well documented. Recent studies have shown that approximately 25% of older adults have both cognitive impairment and depression [92], and the patients with depression have a higher risk of cognitive impairment [93]. Sun found that subthreshold depressed patients had significantly poorer attention and working memory performance, and their study revealed a significant negative correlation between the severity of depressive symptoms and attention and working memory capacity [40]. And Tan have demonstrated that depression can be used as a mediating variable to influence cognitive function [61]. According to these findings, it can be inferred that depressive symptoms can also be an important mediating mechanism for high temperature to impair cognitive functions such as executive function and memory. In addition, it is important to note that 65.06% of the participants in this study had no depressive symptoms. But even so, our results still show that the mediating pathway still holds. This suggests that high temperatures may still negatively affect cognitive function through the mood pathway, even in a sample with better overall mental health. This study not only further enriches previous evidence suggesting that high temperatures are negatively associated with depressive symptoms and cognitive function, but also finds that depressive symptoms play a mediating role in the association between high temperatures and cognitive function, highlighting the broader impact of high temperatures on mental health. Although our findings demonstrate that depression can act as a mediating variable to influence the association between high temperature and cognitive function. However, this may be one of the earlier studies to demonstrate the validity of this pathway of influence, and a large number of future studies are still needed to validate the relationship between temperature, depression and cognitive function.

Strengths and limitations

The main strengths of this study are that the sample size is large and nationally representative, and potential confounders are considered in assessing the association between temperature exposure and cognitive function. And through a reading of the literature, it was found that this study may be one of the earlier studies to test the mediating role of depression in the association between temperature and cognitive function.

However, there are several limitations that should be acknowledged: first, this study was unable to collect precise air temperature data at participants’ specific locations, and the ambient air temperatures used were measured at fixed measurement points without assessing individual exposures. Therefore, there is a possibility of misclassification of temperature exposure. However, in a previous study conducted in Augsburg, Germany, individually measured temperatures were highly correlated with ambient temperatures, so the error was acceptable [94]. In addition, the database did not investigate participants’ use of indoor air conditioners, so the effect of high temperatures on cognitive function may have been underestimated. However, this is a situation that is difficult to avoid in many studies [62, 69]. Second, the air temperature used in this study is the dry bulb temperature measured at the point of measurement, which reflects air temperature only. This may underestimate the effect of humidity on body temperature and physiological burden. Third, due to data issues, we used temperature boxes to represent temperature exposure in the month before participants were surveyed. This may have resulted in an inability to capture the immediate effects of short-term climatic fluctuations (e.g., heat waves) on cognitive function. Fourth, the primary participants in this cohort survey are older adults, and confounding effects need to be taken into account. According to previous studies, participants lost to follow-up are generally in poor physical health [95]. Therefore, effect of high temperatures on cognitive impairment may be underestimated in this study, which may have contributed to the insignificance of the results. Fifth, some exposure variables such as social networks and dietary structure were not considered due to database limitations, which may also have an impact on the results.

Conclusion

The study used a wave of data from a national cohort of middle-aged and older adults to examine the association between exposure to ambient temperature and cognitive function. This study analysed the effects of temperature on cognitive function in middle-aged and older adults using GLM and the temperature box. Our findings show a non-linear relationship between temperature and cognitive function. The cognitive function scores undergo small but statistically significant changes with increasing days of exposure to extreme heat (AT > 30℃). And the study has found that extreme heat exposure is a risk factor for cognitive impairment. In addition, the study found that depression symptoms mediated the association between high temperature and cognitive function through stepwise regression. In the context of climate change worsens, future studies should examine the potential pathway whereby meteorological factors may influence cognitive function through depression.

Although some of the observed estimates are small, the results still have potential implications for public health, policymakers, and clinicians in practice. First, we found a negative correlation between extreme heat exposure and cognitive function. Although the warming trend cannot be stopped, governments can improve local microclimates by investing more in heat protection facilities in urban public spaces. For example, installing more sunshades and spray cooling equipment in parks, squares, and other outdoor places where the elderly are often active. At the same time, urban greening is being rationally planned to increase tree vegetation coverage in order to reduce ground temperatures. For old neighbourhoods, especially those with a high concentration of elderly people, building insulation and renovation projects should be implemented. In addition, for low-income elderly people living alone, community workers should pay regular visits during hot weather to check whether their living environment is safe and provide them with the necessary assistance to protect them from the hot environment. Second, we have found that high temperatures can cause cognitive decline by causing depressive symptoms. Therefore, community-level screening for cognitive function and depressive symptoms can be carried out before the high-temperature season each year to identify people at risk. And timely psychological counselling can be provided to the high-risk elderly population. At the same time, family doctors need to regularly enquire about the physical condition of the elderly during the hot season, especially sleep and mood changes. And they also need to remind elderly people not to go out as much as possible during the hot period and to be prepared for inclement weather.

Author contributions

BL and L-SK conceptualised the paper. L-SK did the statistical analysis and drafted the manuscript. DC and J-DZ conducted the research and data collection. BL and Y-LZ reviewed and edited the writing. BL and X-FC re-reviewed and co-revised the manuscript from the English language perspective. BL is responsible for the overall content as guarantor. All authors made significant intellectual contributions to multiple revisions of the draft. All authors have read and agreed to the published version of the manuscript.

Funding

This research was funded by 2024 Guangdong Provincial Natural Science Foundation Upper Level Project, grant numbers 2024A1515010761; General Project of Philosophy and Social Science Planning of Guangdong Province in 2023, grant number GD23CGL09.

Data availability

The datasets used and analysed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

All subjects signed an informed consent form and CHARLS was approved by the Institutional Review Board of Peking University, the IRB approval number for the main household survey is IRB00001052-11015.

Consent for publication

Not Applicable.

Competing interests

The authors declare no competing interests.