Impact of Climate Warming on All‐Causes Mortality in China: The Role of Non‐Optimal Temperature Exposure and Multi‐Scenario Projections to 2050 Authors: Yujun He, Conglin Song, Dejun Chen, Yuan He, Xiaojun Li, Xuequan Wang, Zhaoyang Liu, Tao‐Hsin Tung Published: 2026 ABSTRACT AbstractGlobal warming intensifies non‐optimal temperature exposure (consisting of high and low temperature), which poses severe threats to human health. As the world's most populous country located in temperate and subtropical zones, China is highly vulnerable to temperature‐related mortality, with distinct yearly and population disparities in health risks. To investigate the impact of non‐optimal temperature exposure on China's all‐causes mortality during 1990–2023 and project future trends (2024–2050) under different climate change scenarios. Integrated CN05.1 gridded observational data, CMIP6 climate data (SSP2‐4.5 and SSP5‐8.5 scenarios), SimmEX extreme temperature indices data, and GBD 2023 disease burden data. We applied Spearman's correlation analysis, Joinpoint regression, Age‐Period‐Cohort analysis, and restricted cubic spline models for trend quantification and future projections. Non‐optimal temperature‐related deaths increased, while age‐standardized mortality rate declined significantly. Low‐temperature exposure remained the primary risk, but high‐temperature‐related health impacts intensified notably. The elderly (≥80 years) and individuals with cardio‐cerebrovascular or respiratory diseases were the most susceptible. Climate warming showed significant correlations with most cause‐specific mortalities. Future projections revealed divergent disease risk trends under the two emission scenarios, with some diseases facing elevated risks. Non‐optimal temperature exposure exerts heterogeneous impacts on China's all‐causes mortality, with high‐temperature risks becoming increasingly prominent amid global warming. Targeted prevention and control strategies tailored to age, disease types, and emission scenarios are crucial to mitigate temperature‐related health burdens. Introduction Non‐optimal temperature exposure includes high‐temperature exposure (deviations above the theoretical minimum risk exposure level (TMREL)) and low‐temperature exposure (deviations below this level) (Liu et al., 2022). On the one hand, high‐temperature exposure systematically increases the mortality rate of various diseases through direct heat stress, altered infectious disease dynamics, deteriorated environmental quality and exacerbated socioeconomic inequality (Miller et al., 2024; X. Yang et al., 2024). On the other hand, short‐term exposure to low‐temperature environments has also been proven to significantly increase the risk of death from various diseases, especially cardiovascular and respiratory diseases (Álvarez‐Miño & Taboada‐Montoya, 2021). At present, global warming is one of the most severe challenges facing the world, and its impact on human health is becoming increasingly prominent. Although global warming can reduce the incidence of low‐temperature exposure events, it will also exacerbate the occurrence of high‐temperature exposure events. The health impacts of global warming are not uniformly distributed, and there are significant group and regional inequalities in the disease mortality risks it causes. Globally, low‐ and middle‐income countries, as well as tropical and subtropical regions, bear a higher disease burden (He et al., 2023). China is not only the most populous country in the world, but also a representative of such regions, and it is highly vulnerable to the health losses induced by exposure to non‐optimal temperatures resulting from global warming. It is reported that the mortality risk of urban Chinese populations caused by high temperatures is higher than the average level of other countries (X. Yang et al., 2024). In addition, large‐scale studies conducted in China have shown that during cold waves, total non‐accidental mortality rate, mortality rate of circulatory system diseases and mortality rate of respiratory system diseases all increased significantly (Sun et al., 2022). Therefore, systematically exploring the disease mortality burden caused by non‐optimal temperature exposure in China is of great significance for the enlightenment of health losses caused by global warming. Despite growing evidence on temperature‐related mortality in China, several gaps remain. First, previous studies have focused on limited outcomes. P. Yin et al. (2024) provided national attributable fractions for non‐optimal temperature but concentrated on overall mortality rather than cause‐specific burdens. Local studies such as Jian‐lei et al. (Jian‐lei et al., 2024) applied non‐linear models in single cities, mainly examining cardiovascular and respiratory mortality. Second, large‐scale national projections have largely relied on linear or piecewise assumptions, oversimplifying the non‐linear temperature‐mortality relationship (Yi et al., 2024; P. Yin et al., 2024). Third, while the GBD framework (Wei et al., 2022; P. Yin et al., 2024), CN05.1 data set, Coupled Model Intercomparison Project Phase 6 (CMIP6) projections and the China Extreme Temperature Indices Gridded Data set (SimmEX) (Yi et al., 2024; Zhang, Han, et al., 2023) have been used separately, few studies have integrated these three data sources in a unified framework from historical observation to future projection. Fourth, although the age‐period‐cohort method has been applied to cardiovascular mortality (Wei et al., 2022), its use for all‐causes mortality under non‐optimal temperature—especially with negative attributable burdens—has rarely been examined. Finally, scenario‐specific projections distinguishing high‐, low‐, and non‐optimal temperature exposures across diverse causes remain scarce. Against the above background, this study comprehensively utilized climate data of the CN05.1 gridded observational data set and the CMIP6, and the SimmEX for robustness verification. Combined with the disease data from the GBD, we explored the changes in the impact of non‐optimal temperature exposure on the all‐causes disease mortality burden in China during the period from 1990 to 2023. Materials and Methods Sources of Climate Data and Disease Burden Data in China At present, the latest CN05.1 gridded observational data set covers the temperature data of China from 1961 to 2022 (J. Wu et al., 2017). The latest CMIP6 temperature data cover the actual observational data from 1850 to 2014 and the projected temperature data up to 2,100 under different emission scenarios (Shared Socioeconomic Pathways (SSP) 5–8.5 and SSP2‐4.5) (Eyring et al., 2016). For the temperature data of CMIP6, the average temperature was calculated using 36 models, and the detailed information of these models is provided in Table S1 in Supporting Information S1. In addition, we adopted the China Extreme Temperature Index Grid Data set (SimmEX, 1961–2020) for extreme temperature indicator extraction. This data set contains a series of internationally recognized extreme temperature indices, including frost days representing low‐temperature exposure and summer days representing high‐temperature exposure. The latest GBD 2023 database includes data from 1990 to 2023 (GBD, 2023 Asia Chronic Respiratory Disease Collaborators, 2026). To ensure the temporal reliability of the analysis of the disease burden data corresponding to the years in the climate database, we extracted the average temperature data of China from 1990 to 2022 from CN05.1. We extracted the actual temperature data from 1990 to 2014 and the projected temperature data under SSP5‐8.5 and SSP2‐4.5 scenarios from 2015 to 2050 from CMIP6. We also retrieved annual extreme temperature indices (frost days and summer days) from 1990 to 2020 from the SimmEX. Meanwhile, disease burden data for all‐causes mortality attributable to non‐optimal temperature exposure in China from 1990 to 2023 were extracted from the GBD 2023 database. Data Analysis Correlation Analysis Between Climate Warming and All‐Causes Mortality Burden in China Firstly, we selected annual average temperature data from CN05.1 for the period 1990 to 2022, as well as data on all‐causes mortality burden attributable to non‐optimal temperature exposure (including overall non‐optimal temperature, high‐temperature, and low‐temperature exposure) in China from 1990 to 2022 from GBD 2023. We used annual average temperature as the climate indicator because the GBD database only provides annual aggregated mortality burden without daily or monthly granularity, which ensures consistency in temporal resolution between exposure and outcome variables. Spearman's correlation analysis was applied to explore the association between them. To further verify the robustness of the associations, we conducted sensitivity analysis using extreme temperature indices derived from the SimmEX. Annual average frost days and summer days were adopted to represent extreme cold and hot weather days, respectively. Furthermore, the sum of annual frost days and summer days was calculated as the total number of extreme non‐optimal temperature days for additional robustness assessment. All analyses and data visualization were performed using R software (Version 4.3.3). Overview of Non‐Optimal Temperature Exposure and All‐Causes Mortality Burden in China From 1990 to 2023 To ensure the reliability of burden estimation, the GBD adopts a Bayesian statistical framework and the Monte Carlo method to calculate the 95% uncertainty intervals (UIs). First, multi‐source data are integrated, and the DisMod MR 2.1 model is used for hierarchical modeling to correct data bias. Secondly, 500 random samplings are performed on each parameter to simulate a reasonable range of values, while retaining and propagating the uncertainties caused by measurement errors, model assumptions and regional heterogeneity. Finally, the 2.5th and 97.5th percentiles of the 500 samplings are used to generate UIs for quantifying the uncertainty of burden estimation. The 95% UIs indicate that the probability of the true value falling within this interval is 95%. The width of the interval reflects the robustness of the estimation: a wider interval indicates sparse data or high model uncertainty, whereas a narrower interval indicates sufficient data and stable model performance (“Global burden of 87 risk factors in 204 countries and territories 1990–2019,” 2020). The trend of the deaths of age‐standardized rate (ASR) was quantified using the estimated annual percentage change (EAPC), which was derived from a generalized linear model assuming a Gaussian distribution (Li et al., 2023). In the calculation of EAPC, the calendar year was set as the explanatory variable (X), and the natural logarithm of the age‐standardized mortality rate (ln [ASR]) was used as the dependent variable (Y). The data were fitted to the regression line Y = a + βX + ε. Then the EAPC was calculated using the fitted regression line parameter β according to the formula EAPC = 100 × (exp(β)−1) (Jiang et al., 2023). The calculation formula for the relative change (RC) is as follows: RC (%) = [(2023 Value − 1990 Value)/1990 Value] × 100% (Wang et al., 2024). In this section, R software (Version 4.3.3) was utilized for graphing. Joinpoint Regression Analysis The Joinpoint regression model consists of a series of linear statistical models, which is used to assess the temporal changes in the disease burden of all‐causes mortality in China induced by non‐optimal temperature exposure. The method of Joinpoint regression analysis is based on the approach proposed by Y. Zhang et al. (2022). This model adopts a mathematical approach that estimates the fluctuation of morbidity using the least squares method, thereby avoiding the inherent subjectivity in traditional trend studies that rely on linear trends. The inflection points of the change trend are determined by calculating the sum of squared residuals between the estimated values and the actual values. This model was developed using Joinpoint software (Version 5.1.0.0; National Cancer Institute, Rockville, Maryland, USA) and JD_GBDR (V2.22, Jingding Medical Technology Co., Ltd.). The software monitors the pattern of time‐series data and then applies the simplest model by connecting line segments on a logarithmic scale. The Average Annual Percentage Change (AAPC) and Annual Percentage Change (APC) are calculated to evaluate the trend; AAPC represents the geometric weighted mean of the annual percentage changes derived from Joinpoint trend analysis, with weights corresponding to the duration of each period within the specified time range (Kim et al., 2000). The 95% confidence interval (CI) is derived based on the linear regression model. An AAPC or APC value and 95% CI exceeding zero indicates an increasing trend in ASR, and vice versa. If the 95% CI of AAPC includes zero, the ASR is considered to remain unchanged over time. Age‐Period‐Cohort Analysis The Age‐Period‐Cohort model has been widely applied in sociology and epidemiology. This model is based on the Poisson distribution and can characterize the temporal trends of health outcomes from three dimensions: age, period, and cohort (Y. Yang et al., 2008). According to the GBD classification, children under 5 years old and individuals aged 95 years and above are classified into separate groups. In the Age‐Period‐Cohort model fitting, age groups are defined as <5 years, 5–9 years, 10–14 years…, ≥95 years, among which the “0” group represents the cohort under 5 years old. Data are aggregated into 5‐year intervals to calculate the total number of cases related to all‐causes mortality across different age groups. In the current analytical epidemiological Age‐Period‐Cohort analysis, the Intrinsic Estimator method is adopted to address the multicollinearity among age, period, and cohort effects. This method effectively disentangles the independent contributions of age, period, and cohort through specific constraints and estimation algorithms, thereby mitigating the estimation bias caused by multicollinearity in the traditional Age‐Period‐Cohort model (Y. Yang et al., 2008). The age effect denotes the variation in disease risk with age; the period effect explains the time‐related changes affecting the health outcomes of all age groups; the cohort effect reflects the outcome differences among individuals born in the same cohort (Luo, 2013). The formula of the log‐linear regression model is as follows: log (Yi) = μ + α × age i + β × period i + γ × cohort i + ε. Where Yi denotes the attributable risk of ASR of all‐causes mortality burden induced by non‐optimal temperature exposure; α, β and γ represent the coefficients of age, period and cohort respectively; μ is the intercept; and ε is the model residual. Model fitting is performed using the Web Tool (https://analysistools.cancer.gov/apc/), and visualization is conducted using R software (Version 4.3.3) and JD_GBDR (V2.22, Jingding Medical Technology Co., Ltd.). Because the APC log‐linear model requires positive ASR values, the presence of negative estimates for overall non‐optimal and low‐temperature exposure initially prevented model fitting. Given the presence of negative attributable mortality values in certain age–year strata for non‐optimal temperature and low‐temperature exposure, we conducted supplementary sensitivity analyses to validate the robustness of our results. we performed a sensitivity analysis in which all negative ASR values were set to zero, allowing the APC model to converge (Mathers et al., 2006). Negative attributable mortality estimates are not errors but mathematically valid outputs of the GBD framework when the current exposure level to a risk factor lies below the counterfactual exposure distribution (GBD, 2023 Disease and Injury and Risk Factor Collaborators, 2025). In such cases, the attributable fraction is negative, indicating a net protective effect of the risk factor relative to the TMREL. Prediction of the All‐Causes Mortality Burden in China Under Two Different Emission Scenarios Given that the relationship between temperature and mortality is widely recognized as nonlinear in epidemiological studies, typically exhibiting a J‐shaped or U‐shaped curve (Alahmad et al., 2025), we employed restricted cubic splines (RCS) for the projection. First, Pearson correlation analysis was performed to evaluate the linear association between projected temperatures and age‐standardized mortality rates attributable to non‐optimal, high‐, and low‐temperature exposure under SSP2‐4.5 and SSP5‐8.5 scenarios, with corresponding scatter plots presented. Subsequently, RCS regression was conducted separately for each scenario to characterize the nonlinear exposure–response relationship using the formula: ASR ∼ RCS (temperature, knots = 3) + Year, where knots were automatically positioned at the 10th, 50th, and 90th percentiles of temperature to flexibly capture nonlinear trends while adjusting for long‐term temporal variations. Model performance was assessed using the coefficient of determination (R 2) and root‐mean‐square error, and fitted curves for historical data were generated with key metrics displayed. The validated RCS models were then applied to predict annual ASRs attributable to non‐optimal, high‐, and low‐temperature exposure under both scenarios from 2024 to 2050. Finally, projected trends under the two emission scenarios were combined and plotted to facilitate direct comparison. All statistical analyses and visualizations were performed using R software (version 4.3.3). In preparing this manuscript, artificial intelligence tools Doubao (https://www.doubao.com/chat/) were utilized solely for language‐related tasks, including translating portions of draft text and refining linguistic expression to enhance clarity and readability. AI did not participate in any substantive aspects of the research, such as study design, data collection, statistical analysis, interpretation of results, or formulation of conclusions. All scientific content and intellectual contributions remain the sole responsibility of the authors. Data Availability and Sources All data used in this study are publicly available from the following repositories: Disease burden data were obtained from the Global Burden of Disease 2023 study (GBD, 2023 Disease and Injury and Risk Factor Collaborators, 2025), accessible via the IHME GBD Results database at https://vizhub.healthdata.org/gbd‐results/. The CN05.1 gridded observation data set is described in (J. Wu et al., 2017) and available from the Nansen‐Zhu International Research Centre at https://nzc.iap.ac.cn/content?cid=24&aid=999. CMIP6 climate model data are documented in (Eyring et al., 2016) and available from the Earth System Grid Federation archive at https://aims2.llnl.gov/search/cmip6/. The China Extreme Temperature Indices Gridded Data set (SimmEX, 1961–2020) used for sensitivity analysis is available in (Chen et al., 2024). Analyses and figure generation were performed using custom R scripts (R Core Team, 2024). The analysis code used in this study is archived with a permanent DOI in (He, 2026), with the development version hosted at GitHub (https://github.com/Yujun‐He666/GBDR_analysis_code). The GBD database data used in this study were obtained and provided by Xuequan Wang, co‐author and member of the GBD Collaborating Network. The CN05.1, CMIP6 and SimmEX data sets were obtained and provided by Dr. Yuan He, co‐author from China University of Geosciences. Results Correlation Analysis Between Climate Warming and All‐Causes Mortality Burden in China Spearman's correlation analysis demonstrated a significant correlation between the annual average temperature from CN05.1 and the ASRs of mortality attributable to various causes from non‐optimal temperature exposure in China during 1990–2022. Distinctly different correlation patterns were observed across various causes of death and temperature‐related risk factors, which are detailed as follows. Specifically, all‐causes mortality, as well as mortality from cardiomyopathy and myocarditis, chronic kidney disease, chronic obstructive pulmonary disease, diabetes mellitus, hypertensive heart disease, ischemic heart disease, lower respiratory infections, and stroke, was significantly negatively correlated with the annual average temperature. This indicates that an increase in temperature was associated with a decrease in ASRs for these causes. Among these causes, low‐temperature exposure showed no significant correlation with mortality from chronic kidney disease, chronic obstructive pulmonary disease, diabetes mellitus, and hypertensive heart disease (P = 0.740, 0.655, 0.063, and 0.096, respectively; all P > 0.05). In contrast, low‐temperature exposure was highly significantly positively correlated with mortality from ischemic heart disease (r = 0.509, P = 0.003). Additionally, non‐optimal and high‐temperature exposure were highly significantly negatively correlated with mortality from all the aforementioned causes (all P < 0.001). Low‐temperature exposure also exhibited a significant or highly significant negative correlation with all‐causes mortality, as well as mortality from cardiomyopathy and myocarditis, lower respiratory infections, and stroke. In contrast, mortality from nine categories of causes (including animal contact, drowning, and exposure to forces of nature) was significantly positively correlated with the annual average temperature, meaning that an increase in temperature was associated with an elevation in ASRs. For these nine causes, non‐optimal and high‐temperature exposure were significantly positively correlated with mortality from exposure to forces of nature (P = 0.015 and 0.016, respectively), while non‐optimal and high‐temperature exposure were highly significantly positively correlated with mortality from the remaining causes (all P < 0.001). Furthermore, low‐temperature exposure showed a significant or highly significant negative correlation with mortality from all these nine causes of death. Overall, the correlation between the annual average temperature and mortality from the vast majority of causes reached a significant level (P < 0.05), with approximately 80% showing a highly significant correlation (P ≤ 0.001). Correlations with high‐temperature exposure were the most stable, whereas those with low‐temperature exposure exhibited considerable variability. The absolute values of the correlation coefficients were mostly in the range of 0.35–0.70. Road injuries, self‐harm, and drowning had the strongest positive correlations with temperature, while cardiomyopathy and myocarditis had the strongest negative correlation with high‐temperature exposure (Table S2 in Supporting Information S1, Figure 1, and Figures S1–S17 in Supporting Information S1). Correlation between the annual average temperature from CN05.1 and the age‐standardized rate of all‐causes mortality attributable to non‐optimal temperature exposure in China from 1990 to 2022. (a): non‐optimal temperature exposure; (b): high‐temperature exposure; (c): low‐temperature exposure. The Spearman correlation analysis between annual extreme hot days and extreme cold days and the age‐standardized mortality rates attributable to non‐optimal temperature exposure in China from 1990 to 2020 revealed distinct patterns. The number of cold days was significantly positively correlated with the attributable ASRs for most causes of death, with the strongest associations observed for all‐causes mortality (r = 0.781, P < 0.001), ischemic heart disease (r = 0.665, P < 0.001), stroke (r = 0.776, P < 0.001), chronic obstructive pulmonary disease (r = 0.775, P < 0.001), and diabetes mellitus (r = 0.772, P < 0.001). In contrast, frost days showed significant negative correlations with injury‐related causes (e.g., drowning, road injuries, self‐harm), with correlation coefficients ranging from approximately −0.78 to −0.82 (P < 0.001). The number of hot days was either negatively or weakly associated with the attributable ASRs for most chronic diseases, except for ischemic heart disease (r = 0.689, P < 0.001) and diabetes mellitus (r = 0.489, P = 0.006), which showed significant positive correlations. Hot days also exhibited significant negative correlations with most injury‐related outcomes (r ≈ −0.47 to −0.60). When both hot and cold days were combined as a single non‐optimal temperature exposure indicator, the correlation coefficients were mostly non‐significant (P > 0.05), suggesting that the opposing effects of hot and cold days may cancel each other out (Table S3 in Supporting Information S1). All‐Causes Mortality Burden Attributable to Non‐Optimal Temperature Exposure in China, 1990–2023 From 1990 to 2023, the number of all‐causes death cases attributable to non‐optimal temperature exposure increased from 379,335.825 cases (95% UI: 326,097.253–454,624.647) to 503,892.550 cases (95% UI: 428,004.927–591,948.999), with a relative increase of 32.835%. The ASR decreased from 58.977 per 100,000 population (95% UI: 51.407–69.659) to 23.174 per 100,000 population (95% UI: 19.407–27.191), with a relative decrease of −60.706%. The overall EAPC was −3.080 (95% CI: −3.235 to −2.925), indicating that the overall population mortality risk decreased significantly after adjusting for the population age structure. In terms of exposure types, low‐temperature exposure remained the dominant risk factor: in 2023, 456,573.341 deaths (95% UI: 393,383.138–516,634.708) were attributable to low‐temperature exposure, which was approximately 9 times the number of deaths caused by high‐temperature exposure (50,637.093 deaths, 95% UI: 21,070.284–98,799.242). In terms of temporal trends, the increase in the number of death cases attributable to high‐temperature exposure (57.406%) was higher than that of low‐temperature exposure (31.139%) during the period from 1990 to 2023. For the decrease in ASR, the magnitude of reduction for low‐temperature exposure (61.947%) was greater than that for high‐temperature exposure (43.081%). Further EAPC analysis showed that the EAPC of ASR for low‐temperature exposure was −3.181 (95% CI: −3.324 to −3.037), while the EAPC of ASR for high‐temperature exposure was −1.812 (95% CI: −2.377 to −1.243). This indicated that the annual reduction in mortality risk from low‐temperature exposure was significantly greater than that from high‐temperature exposure, which confirmed the trend of the gradually increasing health impacts of high‐temperature exposure. In terms of low‐temperature exposure, chronic diseases constituted the primary source of mortality burden: the causes with the highest number of deaths in 2023 included stroke (155,427.279 cases), ischemic heart disease (143,167.160 cases), and chronic obstructive pulmonary disease (111,195.611 cases). From 1990 to 2023, the number of deaths increased for ischemic heart disease (RC: 153.573%) and chronic obstructive pulmonary disease (RC: 33.325%) but decreased for stroke (RC: −7.601%). The ASR of all three causes decreased substantially (stroke: −71.543%; chronic obstructive pulmonary disease: −62.362%; ischemic heart disease: −23.184%), with all EAPCs being negative (stroke: −4.042 [95% CI: −4.200 to −3.884]; ischemic heart disease: −0.565 [95% CI: −0.771 to −0.358]; chronic obstructive pulmonary disease: −3.485 [95% CI: −3.936 to −3.031]). Among these three causes, stroke had the most negative EAPC (−4.042), indicating that its mortality risk decreased the most rapidly. Across all 17 causes, the most negative EAPC was observed for animal contact (−7.220 [95% CI: −7.557 to −6.882]); however, animal contact and other injury‐related causes had negative attributable burdens under low‐temperature exposure, with negative ASR values in both 1990 and 2023, so these EAPCs reflect a narrowing of the protective offset rather than a decline in observed deaths and should be interpreted with caution. In terms of high‐temperature exposure, the mortality burden was relatively concentrated and showed a more notable increase: the causes with the highest number of deaths in 2023 included ischemic heart disease (18,343.048 cases), stroke (12,973.971 cases), and chronic obstructive pulmonary disease (6,940.051 cases). Unlike low‐temperature exposure, deaths from some causes under high‐temperature exposure rose markedly, such as diabetes mellitus (237.728%) and ischemic heart disease (381.337%). Their EAPCs were also the least negative among heat‐related causes and both turned positive—1.422 (95% CI: 0.861–1.986) for ischemic heart disease and 0.553 (95% CI: −0.040–1.150) for diabetes mellitus—with only the former reaching statistical significance. Accordingly, their ASRs were the only two showing a relative increase over the study period (42.929% and 15.616%, respectively), indicating a stagnating or even rising heat‐related mortality risk for these causes. Overall, among these 17 causes, the EAPCs for low‐temperature exposure were generally lower than those for high‐temperature exposure, meaning that the annual reduction in mortality risk of low‐temperature‐associated causes was significantly greater than that of high‐temperature‐associated causes. Furthermore, regardless of high or low temperature exposure, cardio‐cerebrovascular diseases (stroke, ischemic heart disease) and respiratory diseases (chronic obstructive pulmonary disease) were the primary causes of mortality burden, serving as key areas for risk prevention and control. The mortality burden from overall non‐optimal temperature exposure was highly concentrated in cardio‐cerebrovascular and respiratory diseases; the top three causes of death in 2023 were stroke (167,567.219 cases), ischemic heart disease (160,095.313 cases), and chronic obstructive pulmonary disease (117,376.705 cases). Together, these three causes accounted for 88.32% of the total deaths attributable to overall non‐optimal temperature exposure, making them the core areas of mortality burden. Changes in death counts from 1990 to 2023 were heterogeneous: among the eight causes with a positive attributable burden, five increased (most markedly ischemic heart disease and diabetes mellitus) whereas three declined (lower respiratory infections −69.659%, cardiomyopathy and myocarditis −27.453%, and stroke −4.935%); for the remaining nine causes, the burden was negative in both years and its absolute magnitude narrowed over time, reflecting a shrinking protective offset. By contrast, the ASR declined for all eight causes with a positive attributable burden, with relative decreases ranging from 19.188% for ischemic heart disease to 85.525% for lower respiratory infections, and EAPCs ranging from −0.418 (95% CI: −0.624 to −0.212) for ischemic heart disease to −6.649 (95% CI: −6.920 to −6.376) for lower respiratory infections. At the all‐cause level, this decline was driven predominantly by reduced low‐temperature risk, with an EAPC of −3.181 (95% CI: −3.324 to −3.037) for low‐temperature exposure versus −1.812 (95% CI: −2.377 to −1.243) for high‐temperature exposure. Interestingly, the data indicated an association between non‐optimal temperature exposure and a decrease in the mortality rate of various injury‐related causes. Among these, the steepest relative decline​ in ASR was observed for animal contact (EAPC −7.335 [95% CI: −7.661 to −7.008]). The age‐standardized mortality rates of self‐harm and road injuries also decreased, with EAPCs of −5.252 (95% CI: −5.572 to −4.931) and −4.356 (95% CI: −4.873 to −3.836), respectively. Decreases were also observed for interpersonal violence (EAPC −7.299 [95% CI: −7.702 to −6.894]) and other unintentional injuries (EAPC −5.884 [95% CI: −6.473 to −5.291]) (Table S4 in Supporting Information S1). Figure 2 and Figures S18–S19 in Supporting Information S1 show the comparison of heat maps of all‐causes mortality in China attributable to non‐optimal temperature exposure from various specific causes in 1990 and 2023. Figure 3 and Figures S20–S36 in Supporting Information S1 show the temporal trends in the numbers and ASRs of all‐causes deaths in China attributable to non‐optimal temperature exposure from various specific causes during 1990–2023. Heatmap comparison of age‐standardized rates for various causes of death attributable to non‐optimal temperature exposure in China in 1990 (a) and 2023 (b). Annual trends in absolute numbers and age‐standardized rates of all‐causes mortality attributable to non‐optimal temperature exposure (a), high‐temperature exposure (b), and low‐temperature exposure (c) in China from 1990 to 2023. The Contributions of Various Specific Causes to the ASRs of All‐Causes Mortality in China Attributable to Non‐Optimal Temperature Exposure Figure 4 presents the proportions of age‐standardized mortality rates of all‐causes deaths attributable to non‐optimal temperature exposure from various specific causes in China in 2023. Negative impact on chronic diseases (positive correlation): Non‐optimal temperature significantly increased the mortality rates of various chronic diseases. Across all ages, stroke and ischemic heart disease were the most affected conditions, followed by chronic obstructive pulmonary disease. These diseases also ranked the highest in ASRs: stroke, ischemic heart disease, and chronic obstructive pulmonary disease. In addition, the mortality rates of hypertensive heart disease, diabetes mellitus, chronic kidney disease, and lower respiratory infections also showed a significant positive correlation with non‐optimal temperature exposure. Potential protective effect on some injury‐related deaths (negative correlation): Notably, non‐optimal temperature was negatively correlated with injury‐related mortality rates, indicating a potential protective association that requires further analysis in the discussion. Among these, drowning had the strongest negative correlation, followed by self‐harm and road injuries. A negative correlation was also observed for other causes such as interpersonal violence and other unintentional injuries. Composition ratio of deaths from various causes attributable to non‐optimal temperature exposure in China in 2023. Figure S37 in Supporting Information S1 shows the proportions of age‐standardized mortality rates of all‐causes deaths attributable to high‐temperature exposure from various specific causes in China in 2023. Across all ages, ischemic heart disease had the highest attributable mortality rate due to high‐temperature exposure, followed by stroke and chronic obstructive pulmonary disease. Notably, drowning was the only cause where the lower limit of the 95% CI for the attributable mortality rate was greater than zero in all age groups and the age‐standardized group, indicating a statistically significant positive correlation between high‐temperature exposure and an increased risk of drowning death. In addition, cardiomyopathy and myocarditis also showed a significant positive correlation with high‐temperature exposure across all age groups. In contrast, the attributable mortality rates of road injuries and interpersonal violence due to high‐temperature exposure were low, and their 95% CIs crossed zero, suggesting an uncertain or weak association. Figure S38 in Supporting Information S1 displays the proportions of age‐standardized mortality rates of all‐causes deaths attributable to low‐temperature exposure from various specific causes in China in 2023. For both the all ages and ASRs, stroke and ischemic heart disease had the highest attributable mortality rates due to low‐temperature exposure, followed by chronic obstructive pulmonary disease and hypertensive heart disease. The attributable mortality rates of diabetes mellitus and chronic kidney disease were relatively low and comparable. Notably, low‐temperature exposure exhibited a protective association with some injury‐related conditions (e.g., drowning, road injuries, self‐harm), with negative attributable mortality rates observed for these causes. Comparison of All‐Causes Deaths Attributable to Non‐Optimal Temperature Exposure Across Different Age Groups in China in 1990 and 2023 Overall Non‐Optimal Temperature Exposure Number of deaths: In 1990, the peak number of deaths was observed in the 70–74 age group, followed by the 75–79 age group and the 65–69 age group. In 2023, the peak shifted to the 85–89 age group; the 80–84 age group and the 90–94 age group also had significantly higher death counts than other age groups. Notably, compared with 1990, the absolute number of deaths in almost all age groups aged 60 and above increased in 2023, with a marked rise in the advanced elderly population aged over 80 (e.g., the 80–84 age group). In contrast, the death counts in all age groups under 59 were generally lower in 2023, and some younger age groups (e.g., <5 years, 5–9 years) showed negative estimated death counts. These negative values may reflect a net protective effect of temperature exposure in these groups, but they are also mathematically produced by the GBD estimation framework as negative attributable fractions, rather than indicating actual observed reductions in mortality. Mortality rate: In 1990, the mortality rate rose sharply with increasing age, peaking in the age group of 95 years and above; the 90–94 age group and the 85–89 age group also had extremely high mortality rates. The upward trend of mortality rate with age persisted in 2023, with the peak also recorded in the 95 years and above age group, followed by the 90–94 age group (Figure 5). Comparison of trends in absolute numbers and age‐standardized rate of all‐causes mortality attributable to non‐optimal temperature exposure by age group and sex in China, 1990 and 2023. High‐Temperature Exposure For total high‐temperature exposure, both the number of high‐temperature‐related deaths and the corresponding mortality rate rose significantly with age in both 1990 and 2023, showing a distinct J‐shaped curve. The peak of the mortality burden was seen in the oldest age groups. 1990: The peak number of deaths was in the 70–74 age group. The age‐standardized mortality rate peaked in the 95 years and above age group. 2023: The peak death count shifted to the 85–89 age group. The age‐standardized mortality rate still peaked in the 95 years and above age group, yet the mortality rate of most advanced elderly groups decreased compared with 1990. From 1990 to 2023, the mortality rate dropped substantially across almost all age groups, particularly in children and adolescents (e.g., the <5 years age group saw a dramatic decline in mortality rate). Despite the decline in mortality rates, the absolute number of heat‐related deaths in the elderly population aged 65 and above increased markedly (e.g., the 80–84 age group had a significant rise in death count, becoming one of the groups with the highest death counts in 2023). The heat‐related mortality burden, which was relatively dispersed across adult groups in 1990, became concentrated in the elderly population (aged over 75) in 2023, who contributed to the vast majority of the total heat‐related deaths that year (Figure S39 in Supporting Information S1). Low‐Temperature Exposure Number of Deaths: In 1990, the highest number of low‐temperature‐related deaths was recorded in the 70–74 age group, followed by the 75–79 age group and the 65–69 age group. In 2023, the age group with the peak low‐temperature‐related deaths shifted to an advanced age: the 85–89 age group had the highest death count, followed by the 80–84 age group and the 90–94 age group. Notably, the number of cold‐related deaths in the 95 years and above age group rose sharply from 1990 to 2023. In contrast, the estimated cold‐related death counts for all age groups under 50 were mostly negative in both 1990 and 2023 (e.g., 5–9 years, 10–14 years), These negative values suggest that low‐temperature exposure may have a potential net protective effect in younger populations. Meanwhile, such values are also generated by the GBD model as negative attributable fractions during burden calculation, rather than representing actual observed deaths. The <5 years age group had a positive death count in 1990 but turned negative in 2023, representing a substantial improvement. Mortality rate: The mortality rate rose sharply with increasing age. In 1990, the highest mortality rate was in the 90–94 age group, followed by the 85–89 age group and the 95 years and above age group. In 2023, the mortality rate peaked in the 95 years and above age group, with the 90–94 age group ranking second. A key finding was that the mortality rate for low‐temperature‐related deaths decreased significantly in all age groups in 2023 compared with 1990 (e.g., the mortality rate of the peak age group and the 70–74 age group both saw notable declines). This reflects that climate warming has alleviated the mortality burden attributable to low‐temperature exposure (Figure S40 in Supporting Information S1). Joinpoint Analysis Results of the All‐Causes Mortality Burden Attributable to Non‐Optimal Temperature Exposure in China, 1990–2023 Overall Non‐Optimal Temperature Exposure Over the entire study period of 1990–2023, the AAPC in all‐causes mortality attributable to non‐optimal temperature exposure in China was −2.753 (95% CI: −3.514 to −1.993), with P < 0.001. Joinpoint analysis identified four distinct phases with varying rates of change: Phase 1 (1990–2003): The APC was −2.428% (95% CI: −2.927% to −1.926%), P < 0.001. Mortality exhibited a significant and steady decline in this phase; Phase 2 (2003–2006): The APC was −5.130% (95% CI: −10.178%–0.201%), P = 0.059. The rate of mortality decline accelerated, but the 95% CI included 0, indicating the statistical significance of this trend was marginal; Phase 3 (2006–2020): The APC was −3.069% (95% CI: −3.565% to −2.571%), P < 0.001. Mortality showed a significant and rapid decline once again; Phase 4 (2020–2023): The APC was −0.313% (95% CI: −5.617%–5.289%), P = 0.910. The decline in mortality slowed markedly in this phase, with the change being statistically non‐significant, meaning mortality entered a relatively stable plateau (Figure S41 and Table S5 in Supporting Information S1). High‐Temperature Exposure Overall, all‐causes mortality attributable to high‐temperature exposure showed a non‐significant declining trend over the study period, with an AAPC of −0.641% (95% CI: −5.104%–3.822%) and P = 0.778, confirming the decline was not statistically significant. Stratified phase analysis revealed heterogeneity in mortality changes across different periods: Phase 1 (1990–2006): The APC was −0.806% (95% CI: −2.345%–0.757%), P = 0.310. Mortality experienced a slight but non‐significant decline during this period; Phase 2 (2006–2014): The APC was −5.040% (95% CI: −9.557% to −0.298%), P = 0.038. This was the only phase with a statistically significant decline in mortality; Phase 3 (2014–2016): The APC was 18.478% (95% CI: −24.201%–85.188%), P = 0.457. The estimated mortality rate rose sharply in this phase, yet the CI was extremely wide and included zero, indicating this upward trend was highly unstable and statistically insignificant; Phase 4 (2016–2023): The APC was −0.699% (95% CI: −6.452%–5.409%), P = 0.818. Mortality returned to a slight but non‐significant declining trend (Figure S42 and Table S5 in Supporting Information S1). Low‐Temperature Exposure From 1990 to 2023, the AAPC in all‐causes mortality attributable to low‐temperature exposure for the general population in China was −2.972% (95% CI: −3.732% to −2.212%), with P < 0.001. Stratified Phase Trends: Phase 1 (1990–2000): The APC was −2.688% (95% CI: −3.400% to −1.972%), P < 0.001. Mortality presented a stable and significant declining trend; Phase 2 (2000–2017): The APC was −3.362% (95% CI: −3.730% to −2.993%), P < 0.001. The rate of mortality decline accelerated further; Phase 3 (2017–2020): The APC was −5.680% (95% CI: −10.677% to −0.404%), P = 0.035. Mortality continued to decline significantly with the largest average annual drop; however, the relatively wide CI suggested increased volatility or uncertainty in the rate of decline for this phase; Phase 4 (2020–2023): The APC was 1.005% (95% CI: −4.346%–6.655%), P = 0.719. The trend reversed, with a non‐significant average annual increase in mortality rate (Figure S43 and Table S5 in Supporting Information S1). Age‐Period‐Cohort Analysis Results of the All‐Causes Mortality Burden Attributable to Non‐Optimal Temperature Exposure in China From 1990 to 2023 Because the APC model requires positive ASR values, the presence of negative estimates for overall non‐optimal and low‐temperature exposure initially made the analysis impossible. To verify the influence of these negative values, we conducted a sensitivity analysis in which all negative ASR values were set to zero. Under this transformation, the APC model converged and yielded interpretable results. We report the findings from this sensitivity analysis for non‐optimal and low‐temperature exposure in the Supporting Information S1, while the results for high‐temperature exposure (based on the original positive data) are described below. Overall Non‐Optimal Temperature Exposure Age effect: The age‐specific mortality rate exhibited a J‐shaped curve. Mortality was extremely low in children and adolescents aged 0–29 years (<0.001 per 100,000). A slow increase began in the 30–39 age group (<1 per 100,000), followed by an exponential rise from age 40 onwards. The mortality rate increased from approximately 3.2 per 100,000 in the 40–44 age group to 349 per 100,000 in those aged 95 years and older, confirming age as the dominant determinant of mortality risk (Figure S44B in Supporting Information S1). Period effect: Using 2004–2008 as the reference (RR = 1.00), the period‐specific relative risks showed a clear downward trend. From 1989 to 2003, all RRs were above 1.0, indicating higher mortality risk compared with the reference period, although the values decreased over time. From 2009 to 2023, RRs fell below 1.0, declining from 0.92 to 0.72, suggesting sustained improvements in population health during the past 15 years (Figure S44C in Supporting Information S1). Cohort effect: With the 1957–1961 birth cohort as the reference (RR = 1.00), earlier cohorts (1892–1956) had substantially elevated risks, with the highest RR of 6.24 observed in the 1892–1896 cohort. Risk decreased progressively across subsequent cohorts. Later cohorts (born from 1962 onwards) showed RRs below 1.0, falling to near‐zero for cohorts born after 2000, reflecting the long‐term health gains associated with socioeconomic development and improved healthcare (Figure S44D in Supporting Information S1). Net drift and local drifts: The net drift was −2.74%, indicating an average annual decline of 2.74% in all‐causes mortality over the study period after adjusting for age and cohort effects. Local drifts varied by age. For children aged 0–4 years, the trend was not statistically significant (CI crossing zero). For ages 5–14 years, mortality declined significantly by 6%–12% per year. For ages 15–29 years, the decline slowed and became non‐significant (CI included zero). For ages 30–39 years, the trend was initially upward but later stabilized. For all age groups aged 40 years and older, mortality showed significant and persistent annual declines, ranging from −2.1% (40–44 years) to −1.3% (95+ years) (Figure S44A in Supporting Information S1). High Temperature Exposure Age Effect: The all‐causes mortality rate exhibited a typical J‐shaped age distribution pattern. The mortality rate was relatively high in infancy, and then dropped rapidly to the lowest level in childhood and adolescence. Starting from early adulthood, the mortality rate began to rise slowly with increasing age and entered a period of accelerated increase in middle and old age, eventually peaking in the oldest age group (Figure 6b). Results of the Age‐Period‐Cohort analysis for all‐causes mortality attributable to high temperature exposure. (a): Net drifts and Local drifts; (b): Age effect; (c): Period effect; (d): Cohort effect. Period Effect: The period effect reflects the shared impact of specific historical periods on all age groups. With a mid‐2000s period as the reference, the mortality risk was relatively high in the late 1980s to early 1990s. The risk then experienced slight fluctuations near the baseline level before starting to show a steady downward trend from the late 2000s onwards (Figure 6c). Cohort Effect: With a specific mid‐20th century birth cohort as the reference, the mortality risk was significantly higher in earlier birth cohorts. The risk showed a downward trend starting from the early 20th century birth cohorts, and for populations born after the reference cohort, the mortality risk presented a sustained and sharp decline. In particular, the relative risk of cohorts born in the late 20th century and later dropped to a low level (Figure 6d). Local Drift and Net Drift: Net Drift: After adjusting for age and cohort effects, the all‐causes mortality rate attributable to high‐temperature exposure in China decreased significantly annually. Local Drift: The rate of mortality decline varied across different age groups. The young and middle‐aged groups had the fastest decline, and the decline rate slowed gradually with increasing age. Notably, the local drift values turned positive in the oldest‐old population. Although the confidence intervals for these values were wide and included 0, this finding suggests that the mortality decline trend in the oldest‐old population may have plateaued, or even stagnated or slightly increased (Figure 6a). Low Temperature Exposure Age effect: The age‐specific mortality rate followed a J‐shaped pattern. Young populations remained at a low‑risk level. From age 30 years onward, mortality increased steadily, accelerating after age 40 and peaking at 302.79 per 100,000 in the ≥95 age group, demonstrating that aging is the primary driver of cold‐related mortality risk (Figure S45B in Supporting Information S1). Period effect: Using 2004–2008 as the reference (RR = 1.00), period‐specific RRs showed only minor fluctuations. Slightly higher risks were seen in 1989–1993 (RR = 1.26) and 1994–1998 (RR = 1.22), while a marginally lower risk was observed in 2019–2023 (RR = 0.96). However, the 95% confidence intervals of all period RRs included 1.00, indicating that the period effect was not statistically significant for low‐temperature attributable mortality (Figure S45C in Supporting Information S1). Cohort effect: With the 1957–1961 birth cohort as the reference (RR = 1.00), earlier cohorts (e.g., 1892–1896) had a markedly higher risk (RR = 6.81). Risk then declined progressively across subsequent cohorts. Cohorts born from 1962 onwards showed RRs below 1.00. However, from the 1977–1981 cohort onward, RRs began to increase again, with an extreme but highly imprecise peak in the 1987–1991 cohort (RR = 32.33, very wide CI), followed by a sharp decline in later cohorts. This instability in recent birth cohorts likely reflects data sparsity or estimation uncertainty rather than a genuine epidemiological reversal (Figure S45D in Supporting Information S1). Net drift and local drifts: The net drift was −0.95% (95% CI: −9.97%–8.97%), indicating a slight but non‐significant average annual decline in low‐temperature attributable mortality over the study period. Local drifts showed a striking cross‐over pattern. For individuals under 30 years of age, local drifts were positive, with annual increases exceeding 10% in the 20–29 age group. In contrast, for individuals aged 40 years and older, local drifts were consistently negative, with the largest annual decline (−4.32%) observed in the 60–64 age group. This pattern suggests that the reduction in cold‐related mortality has been concentrated in middle‐aged and older adults (Figure S45A in Supporting Information S1). Predictions of the All‐Causes Mortality Burden in China From 2024 to 2050 Based on Future Temperatures Under Two CMIP6 Emission Scenarios Projections for Non‐Optimal Temperature Exposure For all‐causes mortality, the projected ASRs attributable to non‐optimal temperature were 3.166 (SSP2‐4.5) and 4.984 (SSP5‐8.5), with substantially higher risks under the high‐emission scenario. Chronic obstructive pulmonary disease and diabetes mellitus were positive under both scenarios and higher in SSP5‐8.5. Chronic kidney disease, drowning, road injuries, other transport injuries, and exposure to mechanical forces were also positive in both scenarios but slightly lower in SSP5‐8.5. Hypertensive heart disease, cardiomyopathy and myocarditis, and exposure to forces of nature remained negative in both scenarios. Strong scenario‐dependent reversals were found in ischemic heart disease, stroke, lower respiratory infections, self‐harm, interpersonal violence, other unintentional injuries, and animal contact. Projections for High Temperature Exposure For all‐causes mortality, the projected ASRs attributable to high‐temperature exposure were 3.540 (SSP2‐4.5) and 3.399 (SSP5‐8.5), both remaining positive and indicating persistent heat‐related risks. Disease‐specific projections showed that ischemic heart disease, stroke, chronic obstructive pulmonary disease, diabetes mellitus, and hypertensive heart disease had consistently positive ASRs under both scenarios. Most injury‐related outcomes showed near‐zero or weakly positive values. Projections for Low Temperature Exposure For all‐causes mortality, the projected ASR attributable to low‐temperature exposure was −0.128 (SSP2‐4.5), suggesting a mild protective effect, but became 1.767 (SSP5‐8.5), indicating an increased cold‐related risk under high emissions. Stroke, lower respiratory infections, hypertensive heart disease, cardiomyopathy and myocarditis, and exposure to forces of nature showed consistently negative values under both scenarios. By contrast, chronic obstructive pulmonary disease, diabetes mellitus, drowning, road injuries, other transport injuries, exposure to mechanical force, and chronic kidney disease were positive under both scenarios. Scenario‐dependent reversals were observed in ischemic heart disease, self‐harm, interpersonal violence, other unintentional injuries, and animal contact. Overall, the SSP5‐8.5 high‐emission scenario was associated with greater non‐optimal temperature‐related mortality burdens for several major chronic diseases (e.g., all causes, chronic obstructive pulmonary disease, diabetes mellitus, and stroke). Scenario‐dependent risk reversals were common across multiple causes, highlighting the highly nonlinear and heterogeneous health impacts of future climate warming (Figure 7, Figures S46–S98, Table S6 in Supporting Information S1). Association between the age‐standardized rate of all‐causes mortality attributable to non‐optimal temperature exposure in China under two CMIP6 emission scenarios, SSP2‐4.5 (blue) and SSP5‐8.5 (orange), and predictions for 1990–2023 and the future (2024–2050). (a): Pearson correlation analysis. (b): comparison of historical trajectories and future trends under the two scenarios. Discussions and Conclusions Characteristics of the Databases Used in This Study This study conducted an integrated analysis of the CN05.1 gridded observational data set, CMIP6 climate data, and epidemiological data from the GBD database to investigate the impact of non‐optimal temperatures on all‐causes mortality in China under the context of global warming from 1990 to 2023. The CN05.1 gridded observational data set is a high‐resolution meteorological observational data set developed by the Climate Change Research Center, Institute of Atmospheric Physics, Chinese Academy of Sciences, and is mainly applied in China's climate research, hydrological simulation, environmental monitoring and other fields. Specifically, this data set is constructed based on observational data from more than 2,400 national meteorological stations in China, including national meteorological observation stations, national reference climate stations and national basic meteorological stations. It is generated by the anomaly method, which involves superimposing the interpolated climate field and anomaly field respectively. First released in 2013, the data set has been continuously updated since then; it currently covers the period from 1 January 1961 to 31 December 2022 and remains under dynamic expansion (J. Wu et al., 2017). The SimmEX provides long‐term series of extreme temperature indicators over mainland China from 1961 to 2020. This data set includes internationally recognized extreme temperature indices such as frost days for extreme cold exposure and summer days for extreme heat exposure, which were used in the sensitivity analysis to validate the robustness of the correlation results (Chen et al., 2024). CMIP6 is a global climate model intercomparison project organized by the World Climate Research Program, which brings together climate models from numerous research institutions worldwide. Through multi‐scenario and multi‐variable simulations, it integrates socioeconomic factors with greenhouse gas emission pathways, thereby providing a scientific basis for climate change research and policy formulation. Its core scenarios include SSP5‐8.5 and SSP2‐4.5: SSP5‐8.5 is a high‐emission, fossil fuel‐driven development scenario that combines SSP5 (a socioeconomic pathway dominated by traditional fossil fuels) with RCP8.5. Its radiative forcing will reach 8.5 W/m2 by 2100, making it the scenario with the highest radiative forcing in CMIP6, which corresponds to a development model characterized by rapid global economic growth, the absence of systematic emission reduction policies, and a persistently high proportion of fossil energy in the energy mix. In contrast, SSP2‐4.5 is a moderate‐emission, moderate‐development pathway scenario formed by the combination of SSP2 (a moderate socioeconomic pathway following historical trends) and RCP4.5. Its radiative forcing will stabilize at 4.5 W/m2 by 2100, reflecting a development state where policies seek a balance between economic growth and environmental management, the energy structure is gradually diversified, and no breakthrough emission reduction measures are implemented (Eyring et al., 2016). The GBD is a comprehensive global health research program aimed at quantifying the health impacts of diseases, injuries and risk factors at the global, regional and national levels. GBD adopts a holistic approach to estimate disease burden, which includes integrating multi‐source data, conducting disease modeling with the DisMod‐MR model, and quantifying uncertainties using Bayesian statistical frameworks and Monte Carlo methods. These methods together form a systematic analytical framework that can generate comparable and consistent estimates of disease burden (“Global burden of 87 risk factors in 204 countries and territories 1990–2019,” 2020; Liang et al., 2024). Discussion on the Correlation Between Climate Warming and the All‐Causes Mortality Burden in China In this study, we used annual average temperature as a proxy for climate warming in our correlation analysis with GBD‐attributed non‐optimal temperature mortality. It should be noted that non‐optimal temperature exposure in the GBD framework is defined based on daily temperature deviations from the TMREL, which more directly captures extreme heat and cold events. However, several considerations support the use of annual average temperature as a proxy. First, temporal consistency is an important consideration. GBD outputs attributable mortality data as annual age‐standardized mortality rates, not daily or monthly data. Consequently, regardless of which temperature metric is chosen, only annual aggregates are available for correlation analysis. Linking annual climate indicators with annual health indicators is therefore consistent at the data aggregation level. Second, empirical evidence shows that annual average temperature is significantly correlated with extreme temperature indices in China (Dai et al., 2016). More importantly, the official GBD methodology itself uses annual mean temperature to define temperature zones, within which the effects of daily temperature exposure are modeled (Burkart et al., 2021; Liu et al., 2022). Furthermore, the non‐linear J‐shaped temperature–mortality curve implies that changes in mean temperature affect the proportion of days falling into the extreme tails of the distribution (Gasparrini et al., 2015). Based on these considerations, annual average temperature serves as a reasonable proxy for assessing the broad association between long‐term climate warming and temperature‐attributable mortality. The consistently strong correlations observed in this study (most |r| > 0.6) support the validity of this approach. We acknowledge that this metric cannot fully capture details such as the frequency, intensity, or duration of extreme temperature events; these limitations are discussed in Section 5.9. To further validate the robustness of our conclusions, we conducted sensitivity analyses using extreme temperature indicators. The correlation analysis using extreme temperature days revealed that cold days were significantly positively associated with attributable mortality risks for most chronic diseases, particularly cardio‐cerebrovascular diseases, chronic obstructive pulmonary disease, and diabetes mellitus, which is consistent with the conventional understanding that cold exposure increases chronic disease mortality by raising blood pressure, blood viscosity, and the risk of respiratory infections (Guoqing Zhang et al., 2023). In contrast, cold days showed negative correlations with injury‐related causes such as drowning and self‐harm, possibly because cold weather reduces outdoor activities and the occurrence of related accidents. Hot days exhibited a weak overall association with chronic diseases but remained significantly positively correlated with ischemic heart disease and diabetes mellitus, suggesting that sustained high temperatures may induce cardiovascular stress and metabolic disorders (Meo et al., 2025). Notably, when hot and cold days were combined into a single non‐optimal temperature exposure indicator, most correlations became non‐significant, reflecting the opposing effects of the two temperature extremes that cancel each other out. This finding also indicates that analyzing extreme cold and hot days separately reveals the true health impacts more effectively than using a composite indicator. Climate warming was significantly correlated with the mortality burden of most causes of death, with distinct correlation patterns. Mortality from most chronic diseases and all causes was negatively correlated with air temperature, primarily because low temperature constitutes a key risk factor for such chronic diseases, and climate warming can mitigate the health hazards induced by low temperature (Chang et al., 2022). Furthermore, the impact of low‐temperature exposure on different chronic diseases is heterogeneous, which stems from the divergent pathophysiological mechanisms underlying the effects of low temperature on various diseases. In contrast, mortality from nine categories of causes associated with accidents and environmental exposure was positively correlated with air temperature: the rise in air temperature prolongs the activity period of disease vectors, increases the incidence of outdoor activities and extreme weather events, thereby elevating the corresponding mortality risk (Parums, 2024; Pavia et al., 2025). The correlation between high‐temperature exposure and mortality rate was more stable, while that of low‐temperature exposure showed greater variability. This is due to the fact that high temperature has a more direct association with changes in average air temperature, whereas low temperature is susceptible to such factors as regional climate and human adaptive behaviors. The correlation coefficients for the above associations were mostly moderate to strong. Core Findings on the All‐Causes Mortality Burden Attributable to Non‐Optimal Temperature Exposure in China (1990–2023) This study found that the total number of all‐causes death cases attributable to non‐optimal temperature exposure in China increased from 1990 to 2023, while the ASR declined. This contradictory trend between the absolute death count and ASRs can be explained by two key factors: the reduction in low‐temperature exposure against the backdrop of global warming, and changes in China's population structure and public health standards. On the one hand, China's total population rose rapidly from 1990 to 2023, accompanied by an accelerated pace of population aging (Long et al., 2025). The elderly population is more sensitive to temperature changes, which may have led to the rise in the absolute number of temperature‐related deaths. On the other hand, the significant decline in ASRs and the negative EAPC indicate that, after adjusting for population age structure, the overall mortality risk of population exposure to non‐optimal temperatures has been effectively controlled. This is closely associated with the improvement of medical and health services, the popularization of health education, and the strengthening of temperature‐related health interventions (e.g., heating support in cold regions, heatstroke prevention and cooling measures in hot regions) in China over the past three decades (B. Yin et al., 2024; Z. Yin et al., 2024). Nevertheless, the continuous growth in the absolute number of deaths also reminds us that the pressure of temperature‐related mortality burden cannot be ignored against the backdrop of global warming. In terms of exposure types, low‐temperature exposure remained the dominant risk factor for temperature‐related deaths in China in 2023, with the number of deaths approximately nine times that of high‐temperature exposure‐related deaths. This is consistent with China's geographical location and climatic characteristics: most parts of China are located in temperate and subtropical zones with long cold seasons and significant low‐temperature impacts, particularly in northern regions (Zheng et al., 2021). A more noteworthy trend, however, is that the rate of increase in the number of high‐temperature exposure‐related deaths during the study period was significantly higher than that of low‐temperature exposure, while the decline in low‐temperature exposure ASRs was far greater than that of high‐temperature exposure. In addition, the EAPC of low‐temperature exposure ASRs was lower than that of high‐temperature exposure, indicating that the annual rate of decline in mortality risk attributable to low‐temperature exposure was significantly greater than that of high‐temperature exposure. This trend is closely linked to global warming: China's observed annual temperature increase was 1.54 degrees Celsius between 1901 and 2018, which was faster than the global average (H. Yin & Sun, 2023). This has led to an increase in the frequency, intensity and duration of extreme high‐temperature events (e.g., heatwaves). At the same time, the impact of global warming on low‐temperature events is relatively complex—although the overall temperature has risen, extreme cold events may still occur in some regions, yet the long‐term trend in the intensity and frequency of low temperatures may have weakened (Tang et al., 2013). Combined with the continuous improvement of heating conditions in China, the mortality risk related to low temperatures has decreased more significantly. In contrast, the mortality risk related to high temperatures has declined relatively slowly, and some causes have even shown an upward trend (e.g., diabetes mellitus and ischemic heart disease). This reflects that the health impacts of extreme high temperatures induced by global warming are gradually emerging and intensifying, and that China's adaptive measures for high‐temperature risks are still inadequate compared with those for low‐temperature risks. Analysis of the disease spectrum revealed that cardio‐cerebrovascular diseases (stroke, ischemic heart disease) and respiratory diseases (chronic obstructive pulmonary disease) were the main sources of the mortality burden attributable to non‐optimal temperature exposure. Temperature changes can directly affect the human cardio‐cerebrovascular and respiratory systems: low temperatures can cause vasoconstriction and elevated blood pressure, triggering acute episodes of stroke and ischemic heart disease, while exacerbating respiratory infections and chronic obstructive pulmonary disease (Barrak et al., 2024; Daiwei et al., 2025; Jia et al., 2023); high temperatures can lead to increased blood viscosity and water and electrolyte imbalance, which can also induce acute cardio‐cerebrovascular and respiratory events (Chang et al., 2022; L. Jingwen et al., 2022). Against the backdrop of global warming, the interaction between temperature changes and chronic diseases has become more pronounced: with the intensification of population aging, the prevalence of chronic diseases in China has continued to rise, and the number of chronic disease patients who are sensitive to temperature changes has increased accordingly, which has further amplified the mortality burden of non‐optimal temperature exposure. Notably, under high‐temperature exposure, both the number of deaths and ASRs related to ischemic heart disease increased significantly. An interesting finding of this study is that the non‐optimal temperature attributable mortality rates for multiple injury‑related outcomes (e.g., drowning, self‑harm, road injuries) declined over the study period. This pattern may be largely attributable to substantial improvements in public safety governance and health‑related interventions across China over the past three decades. For instance, widespread drowning‑prevention education, stricter road‑traffic safety regulation, and expanded mental‑health services have collectively lowered the total background mortality from these injury causes. Such secular declines may partly offset the detrimental health impacts of non‑optimal temperature exposure on injury‑related deaths (Zi et al., 2023). Notably, many injury categories presented negative attributable mortality burdens under non‑optimal temperature exposure, which warrants cautious interpretation of their EAPC estimates. While nearly all injury‑related causes showed decreasing temperature‑attributable mortality, chronic conditions such as ischemic heart disease exhibited positive EAPCs under high‑temperature exposure, signalling elevated mortality risks linked to heat; accordingly, targeted public‑health interventions should be prioritized for heat‑sensitive chronic diseases. Discussion on the Causes of Age Characteristics of All‐Causes Mortality Attributable to Non‐Optimal Temperature Exposure in China, 1990–2023 The characteristics of deaths attributable to non‐optimal temperature exposure across different age groups in China from 1990 to 2023, namely the shift toward advanced age, the decline in population‐wide mortality rates and disparities in the impacts of high‐ and low‐temperature exposure, are fundamentally the combined outcomes of population aging, climate warming and the improvement of the public health system. Population aging has led to a substantial expansion of the oldest‐old population. Coupled with their declined physiological functions, high prevalence of underlying diseases and extremely low tolerance to temperature stress, this constitutes the fundamental cause of the shift in mortality burden to individuals above 80 years of age (Hua et al., 2024). Advances in the public health system, including the standardized management of chronic diseases, the improvement of health care services for the elderly and children, and the strengthening of early warning and intervention for extreme weather events, have effectively reduced population‐wide temperature‐related mortality rates. In particular, they have driven a sharp drop in high‐temperature mortality among younger age groups and achieved a fundamental improvement in low‐temperature‐related deaths among infants and young children. The negative attributable mortality estimates observed in some younger age groups warrant careful methodological interpretation. In the GBD comparative risk assessment framework, the population attributable fraction (PAF) can become negative when the observed cause‐specific mortality rate in a particular age‐sex‐year stratum is lower than the expected mortality at the TMREL (GBD, 2023 Disease and Injury and Risk Factor Collaborators, 2025). A negative PAF indicates that the current exposure level—in this case, to non‐optimal temperature—is associated with a lower disease burden than the TMREL counterfactual, rather than suggesting a direct “protective” biological effect of temperature itself. From a biological perspective, this pattern is consistent with the well‐documented finding that elderly populations are far more vulnerable to both heat‐ and cold‐related mortality than younger age groups (Zeng et al., 2022). Epidemiological evidence from China confirms that the mortality effect of non‐optimal temperature on years of life lost rates is substantially smaller for younger populations than for older ones (Zeng et al., 2022). Although the precise physiological and socioeconomic factors that contribute to this lower vulnerability in young individuals are not fully delineated in the GBD framework, the net statistical offset remains a plausible explanation for the negative estimates observed in younger age strata. Climate warming has significantly alleviated the intensity and duration of low‐temperature exposure, acting as the key factor behind the sharp drop in low‐temperature‐related mortality rates. Meanwhile, the concentration of high‐temperature mortality burden in the oldest‐old population is due to the comprehensive heat protection measures for young and middle‐aged groups, whereas the oldest‐old population have limited protective effects against high temperatures as a result of their severe decline in physiological functions. Driving Factors and Causes of Disparities in the Evolution of Mortality Rates Attributable to Non‐optimal Temperature Exposure From 1990 to 2023, the all‐causes mortality rates attributable to the three categories of temperature exposure in China exhibited distinct evolutionary trends, which are fundamentally driven by the differentiated and synergistic effects of factors including the improvement of the public health system, climate warming, population aging and fluctuations in extreme climate events. The mortality rate attributable to overall non‐optimal temperature exposure saw a significant decline in the early stage, driven by the dual positive effects of the improved public health system—such as the implementation of chronic disease management and extreme weather early warning systems—and climate warming that mitigated low‐temperature exposure. It then leveled off after 2020, because the intensified population aging has expanded the scale of the highly vulnerable oldest‐old population (L. Zhang et al., 2024), and the increase in their mortality has offset the aforementioned protective effects. The mortality rate attributable to high‐temperature exposure showed an overall non‐statistically decline, with only a brief and significant decrease in a single phase. This is because the protective effects brought by the implementation of heat protection policies and the popularization of cooling facilities in that phase were continuously offset by the adverse effects of increased frequency and intensity of extreme high temperatures induced by climate warming. (Miao et al., 2022). The fluctuations in estimates during certain periods also stemmed from the synergistic effects of interannual variations in extreme high temperatures and population aging. The mortality rate attributable to low‐temperature exposure recorded the most significant decline among the three categories, with a sustained and accelerated decrease in the early stage. This is mainly due to climate warming, as the dominant factor, which greatly reduced the intensity and frequency of low‐temperature exposure, coupled with the synergistic effects of public health measures such as the improvement of heating facilities and the strengthened prevention and control of chronic diseases in winter. The non‐statistically significant slight increase after 2020 is merely the combined short‐term effects of short‐term fluctuations caused by severe cold waves, temporary disruptions to public health services in the post‐pandemic period and population aging, and has not formed a substantive trend. In addition, the evolution of the three exposure types also shares common characteristics and core differences. Their common ground lies in that all were positively driven by the improvement of the public health system in the early stage and constrained by population aging and fluctuations in extreme climate events in the later stage. The core differences are shaped by the differentiated impacts of climate warming: it exerts a significant mitigating effect on low‐temperature exposure, while bringing about the adverse effect of an increase in extreme events for high‐temperature exposure, which has become the key reason for the divergent trends of mortality attributable to high and low‐temperature exposure. Discussion on Results of Age‐Period‐Cohort Analysis Overall, the Age‐Period‐Cohort results revealed consistent patterns and heterogeneous characteristics of mortality attributable to non‐optimal, high‐, and low‐temperature exposure across age, period, and cohort dimensions. The age effects for all three temperature exposure types exhibited a typical J‐shaped curve. Mortality risk was low in children and young adults, increased slowly after middle age, rose exponentially in the elderly, and peaked in the age group of 95 years and older. This pattern reflects the profound impact of physiological aging: reduced thermoregulatory capacity and a high prevalence of comorbidities make the elderly the most vulnerable population to temperature stress (Azan et al., 2025). Young adults, by contrast, have stronger physiological adaptability and consequently bear a relatively lower mortality risk. Cohort relative risks showed a long‐term declining trend across all temperature indicators. Earlier birth cohorts had significantly higher mortality risks, while later birth cohorts—particularly those born after 1980—experienced sustained and substantial risk reductions. This reflects the fundamental improvements in nutrition, public health, medical services, health literacy, and living conditions in China over recent decades. For low‐temperature‐related mortality, the trend in recent birth cohorts displayed some instability, which is more likely attributable to data sparsity and estimation uncertainty than to a genuine reversal of health risk. Period effects differed markedly across exposure types. For non‐optimal and high‐temperature exposure, period relative risks were highest in the early part of the study period and declined steadily after 2009, indicating continuous improvements in China's heat‐health early warning systems, urban cooling facilities, heat‐protection policies, and healthcare accessibility. For low‐temperature exposure, the period effects were not statistically significant in any phase, suggesting that changes in cold‐related mortality are primarily driven by age structure and cohort improvements rather than by nationwide period‐level interventions. Net drift results showed that for non‐optimal and high‐temperature exposure, the net drifts were significantly negative, indicating a year‐on‐year decline in related mortality after adjusting for age and cohort effects. Local drifts revealed that the most pronounced declines occurred in children and adolescents (especially those aged 5–14 years); the rate of decline gradually slowed with increasing age, and in the oldest‐old population the downward trend plateaued or even reversed slightly. Among the oldest‐old population, local drifts turned positive (although not statistically significant), suggesting that the decline in mortality has entered a plateau or slightly increased due to extreme frailty and the limited effectiveness of conventional protective measures. For low‐temperature exposure, the net drift was not statistically significant, but the local drifts showed a striking bipolar pattern: mortality risk increased in people under 30 years of age, while it continued to decline in those aged 40 years and older. This divergence suggests that the health benefits of climate warming have mainly accrued to the elderly, whereas cold protection and safety interventions for younger populations remain inadequate. In summary, the Age‐Period‐Cohort analysis demonstrates that age, period, and cohort effects jointly shape the long‐term trends of temperature‐related mortality in China. Aging is the central determinant of temperature vulnerability, while socioeconomic development and public health advances have driven long‐term risk reductions. In addressing future climate change, more targeted and intensive interventions are needed for both the oldest‐old and young adults to consolidate and expand health gains. Disparities in Predicted All‐Causes Mortality Burden Attributable to Non‐Optimal Temperature Exposure and Their Driving Factors Under Two CMIP6 Emission Scenarios (2024–2050) The divergent projected mortality burdens attributable to non‐optimal, high‐, and low‐temperature exposure under the SSP2‐4.5 and SSP5‐8.5 scenarios from 2024 to 2050 mainly stemmed from climate warming‐induced changes in average temperature levels, the increasing frequency and intensity of extreme temperature events, as well as the heterogeneous nonlinear effects of temperature exposure on different diseases. Under non‐optimal temperature exposure, the projected all‐causes mortality risk was significantly higher in the high‐emission SSP5‐8.5 scenario than in the moderate‐emission SSP2‐4.5 scenario, indicating that unabated greenhouse gas emissions would further exacerbate the overall temperature‐related health burden. Chronic obstructive pulmonary disease and diabetes mellitus showed consistently elevated risks under both scenarios, with higher values in the SSP5‐8.5 scenario, reflecting the high and sustained vulnerability of these chronic diseases to long‐term climate warming. By contrast, chronic kidney disease, drowning, road injuries, other transport injuries, and exposure to mechanical forces also presented positive projections in both scenarios, suggesting persistent health risks; however, their risk magnitudes were slightly lower under SSP5‐8.5, which may be related to the offsetting effects whereby certain risks induced by extreme heat are counterbalanced by reductions in cold‐related harms under different scenarios. Meanwhile, hypertensive heart disease, cardiomyopathy and myocarditis, and exposure to forces of nature showed consistently negative projections, implying net protective effects under long‐term temperature rise. Notably, ischemic heart disease, stroke, lower respiratory infections, self‐harm, interpersonal violence, other unintentional injuries, and animal contact exhibited obvious scenario‐dependent risk reversals, which further confirmed the nonlinear and complex association between temperature and mortality. For high‐temperature exposure, all‐causes mortality showed persistent positive risks under both scenarios, confirming that heat‐related health hazards will continue to rise with future climate warming. Ischemic heart disease, stroke, chronic obstructive pulmonary disease, diabetes mellitus, and hypertensive heart disease were all positively associated with high‐temperature exposure, suggesting that patients with cardio‐cerebrovascular and respiratory diseases remain the key high‐risk groups for heat damage. Most injury‐related outcomes showed near‐zero or weakly positive effects, indicating that their direct heat‐related mortality risks were relatively limited. For low‐temperature exposure, striking disparities were observed between the two scenarios. The all‐causes mortality projection was slightly negative under SSP2‐4.5, suggesting a mild protective effect brought by reduced cold stress under moderate climate warming, whereas it turned positive under SSP5‐8.5, indicating that intensified climate variability may increase the frequency and intensity of extreme low‐temperature events, thereby re‐elevating cold‐related health risks (S. Wu et al., 2025). Stroke, lower respiratory infections, hypertensive heart disease, cardiomyopathy and myocarditis, and exposure to forces of nature showed sustained negative projections, demonstrating that long‐term warming could continuously reduce the mortality burden attributable to low‐temperature exposure for these diseases. In contrast, chronic obstructive pulmonary disease, diabetes mellitus, drowning, road injuries, other transport injuries, exposure to mechanical force, and chronic kidney disease remained positively correlated with low‐temperature exposure, indicating that these diseases still have long‐term vulnerability to cold stress. Additionally, ischemic heart disease, self‐harm, interpersonal violence, other unintentional injuries, and animal contact showed clear scenario‐dependent reversals, which further verified the unstable and nonlinear impact of low‐temperature exposure under changing climates. Overall, the SSP5‐8.5 high‐emission scenario would lead to a heavier non‐optimal temperature‐related mortality burden for most major chronic diseases, including chronic obstructive pulmonary disease, diabetes mellitus, and stroke. The widespread scenario‐dependent risk reversals across diseases suggest that the health impacts of future climate warming are highly nonlinear, heterogeneous, and sensitive to emission pathways. These findings underline that medium–low emission pathways represented by SSP2‐4.5 can effectively mitigate the overall temperature‐related health risks, while targeted intervention strategies should be developed for high‐risk populations and high‐sensitive diseases to cope with the dual threats of increasing heat stress and residual cold risks under future climate change. Policy Recommendations Based on the evolutionary characteristics of the all‐causes mortality burden attributable to non‐optimal temperature exposure in China from 1990 to 2023, the predicted results from 2024 to 2050, and combined with the laws of temperature‐related health risks under the background of climate warming, the following targeted policy recommendations are proposed: First, establish a precise protection system differentiated by age and disease. Focus on identifying the elderly population aged 80 and above, patients with cardio‐cerebrovascular diseases (stroke, ischemic heart disease) and chronic obstructive pulmonary disease as core susceptible groups. For low‐risk groups such as infants, young children and young and middle‐aged people, carry out health education on water‐related safety in high temperatures and cold protection and warmth preservation in winter through multiple channels to enhance their awareness of active protection and reduce incidental risks. Second, implement differentiated prevention and control strategies for high and low temperatures. In terms of high‐temperature protection, in light of the predicted trend of increasing high‐temperature intensity in the future, expand urban public cooling facilities, strengthen the safety supervision and publicity and education of water‐related activities for children and adolescents in summer, and targetedly reduce the risk of high‐temperature‐related injuries such as drowning. In terms of low‐temperature protection, consolidate the achievements in reducing the mortality burden brought about by climate warming, continue to optimize heating support, and establish emergency patrol and medical support mechanisms for severe cold waves. Third, strengthen scenario adaptation and synergy with source emission reduction. In response to the predicted increase in risks related to exposure to natural forces and extreme low‐temperature events under the SSP5‐8.5 high‐emission scenario, incorporate climate risk assessment into national public facility construction standards, promote the coordinated advancement of emission reduction policies and health protection, support green energy substitution and low‐carbon city construction, reduce the frequency of extreme temperature events from the source, and lay a solid environmental foundation for the prevention and control of temperature‐related health risks (Romanello et al., 2024). Fourth, improve monitoring and early warning. Establish a regular linkage analysis mechanism between CN05.1, CMIP6 climate data, SimmEX extreme climate data and GBD health data, dynamically update the temperature‐mortality exposure‐response relationships for different populations and diseases, and continuously optimize the accuracy of early warning models. Limitations This study also has some limitations. First, our future projections employed RCS models to characterize the nonlinear temperature–mortality relationship, which is more theoretically appropriate than simple linear models. Nonetheless, some uncertainties remain, as real‐world exposure–response relationships may involve more complex threshold or interaction effects that are difficult to fully capture, especially in scenarios where extreme temperature events occur frequently under high‐emission scenarios. Second, this study uses national‐scale temperature grid data and does not consider regional microclimatic characteristics such as the urban heat island effect and topographic differences (e.g., mountainous areas and plains), nor the synergistic effects between air pollutants (e.g., PM2.5) and temperature (Rahman et al., 2022). These local environmental factors may amplify or weaken the health impacts of temperature, leading to biases in the assessment of mortality burden across different regions. Third, the GBD database has some inherent limitations, such as modeling dependencies and uncertainties, as well as methodological limitations (Bhutta, 2025; GBD, 2023 Disease and Injury and Risk Factor Collaborators, 2025). Fourth, negative attributable mortality estimates appeared in certain age–year strata, particularly among younger populations under non‐optimal and low‐temperature exposure. These negative values are mathematically valid outputs of the GBD counterfactual framework—they indicate that current exposure levels are lower than the TMREL—but they do not represent actual negative deaths. Critically, because the Age‐Period‐Cohort log‐linear model requires positive age‐standardized mortality rates, the presence of negative estimates initially prevented us from performing direct Age‐Period‐Cohort analyses for overall non‐optimal and low‐temperature exposure. To overcome this constraint and enable model convergence, we conducted a sensitivity analysis by setting all negative ASR values to zero and reported the Age‐Period‐Cohort results for these two exposure types in the Supporting Information S1. This pragmatic transformation may introduce bias, especially for recent birth cohorts where negative values were more frequent and UIs were wide (e.g., the extreme relative risk of 32.33 for the 1987–1991 cohort under low‐temperature exposure). Therefore, the Age‐Period‐Cohort findings for non‐optimal and low‐temperature exposure should be interpreted as exploratory. Future studies using alternative modeling approaches that can accommodate zero or negative values or employing refined data with fewer negative estimates are needed to validate these results. Last, our correlation analysis used annual average temperature as a proxy for climate warming. This metric cannot fully capture the frequency, intensity, or duration of extreme temperature events, nor intra‐annual temperature variability or seasonal asymmetric warming. Consequently, our correlation results reflect a broad association between long‐term warming and temperature‐attributable mortality, rather than a precise quantification of risk from individual extreme events. Therefore, using annual average temperature may lead to potential information loss regarding short‐term extreme events and temperature fluctuations. Future studies incorporating daily temperature data and multiple extreme temperature indices are needed to refine these estimates. Conclusions This study integrated CN05.1, CMIP6, SimmEX, and GBD 2023 data to explore non‐optimal temperature exposure's impact on China's all‐causes mortality (1990–2023) and project trends to 2050. Non‐optimal temperature‐related deaths and age‐standardized mortality showed opposite trends, with low‐temperature exposure remaining the primary risk while high‐temperature impacts gradually intensified. The elderly and individuals with cardio‐cerebrovascular or respiratory diseases were the most vulnerable groups. Climate warming was significantly correlated with mortality from most causes. Future projections under different emission scenarios revealed divergent risk trends for various diseases. These findings underscore the necessity of targeted temperature risk prevention and control strategies tailored to age, disease, and emission scenarios amid global warming. Conflict of Interest The authors declare no conflicts of interest relevant to this study. Supporting information Supporting Information S1