Health cost impacts of extreme temperature on older adults based on city-level data from 28 provinces in China

Extreme temperature exposure can have a considerable impact on the health of older adults. China, which has entered a deeply aging society, may be obviously threatened by extreme weather. Based on data obtained from the China Health and Retirement Longitudinal Study, we apply a panel fixed effect model to investigate the impact of extreme temperature on medical costs for older adults. The results reveal a U-shaped relationship between temperature and older adults’ medical costs. Heterogeneity analysis indicates that medical costs for older adults in the South and older adults in rural areas are more significantly affected by low temperatures, mainly due to lower per capita heating facilities. Furthermore, the medical costs of older people with lower education levels are also more susceptible to temperature fluctuations. Our simulated prediction indicates that the medical costs of older adults in 2050 will be 2.7 trillion Chinese yuan under the RCP8.5 scenario, but can be reduced by 4.6% and 7.4% following RCP4.5 and RCP2.6 scenarios, respectively. Compared with base period, the medical costs of older adults in western provinces such as Guangxi and Sichuan will more than triple by 2050. Policymakers should prioritize addressing the health needs of these vulnerable groups and less developed regions with less adaptive capacity.


Introduction
Climate change is the greatest health threat of the 21st century, and corresponding global public health damage continues to worsen (Costello 2009, Marina et al 2022, Lohmann and Kontoleon 2023).Specifically, the health impacts of extreme temperature exposure are the most dominant and immediate of all climate change impacts (Watts et al 2017, Yao et al 2023).In the context of advancing climate change, the increasing frequency, intensity, and duration of heatwaves have catastrophic impacts on public health (Perkins et al 2012, IPCC 2022).This is especially true for older adults, as extreme heat affects the thermoregulatory mechanisms of older adults' body, leading to hyperthermia diseases and exacerbating cardiovascular, respiratory, kidney, and psychiatric diseases (World Health Organization 2015, Watts et al 2019).As older adults generally have lower incomes and are more vulnerable to extreme weather events, so climate change may cause them to have a greater financial healthcare burden compared to other age groups.Therefore, there is an urgent need to assess the health cost impact of extreme temperature exposure on older adults to strategically reduce the health risks associated with climate change in older adults, and to tailor timely and effective health adaptation actions for these clinically vulnerable groups.
A majority of previous research on the health effects of extreme temperature exposure on older adults has focused on the effect of extreme temperature exposure on mortality (Chien et al 2016, Zhang et al 2021).In recent years, a small number of estimates of the health costs of extreme temperature exposure have also been developed.However, most of these studies are conducted in developed countries (Cheng et al 2018), and have examined developing countries to a lesser extent.And these studies primarily use the statistical life value (i.e. the additional cost an individual is willing to bear to reduce the risk of death) to examine the health costs of extreme temperature exposure for older adults (Barreca et al 2015).As it does not refer to actual health costs incurred, statistical life value is inappropriate for assessing the related strain on the health system.Estimating the specific healthcare cost impact of temperature fluctuations will allow health authorities to better understand the economic burden associated with the health of older adults and implement adaptive strategies to address this challenge (Cai et al 2022).However, a limited number of studies have used specific healthcare costs to measure it (Wang et al 2021, Wondmagegn et al 2021).And these studies focused on climate impacts in areas with small latitude and longitude spans, and the results are insufficient to support climate change policies for larger regions across different climate zones and urban-rural arrangements.
The age structure of China's population is undergoing profound change, making it a typical case that requires in-depth research on the health costs of climate change.First, the Chinese population is aging rapidly.China has entered an aging society since 2000, and has entered a deep aging society in 2021.By the end of 2022, the number of people aged 60 and above in China reached 280 million, accounting for 19.8% of the total population (National Bureau of Statistics 2022b).According to the projection of World Population Prospects (2022), China's older population is expected to reach 370 million and 510 million, accounting for one-quarter and one-third of the country's total population in 2030 and 2050 (United Nations 2022).Second, in addition to a few developed coastal provinces and cities, most of China's provinces and cities are located in inland plains and hilly areas that are vulnerable to extreme temperatures (Cai et al 2021).Third, China has a massive scale of medical consumption of older adults.Older adults' demand for medical care is significantly greater than that of the nonolder people, and its per capita medical expenditure is about three times that of non-older people (Yang et al 2021).
In the end, China's age-bearing foundation is currently weak, particularly in terms of medical security.Medical resources are generally insufficient and unevenly distributed, with minimal coverage of the medical security system (National Health and Family Planning Commission 2017).For example, East China and Central South China, which are economically developed, have the highest concentration of hospitals, accounting for about 50% of total hospitals in the country in 2021.And the number of hospitals in the northwestern provinces is relatively small, accounting for only 8.6% of the total hospitals (National Bureau of Statistics 2022a).In addition, the average number of doctors per 1000 people in the western region is the lowest among all regions, at 2.88 in 2021, which is lower than the average number of doctors per 1000 people in the eastern region (3.19) (National Bureau of Statistics 2022a).In particular, the number of doctors per 1000 people in western provinces such as Sichuan and Xinjiang is only 2.68 and 2.73, which is significantly lower than the number of doctors per 1000 people in Beijing in the eastern region (5.13) (National Bureau of Statistics 2022a).Moreover, the medical costs burden rate (the medical costs/provincial GDP) varies among provinces.For example, Sichuan Province and Xinjiang Province in the western region have almost twice medical costs burden rate than Jiangsu and Tianjin in the eastern region in 2018 (see figure S5(a) in supplementary materials 5 for more details).Therefore, considering the vulnerability of older adults and high dependence on the healthcare system, it is urgent to assess the health impact of extreme temperature exposure on older adults in China.
The research content of the study is divided into three parts.First, given that existing studies consider only the effects of outpatient or inpatient medical costs, and focus on the areas with small latitude and longitude spans, making it difficult to accurately depict total health economic burden of temperature effect in a whole country, this study analyzes the impact of extreme temperature on the total medical cost of older adults through a panel fixed effect model, which based on the China Health and Retirement Longitudinal Study (CHARLS) data that includes 125 cities in 28 provincial administrative regions in China.Second, considering the lack of heterogeneity analysis of the response of medical costs to temperature changes in previous studies, making it difficult to identify vulnerable groups with temperature effects, we distinguished the heterogeneity of northern and southern regions, urban and rural areas, and different education levels.Third, in view of the lack of prediction for the temperature effect on the medical costs of older adults and the inability to provide direct decision-making support for future medical reforms, this study predicts the future changes of medical expenses of older adults under climate change.

Research methods
This study uses a panel fixed effect model to investigate the impact of extreme temperature on the medical expenses of older adults in China.The panel fixed Y-Y Yu et al effect model can control for differences in unobservable values, eliciting consistent estimates of temperature coefficients (Auffhammer andAroonruengsawat 2011, Zhang et al 2022).To determine the health effects of extreme temperature exposure on older adults, this study combines microdata from the CHARLS with municipal temperature data for empirical analysis.The detailed data sources and processing in the study are in supplementary materials 1.
The medical cost data of older adults used in this study is obtained from CHARLS microdata in 2011, 2013, 2015, and 2018.Total annual medical expenses are selected.Total annual medical expense is the total of bills to medical insurance plus the out-of-pocket expense.It consists of three parts: outpatient medical costs, hospitalization costs, and the separate cost of purchasing medication.And the number of days that the average daily temperature was in each temperature bin in the survey year serves as the core explanatory variable (recorded as Temp).We set the range of each temperature bin to 5 • C, and divide the daily average temperature into 10 groups (<−10 temperature bin as the reference group since medical cost is least affected by temperature at this temperature bin (Hou et al 2022).We also control for other relevant factors in the model, including individual variables (H m,i ) such as age, household registration, and income, as well as relevant climate effects (Rh m,c ), including rainfall, wind speed, humidity, duration of daylight, and atmospheric pressure.Since rainfall is a non-linear, non-stable time series, we refer to Aufhammer & Aroonruengsawat (2011), Hou et al (2022) to control for rainfall using a secondorder polynomial in all regressions.Particularly, we also controlled the annual average pollutant concentration.In addition, temperature changes may be accompanied by changes in economic development and energy consumption.City's GDP per capita and greening rate are added to the model as city-level control variables (φ m,c ).With these variables, we construct a health cost impact model for temperature, presenting the specific formula of the model in equation ( 1) (1) Here, Lnexpense m,c,i represents the logarithm of medical expenses for resident i of city c in year m; Temp m,c,j indicates the number of days that resident i in city c has experienced temperature bin j in year m; γ m represents the year fixed effect, and δ i is the individual fixed effect.β µ and ρ are the regression coefficients for H m,i , Rh m,c and φ m,c , respectively.ε m,c,i represents a random error term.
To project the impact of climate change on healthcare costs, we first calculate the predicted change in the number of days per year in each temperature bin, multiply it by the corresponding impact on medical costs obtained from the regression results to obtain the degree of impact of each temperature bin, then multiply it by the number of older residents and basic medical costs to obtain the medical costs change (△expense) correlated with climate change referencing Agarwal et al (2021).The corresponding formula is presented in equation ( 2) (2) Here, population older represents older adults' population of city c in year m, and expense basic is the medical cost for the base period, compiled from the average per capita medical cost of each city from 2016 to 2020.

Baseline estimates for temperature and medical costs
As shown in figure 1, The relationship between older adults' medical expenses and temperature is roughly expressed as a U-shaped curve, indicating that daily average extreme temperatures increase older adults' medical expenses.For example, as shown in figure 1, an extra day in a temperature bin above 30 • C would lead to a 1.3% increase in annual medical costs for older adults (relative to the 10 • C-15 • C reference temperature bin), which is equivalent to an annual increase of 5.10 million Chinese yuan (CNY) in medical expenses per 100 000 older people (51.05 CNY per capita, according to the constant price in 2018).Li et al (2023) examined the causal effects of extreme temperatures on out-of-pocket health expenditure.It was found that out-of-pocket medical expenditures increased by 0.85% and 2.80% at temperatures of 16 • C-21 • C and above 32 • C (relative to the reference temperature chamber), respectively.This is similar to the results of this study.In addition, temperature bins with daily average temperatures between −5 • C and 20 • C are not significantly associated with medical costs.The detailed regression results and robustness tests of regression results are shown in supplementary materials 3 and 4, respectively7 .

Heterogeneity analysis
Vulnerability to climate change depends on the type of climate hazard as well as demographic characteristics (Lutz and Muttarak 2017).Among the subgroups of older adults, different types of older adults may be vulnerable to differing degrees of climate risk.The characteristics of these different types of older population primarily include Geographical features, household registration, gender, age, education level, etc.Some studies found that the effect of temperature on mortality varies under different regions and household registrations (Wang et al 2018, He et al 2023), which may also have a regional heterogeneous effect on residents' health care costs.In addition, individual characteristics such as gender, age, and education level have also shown an impact on temperature effects (Xing et al 2022, Yang et al 2022).Therefore, according to the characteristics of CHARLS data, we further investigate the heterogeneity of the results between northern and southern regions, urban and rural areas, and different education levels8 .The figure 2(a) shows the relationship between extreme temperatures and medical costs for older adults in northern and southern China.We found a U-shaped relationship between temperature and medical costs among older people in the south.However, the relationship between extreme cold temperature in the north and the medical cost of older adults is not statistically significant.This may be because northern China provides public coal-fired heating in winter, and we can also see from our sample that the north has much more heating installations per capita than the southern population (As shown in figure 3(a)).Stable and continuous heating situation can reduce the harm of extreme cold temperatures to older adults, which in turn lead to lower medical costs.
To investigate urban-rural differences in the effect of temperature on the medical expenses of older adults, we compared the results in figure 2(b), revealing a U-shaped relationship between temperature and older adults' medical expenses in urban and rural groups.For example, for older adults in urban areas, extremely hot weather (above 25 • C) significantly increases the medical expenses of older adults.This may be due to the heat island effect (Jenerette et al 2016, Schinasi et al 2018).However, extreme cold weather does not affect them significantly.Zhang et al (2023) also found that rural residents are affected more by extreme cold weather than urban residents, mainly because their per capita disposable income is too low to have access to safe and reliable energy, and the airtightness and warmth in rural homes are worse than in urban areas.Some of the data from this study give similar evidence.As shown in figure 3, during the study period, the per capita air conditioning and heating facility ownership rates of rural older adults are much lower than that of urban older adults.
Figure 2(c) shows the effects of temperature on medical expenses of older adults with different education levels, revealing that the medical expenses of older adults with less than nine years of education are more significantly affected by temperature fluctuations.This may be because older adults with lower education levels are more likely to live and work in relatively poor environments.Older adults with lower education levels are more likely to have worked in individual production and agricultural activities, leading to unfixed workplaces or relatively poor working environments, making them more vulnerable to extreme temperature exposure (see figure S1 in the supplementary materials 3 for more details).Note: Authors' calculations from CHARLS data (2011, 2013, 2015, and 2018).In conclusion, the medical costs for older adults in the South and older adults in rural areas are more affected by low temperatures.In addition, older adults with lower education levels are also more affected by temperature fluctuations.It is necessary to establish measures to older adults against extreme temperatures.
In addition, we added an alternative specification that interacts group dummy variables of interest with the Temp variables.Specifically, we added these three interaction terms in equation ( 1) and reperformed the regression.The results are shown in table 1.At the regional level, the interaction coefficient between the extreme low temperature chamber and the regional level was negative and significant, which indicating that the medical costs for older adults in the North will reduce by 0.3% compared to that in the South at extremely low temperatures.Under extreme high temperatures, the interaction term coefficient at the regional level is not significant.This suggests that there is no significant difference in medical costs among older people in different regions under extreme heat.In terms of the urban-rural level, the interaction coefficient was also negative and significant, indicating that older adults' medical costs in the urban areas will be reduced by 0.4% compared to that in the rural areas at extremely low temperatures.Similar to regional heterogeneity, there was no significant difference in extreme high temperature impact on the medical costs of urban and rural older adults.In terms of education level, differences in education levels can have a significant impact on older adults' health expenditures, regardless of exposure to extreme heat or extreme cold.In the extreme heat exposure, higher levels of education can reduce health care expenditures by 0.7%; In the extreme cold exposure, higher levels of education can reduce health care expenditures by 0.1%.

Predicting older adults' future medical costs
Figure 4(a) presents the projected temperature distributions for 2030 and 2050.In the future, populations in 2030 and 2050 are likely to face less cold weather and more hot weather than in the base period.Figure 4(b) illustrates the distribution of medical expenses for older adults in China in 2030 and 2050.In the context of climate change, if the world follows the RCP 8.5 scenario and does not initiate more active carbon emissions reduction measures, the medical expenses of older adults will reach 1.9 trillion CNY in 2030 and 2.7 trillion CNY in 2050, and the per capita medical expenses of older adults in these two periods are 5135.1 CNY and 5294.2CNY, respectively, respectively.If the world follows the RCP 4.5 scenario and achieves relatively modest carbon reduction measures, older adults' medical costs will fall by 27.5 billion CNY in 2030 and 37.2 billion CNY in 2050 compared with RCP 8.5 scenario.If the world follows the RCP 2.6 scenario in which a global temperature increase of less than 2 • C in 2100 compared with the pre-industrial era, the costs of older adults in 2030 and 2050 will be reduced by 39.8 billion CNY and 83.7 billion CNY, respectively.
The specific effects of the temperature day changes are presented in table 2. First, the results are consistent across all scenarios, demonstrating that a change in the hot days (over 25 • C) would significantly increase healthcare costs for older adults, while changes in extremely low temperature bins (below −5 • C) would reduce healthcare costs for older adults.Second, the predicted total temperature impact is positive for all scenarios.Third, the differences between medium-and long-term impacts are clear across scenarios; for example, under the RCP 4.5 scenario, the long-term temperature impact is almost twice as large as the medium-term impact, suggesting that if global warming continues to deepen, older adults' health costs related to temperature change will also increase significantly.
As is shown in figure 5, the burden of medical costs incurred by older adults in the context of climate change exhibits strong spatial heterogeneity.In all regions, older adults in eastern China generate the most medical expenses, and more than one-third of the older adults' medical costs (37.5%-38.4%) in 2050 are concentrated in the eastern region of China, which is primarily attributable to the fact that the eastern region of China will face more serious extreme high temperatures in the future.In addition, the change in total medical cost due to climate change in the eastern region is the highest in all regions, accounting for 34.5%-35.8%.And the change in medical cost per capita in this region is above average level.Specifically, Shandong Province and Guangdong Province will bear the heaviest health burden of older adults in the eastern region, accounting for 8.1%-8.7% and 5.6%-7.0% of the total medical costs of older adults, respectively.And the change in total medical cost due to climate change in Shandong Province is the highest in eastern region, accounting for 7.8%-10.1%.Notably, compared with the baseline period, the medical costs of older adults in less developed western provinces with weak adaptive capacity (primarily Guangxi Province and Sichuan Province) will more than triple in 2050.Especially, they have the largest per capita increase in medical cost due to climate change.In addition, we find most western provinces' medical burden rates of older adults are higher than the national average level (see figure S5(b) in the supplementary materials 5 for more details).This suggests that climate change will exacerbate inequality across provinces, adding more difficulties to the economically backward regions.

Conclusions and recommendations
This study conducts a comprehensive investigation based on data covering most provinces and cities in China, revealing a clear U-shaped relationship between temperature and medical expenses of older residents.Specifically, older adults in the South and older adults in rural areas are more affected by low temperatures.And the medical costs of older adults with low-education level are also more affected by extreme temperature compared with the higheducation level.The prediction results show strong spatial disparities in the temperature-related health burden across China, medical costs for older adults in less developed western provinces (primarily Guangxi and Sichuan) will more than triple by 2050 compared with base period.Governments should focus on tailoring policies for less adaptable provinces to mitigate the health threats of extreme temperature exposure to older adults.This study predicts the healthcare cost implications of temperature based on long-term panel data from 2011, 2013, 2015, and 2018.In this case, we only consider the baseline health costs and fail to consider the advancements in healthcare and the increased availability of adaptation facilities in the future.It may overestimate the changes in healthcare costs due to climate change in 2050.In addition, due to data limitations, we only consider the total medical expenses of older adults, and could not consider the composition of each disease in total medical expenses and temperature-associated health costs arising from seasonal variations.In the future, we hope to enrich our research based on the availability of relevant data.Moreover, the CHARLS sample currently only has national data for a wide range of provinces and cities in 2011, 2013, 2015 and 2018.The availability of more comprehensive data in the future should further enrich our understanding of the relationship between temperature and medical costs for older adults in developing countries.

Figure 1 .
Figure 1.Effect of temperature on the older adults' medical expenses.Note: Two dashed lines indicate 90% confidence intervals.

Figure 4 .
Figure 4. Temperature and healthcare cost projections under three RCP scenarios.Note: CNY: Chinese yuan.

Figure 5 .
Figure 5. Projected distribution of medical costs for older adults in 2050 by province (10 8 CNY).