Assessment of health economic losses caused by PM_2.5 and ozone pollution in Beijing and Tianjin

Abstract

This study assesses the health economic losses caused by PM_2.5 and ozone pollution in Beijing and Tianjin from 2016 to 2021 using Poisson regression relative risk model and value of statistical life method. The results show that the health economic losses caused by PM_2.5 pollution in Beijing and Tianjin fluctuating decrease from 6.115 billion and 2.699 billion RMB in 2016 to 4.39 billion and 2.298 billion RMB in 2021. The health economic losses caused by ozone pollution in Beijing increase from 6.358 billion RMB in 2016 to 11.333 billion RMB in 2019, with a maximum increase of 25.91% during the period, and then decrease to 7.825 billion RMB in 2021, but the loss value is still higher than that in 2016. In Tianjin, the health economic losses caused by ozone pollution with the same change trend as that in Beijing. Besides, the study shows that although the health economic losses caused by PM_2.5 and ozone pollution in Beijing and Tianjin have fluctuating decreased in recent years as a proportion of GDP in the corresponding years, the loss values are still relatively large, the proportion of health economic losses caused by ozone pollution in total losses is becoming higher and higher.

Introduction

Numerous studies have shown that air pollution can seriously affect human health, and the harm of PM₁₀, PM_2.5 and ozone pollution is particularly significant. They are easy to cause lesions in various organs after being inhaled by the human body. With the increase of the amount of inhalation, they will continue to accumulate in the respiratory system, infringe the respiratory system, and induce dyspnea, asthma and other diseases. In addition, epidemiological investigations have found that air pollutants have acute and chronic health effects on the human body. Acute health effects are mainly reflected in the significantly increased risk of acute respiratory diseases, cardiovascular and cerebrovascular diseases in the environment with high exposure to pollutants. Chronic health effects are reflected in the fact that pollutants may induce chronic diseases such as chronic obstructive pneumonia and lung cancer, and even affect the human immune system and nervous system^1,2,3,4. As a developing country with a large population, China has also encountered air pollution problems while developing its economy, which has caused certain health economic losses^5,6,7. In fact, not only China but most countries are facing the varying degrees problem of the impact of air pollution on people’s health, so it is particularly important to assess the health economic losses caused by air pollution.

As the main air pollutants, PM_2.5 and ozone have been deeply studied by many scholars on the health economic losses caused by them. Some scholars used statistical methods such as descriptive statistics, distribution function and its parameter optimization to assess the health impacts caused by air pollutants, and found that compared to high-income areas, the health impacts and economic burdens on low-income areas were more severe^8,9,10. Additionally, several scholars combined epidemiological and mathematical methods to quantify the impact of PM_2.5 and ozone exposure on health, and found that the economic losses caused by them were both close to 100 million RMB^11,12. There are also studies employed the cost of illness approach and generalized additive models to unveil the impacts of air pollution on health, encompassing multi-dimensional economic losses such as hospitalization expenses for depression, health expenditures and productivity losses among the working population, and excess deaths resulting from regional ozone exposure^13,14,15.

From the process of literature review, it can be found that in the study of health economic losses caused by air pollution, scholars mainly use mathematical or statistical models^7,8,9,12,15, environmental economics^16,17 and epidemiological methods^10,11,13,14 to estimate the economic losses and governance costs caused by air pollution. The calculation mode of these methods is relatively single, and they mainly use non pollutant concentration data to directly calculate economic losses, this will result in varying degrees of deviation in the calculated loss values, so it is necessary to fully explore the potential relationship between pollutant concentration data and economic losses, and establish more refined models to calculate economic losses. Combining Poisson regression with value of statistical life can effectively achieve this goal.

Poisson regression is frequently used to solve count-related problems influenced by exposure levels, such as the number of cases of a certain disease in a specific region or the number of hospital visits^18,19. It is also applied in risk assessment and prediction in fields like public health and insurance^20,21,22. Additionally, some researchers have employed it to explore the associations between air pollution and outpatient visits, hospitalizations, and mortality rates^23,24,25. Obviously, Poisson regression can only predict the number of deaths or hospitalizations caused by pollution and cannot be directly used to evaluate economic losses. This study further combines the value of statistical life to monetize life value, thereby converting health risks into quantifiable economic losses. Specifically, we selects two representative cities in transmission channel cities of air pollution in China, Beijing and Tianjin, which have serious pollution, innovatively combines value of statistical life and Poisson regression to model pollutant concentration data to evaluate the health economic losses caused by PM_2.5 and ozone pollution, indirectly confirming the importance of pollution prevention and control policies in reducing the burden of air pollution in high-risk urban agglomerations.

The rest of this study is arranged as follows. Section"Statistical methods"introduces the statistical methods used to assess the health economic losses caused by PM_2.5 and ozone pollution. Section"Assessment model and results"gives the assessment model used in this study and the setting of model parameters. Section"Discussion"presents the assessment results of health economic losses and conducts in-depth analysis of the results. Section"Conclusions"summarizes and further discusses this study.

Statistical methods

In this study, the product of health impacts and unit economic losses was used to estimate the total economic losses caused by PM_2.5 and ozone pollution²⁶. The health impacts refer to the number of people with health damage caused by PM_2.5 and ozone pollution, which is a counting event. The Poisson regression relative risk model is most suitable for modeling when the result is a counting event, which can assess the relative risk of a certain event between different exposure levels. Therefore, compared with other methods, the Poisson regression relative risk model is more suitable for modeling PM_2.5 and ozone pollution data to estimate the relative risk population exposed to different pollutant concentrations. In addition, this study uses value of statistical life to indicate the cost that people are willing to pay to reduce the risk of death caused by pollutants, namely, the unit economic loss of death caused by PM_2.5 and ozone pollution.

Poisson regression

Poisson regression²⁷ is a type of regression analysis used to model counting data, contingency tables, etc. Poisson regression assumes that the response variable Y follows a Poisson distribution, that is, Y satisfies stationarity, independence and generality, and assumes that the logarithm of its expected value can be modeled by a linear combination of unknown parameters.

Specifically, let $x_{0}$ be a multidimensional column vector composed of n independent variables, then the mathematical form of Poisson regression is

(1)

where x is an (n + 1) dimensional column vector, consisting of n original independent variables with a constant vector of element 1, and $\theta$ is the column combination of β and $c_{0}$.

Therefore, if the coefficient θ and characteristic variable x in the Poisson regression model are known, the expected value of the response variable satisfying the Poisson distribution can be predicted by the following formula:

$$E\left( {Y\left| {\text{x}} \right.} \right) = e^{\theta ^{\prime}x} .$$

(2)

The value of the parameter θ can be obtained by maximum likelihood estimation, that is, y_i is assumed to be the observed value of the response variable Y and the corresponding characteristic variable is x_i, i = 1, 2, …, m. Since Y follows the Poisson distribution, its probability density function is

(3)

For m observations (x_i,y_i), maximum likelihood estimation is to find the θ that maximizes the joint probability based on the current observations, therefore, there is

(4)

take the logarithms on both sides of Eq. (4) and simplify it to

(5)

That is, the problem becomes to estimate the value of the parameter θ that maximizes the Eq. (5). To solve this problem, the first derivative of l(θ|X,Y) with respect to θ can be set to 0, and the Newton–Raphson algorithm can be used to initialize θ randomly, and then the θ can be iteratively updated until its convergence.

Value of statistical life

The value of statistical life (VSL) in the economic sense refers to the value of life measured by physical objects, money, etc. From a statistical point of view, considering the correlation between the probability of reducing risk and the cost of paying, the value of an individual’s life is measured by the cost that an individual is willing to pay to reduce the risk of death. That is, the cost of payment is used to replace the risk of death, thus putting forward the value of statistical life method²⁸.

The assessment methods for value of statistical life mainly include wage risk method, conditional value method and choice experiment method. The assessment results of each method vary in different application scenarios^29,30,31, but the core step of the assessment is to design the questionnaire. Arrow et al. believed that when designing a questionnaire, the binary method is more aligned with people’s daily consumption decision-making behavior, and the willingness to pay obtained is closer to the true value³². Therefore, this study uses a binary method to calculate the value of statistical life of health economic losses caused by PM_2.5 and ozone pollution³³. Specifically, according to utility theory, the expected utility function U consists of two parts: random change and non-random change, namely, the utility fixed function and the utility probability variable function, which are expressed by the formula as follows:

$$U\, = \,V\, + \,\varepsilon$$

(6)

(7)

where the variable V is the utility fixed function and ε is the utility probability variable function. c is the constant term, β represents the regression coefficient of bid_I, bid_I represents the initial bid value, b is the regression coefficient of r, r represents the level of risk of death, γ_m represents the regression coefficient of X_nm, and X_nm is the m-th characteristic of an individual.

In fact, the distribution of utility probability variable function ε follows a Logit model. When applied to the value of statistical life, its value usually has two cases, which can be regarded as a discrete distribution series of two points with values of 0 and 1. The probability when the value is 1 can be expressed as

(8)

According to Hanemann³⁴, when the willingness to pay (WTP) is greater than or equal to 0, the average willingness to pay of the interviewees is

(9)

According to the definition of value of statistical life, the expression of value of statistical life obtained by the binary method is as follows:

(10)

Where T represents the number of years of payment and ∆R represents the change in risk of death.

Assessment model and results

This section present the health impact assessment model for PM_2.5 and ozone concentrations, the health economic loss assessment method, the values of each parameter in the model and the estimation method.

Health impacts assessment

In this study, the relative risk model is used to assess the health impacts of reducing PM_2.5 and ozone concentrations. The model equation is as follows:

$$\Delta y\, = \,[(RR\left( c \right){-\!\!-}{1})/ \, RR\left( c \right)]\, \times \,y\, \times \,Pop,$$

(11)

here.

$$RR\left( c \right)\, = \,exp[\beta (c{-\!\!-}c_{0} )].$$

(12)

Equation (12) is the Poisson regression equation, where ∆y represents the change in health impacts, the term (RR(c)—1)/RR(c) is the attributable fraction and RR(c) represents the relative risk³⁵, c is the concentration of pollutants, y represents the health impacts when the concentration of pollutants is c, Pop is the number of exposed populations, β is the exposure–response coefficient, and c₀ is the pollutant concentration limit.

Assessment method for unit health economic losses

The research on the value of statistical life related to air pollution in various cities in China is not complete. Combined with the research conclusions of some scholars, this study uses the following formula to estimate the value of statistical life of the cities included in the study³⁶:

$$VSL_{A,t = } VSL_{BJ} \times \left( {I_{A,t} /I_{BJ} } \right)^{a} \times \left( {1 + \% G_{t} + \% P_{t} } \right)^{a} .$$

(13)

Here VSL_A,t represents the value of statistical life of PM_2.5 and ozone pollution of city A in year t, VSL_BJ represents the value of statistical life of haze pollution in Beijing in 2016³³. I_A,t represents the per capita disposable income of City A in year t, while I_BJ represents the per capita disposable income of Beijing in 2016. %∆G_t represents the per capita GDP growth rate of Beijing from 2016 to year t, and %∆P_t represents the CPI growth rate of Beijing from 2016 to year t. a represents the income elasticity coefficient, which is taken as 1 in this study.

The health economic losses caused by hospitalization are usually estimated by cost of illness approach, which mainly includes the medical expenses required for the treatment of diseases and the economic losses caused by missed work. The specific formula is as follows:

$$CI_{k} \, = \,\left( {CI_{pk} \, + \,G\, \times \,D_{k} } \right)\, \times \Delta y\,$$

(14)

in the formula, CI_k represents the total disease cost caused by health terminal k, and CI_pk represents the total medical expenses (direct and indirect medical expenses) per unit case of health terminal k. G represents the daily per capita GDP of a certain region, and D_k represents the time of missed work due to health terminal k, in this study, which is calculated based on the average hospitalization days of a health terminal, where ∆y represents the change in health impacts.

Model parameters setting and data sources

According to the Meta-analysis³⁷ and the research results of some scholars, this study select all-cause (AC) death, respiratory system-related (RSR) death, cardiovascular and cerebrovascular system-related (CCSR) death, respiratory system disease-related (RSDR) hospitalization, cardiovascular and cerebrovascular disease-related (CCDR) hospitalization as the health terminal, and combines the aforementioned models to assess the health impacts of reducing PM_2.5 and ozone concentrations. The specific parameters of the model are shown in Table 1.

Table 1 PM_2.5 and ozone exposure–response coefficient of health terminal.

This study uses the annual average PM_2.5 concentration and the 90 th percentile of the daily maximum 8-h average ozone concentration in Beijing and Tianjin from 2016 to 2021 to assess the health economic losses caused by these two pollutants. Their specific values and trends are shown in Fig. 1, and the data are obtained from the Environmental Status Bulletins published annually by the Beijing Municipal Ecology and Environment Bureau and Tianjin Ecology and Environment Bureau. These bulletins provide official air quality monitoring data collected from government-operated stations across both municipalities.

Fig. 1

Annual average PM_2.5 concentration and the 90 th percentile of the daily maximum 8-h average ozone concentration in Beijing and Tianjin.

The values of y of each health terminal under the death item in the Poisson regression relative risk model, namely the baseline incidence rate, are derived from statistical yearbooks and statistical bulletins of Beijing and Tianjin for various years, with the specific values listed in Table 2. The baseline incidence rate of each health terminal under the hospitalization item is based on the research results of Lv and Li⁴³.

Table 2 Baseline incidence rate of each health terminal (%).

The economic losses caused by death-related factors in various health terminals are estimated using the value of statistical life method. The specific formula is given in Section"Assessment method for unit health economic losses". From the statistical yearbooks and statistical bulletins of Beijing over the years, it can be obtained that the per capita GDP of Beijing from 2016 to 2021 is 123,391, 136,172, 150,962, 161,776, 164,889 and 183,997 RMB, respectively. Therefore, the per capita GDP growth rates from 2017 to 2021 relative to 2016 are 10.36%, 22.34%, 31.11%, 33.63% and 49.12%; taking 2016 as the base year, the CPI growth rates from 2017 to 2021 are 1.9%, 4.4%, 6.8%, 8.7% and 9.9% in turn.

The per capita disposable income of Beijing and Tianjin in each year is shown in Fig. 2, and the data are extracted from the Beijing Statistical Yearbook and Tianjin Statistical Yearbook (2016–2021 editions), which are publicly accessible through the official portals of the Beijing Municipal Bureau of Statistics and Tianjin Bureau of Statistics. According to Eq. (13), the value of statistical life of Beijing and Tianjin in each year can be calculated.

Fig. 2

Per capita disposable income.

For the assessment of economic losses caused by hospitalization of various health terminals, this study uses the cost of illness approach. Disease costs include all direct costs and indirect costs during patient visits. The average hospitalization costs and average hospitalization days for respiratory diseases are replaced by the average hospitalization costs and average hospitalization days for tuberculosis, pneumonia and pulmonary heart disease. The hospitalization costs for corresponding diseases can be obtained from the China Health Statistics Yearbook for each year. Specifically, for the average hospitalization costs in Beijing and Tianjin, this study assumes that the ratio of the per capita disposable income of the residents in the country to the hospitalization costs of various diseases is equal to the ratio of the per capita disposable income of the residents in the two cities to the corresponding hospitalization costs of various diseases, thus estimating the average hospitalization costs of various diseases in the two cities. Similarly, the average hospitalization costs of acute myocardial infarction are used to replace the average hospitalization costs of cardiovascular and cerebrovascular diseases. Specifically, for Beijing and Tianjin, the calculation method is the same as that of respiratory system diseases. The average hospitalization days caused by respiratory system diseases and cardiovascular and cerebrovascular diseases in the two cities are replaced by the average hospitalization days of corresponding diseases in the country.The specific values are listed in Table 3.

Table 3 Average hospitalization days and medical expenses of each health terminal.

The daily per capita GDP for Beijing and Tianjin in each year is given in Fig. 3. In addition, the number of permanent residents in Beijing and Tianjin at the end of the year (the data is sourced from the statistical yearbooks over the years) is used as the exposed population in the Poisson regression relative risk model. After reviewing and estimating all the above data, the unit economic losses for different health terminals in each year can be obtained according to Eqs. (13) and (14), as shown in Table 4.

Fig. 3

Daily per capita GDP of Beijing and Tianjin.

Table 4 Unit economic losses of each health terminal.

In general, the health impacts (∆y) calculated via Eqs. (11) and (12) serve as the foundation for estimating total economic losses. Mortality-related losses are quantified using the Value of Statistical Life (VSL), where the number of deaths averted (e.g., AC, CCSR, RSR deaths in Tables 5–6) is multiplied by the corresponding VSL (Table 4). For hospitalization-related losses, the reduction in cases (e.g., RSDR, CCDR hospitalizations) is combined with unit costs (direct medical expenses and productivity loss) from Eq. (14). This dual-method approach ensures that both fatal and non-fatal health outcomes are monetized, directly linking pollutant concentrations to economic burdens.

Table 5 Health impacts of PM_2.5 concentration reaching the first-level limit of air quality standard (person).

Table 6 Health impacts of ozone concentration reaching the first-level limit of air quality standard (person).

Further, this study select the PM_2.5 first-level concentration limit of 15 µg/m³ and the ozone first-level concentration limit of 100 µg/m³ as the pollutant concentration limits specified in the Ambient Air Quality Standards of China (GB3095-2012).

Discussion

Based on the data of pollutant concentrations, concentration limits, expose-response coefficients, baseline incidence rates of various health terminals, and the number of permanent residents at the end of the year, combined with the calculation formula of the Poisson regression relative risk model, the change of health impacts with pollutant first-level concentration limit as the baseline is estimated. Then, combined with the unit economic losses of various health terminals, the total economic losses of the corresponding health terminals in each year are estimated. The following is divided into two parts: evaluating and analyzing the health impacts of PM_2.5 and ozone concentrations reaching the first-level limit of air quality standard in Beijing and Tianjin from 2016 to 2021, and the health economic losses caused by PM_2.5 and ozone concentrations not reaching the first-level limit.

Health impacts analysis

Table 5 shows the health impacts of each health terminal when the PM_2.5 concentration in Beijing and Tianjin reaches the first-level concentration limit from 2016 to 2021. From the table, it can be seen that the health impacts of various health terminals in Beijing and Tianjin show a downward trend from 2016 to 2021, and the decline is significant. Compared with 2016, the total health impact in Beijing in 2021 decreased by 68.62%, while that in Tianjin decreased by 57.72%. Obviously, due to differences in pollutant concentrations, exposure–response coefficients of each health terminal and baseline incidence rates, the degree of PM_2.5 impact on health terminals varies significantly. The health impacts of various health terminals in Beijing are significantly higher than those in Tianjin during the same period.

Specifically, the most obvious health impact brought by reducing PM_2.5 concentration in Beijing is the number of respiratory system disease-related hospitalizations. And the health impacts from 2016 to 2021 are 25,805 persons (95% confidence interval: 5573, 45,437), 19,292 persons (95% confidence interval: 4165, 34,175), 16,310 persons (95% confidence interval: 3550, 28,949), 12,211 persons (95% confidence interval: 2632, 21,777), 10,153 persons (95% confidence interval: 2102, 18,221) and 8098 persons (95% confidence interval: 1076, 14,534), its proportion in total health impacts remains basically at 61%. The health impact of cardiovascular and cerebrovascular disease-related hospitalization in each year accounts for about 20% of the total health impact, while the health impacts related to death are relatively low, that is, when the PM_2.5 concentration reaches the first-level concentration limit, the number of people who are hospitalized benefited more, accounting for approximately 81% of the total health impact. Regardless of the type of health terminal, the number of beneficiaries show a decreasing trend from 2016 to 2021, but the proportion of the total health impact remains almost unchanged.

For Tianjin, the change trend of various health terminals is the same as that of Beijing, but the proportion structure of the total health impact is different from that of Beijing. For example, the proportion of respiratory disease-related hospitalization is around 59%, slightly lower than that of Beijing.The health impact rankings of other health terminals are also the same as those of Beijing, but the number of beneficiaries in the same health terminal is significantly lower than that of Beijing. As shown in Fig. 1, the average annual PM_2.5 concentration in Tianjin is almost higher than that in Beijing in each year, but the number of beneficiaries from reducing PM_2.5 concentration is less than that in Beijing. This difference arises because health impacts depend on multiple factors: PM_2.5 concentration, pollutant concentration limit, exposed population, exposure–response coefficient and health terminal baseline incidence rate. Although the average annual PM_2.5 concentration and the reduction of relevant variables in each year are different in the two cities, the exposed population in Beijing is higher than that in Tianjin, the baseline incidence rate related to hospitalization in Beijing is higher than that in Tianjin, and the proportion of the health impact related to hospitalization in the total health impact exceeds 80%.

In addition, the total health impact of Beijing and Tianjin in each year is as follows: Beijing: 25,805, 19,292, 16,310, 12,211, 10,153, 8098 persons, with an average annual decrease of 20.59%; Tianjin: 13,925, 11,710, 9110, 8825, 8135, 5887 persons, with an average annual decrease of 15.34%. The difference in the decrease between the two cities is significant, which also reflects the significant effect of Beijing’s governance on PM_2.5 concentration in recent years, a series of measures have played a positive role in prevention and control. For any healthy terminal, no matter which city, taking corresponding measures to reduce PM_2.5 concentration can greatly improve people’s health impacts, and the health impacts caused by it are very considerable, which is of great significance for improving residents’ lives and air environment.

This study also estimate the health impacts of reducing ozone concentration to the first-level concentration limit, as shown in Table 6. Similar to the health impacts of PM_2.5, the health impacts of ozone on respiratory system disease-related hospitalization is the most pronounced, with the proportion of respiratory system disease-related hospitalization health impacts in Beijing accounting for 44.72%, 44.66%, 43.92%, 44.36%, 46.19% and 44.28%, respectively, with an average of 44.69%. For the same health terminal, in contrast, the proportion of health impacts caused by reducing PM_2.5 concentration to the total health impacts is much higher. In addition, compared with the data related to PM_2.5, the health terminal that ranks second has changed to all-cause death, with an average proportion of 21.50%, showing a significant increase. The proportion of respiratory system disease-related hospitalization health impacts in Tianjin in each year accounts for 39.48%, 40.64%, 40.55%, 40.94%, 40.42% and 40.17%, respectively, which is basically around 40%, accounting for the largest proportion among the five health terminals. And the number of beneficiaries from reducing ozone concentration to the first-level concentration limit in Tianjin is lower than that in Beijing in each year, mainly due to the fact that the number of exposed population in Beijing is significantly higher than that in Tianjin.

In Beijing, the total health impacts of reducing the ozone concentration to 100 µg/m³ are 12,482, 11,743, 11,798, 11,547, 9029 and 6253 persons, respectively. Except for a slight increase in 2018 compared to the previous year, the rest of the years have a decreasing trend, with annual change rates of −5.920%, 0.47%, −2.13%, −21.80% and −30.74%. The decrease in the past two years has increased significantly, with an average annual change rates of −12.02%, that is, taking 2016 as the base year, the total health impact in Beijing has decreased by 12.02% annually. The total health impacts in Tianjin are 4541, 6930, 7473, 7338, 6707 and 4469 persons, showing an increasing trend from 2016 to 2018, and then begin to decline after 2018. And the annual change rates are 52.62%, 7.83%, −1.80%, −8.60% and −33.37%, respectively. Compared with 2016, the total health impact in 2017 increase by more than half, and the average annual change rate is 3.33%. In other words, the average annual increase of total health impact in Tianjin from 2016 to 2021 is 3.33%, but in recent years it has shown a downward trend, and the decline of it and Beijing reach more than 30%.

By comparing Table 5 and Table 6, it is easy to find that the health impacts of reducing ozone concentration in Beijing and Tianjin are almost lower than those of reducing PM_2.5 concentration, except for cardiovascular and cerebrovascular system-related death health terminal, which is much higher. This also indicates that PM_2.5 is still one of the main pollutants that damage human health, but the health problems caused by ozone cannot be ignored. In Tianjin, the proportion of hospitalization-related health impacts caused by ozone pollution accounts for 53.1%, which is also significantly lower than the proportion of hospitalization-related health impacts caused by PM_2.5. Therefore, it can be considered that reducing PM_2.5 concentration will benefit more people in hospitalization-related diseases, while reducing ozone concentration may bring more death-related health impacts. In terms of the total health impact, the annual change rate of the total health impact brought by reducing PM_2.5 concentration in the two cities is almost above 15%, and the change range is relatively stable. However, the fluctuation range of the average annual change rate of the total health impact brought by reducing ozone concentration is relatively large, and there is a positive change rate in the early years, that is, the number of beneficiaries increase compared with the previous year. This indicates that the impact of PM_2.5 pollution on people’s health is gradually decreasing. Although the impact of ozone pollution has been decreasing in recent years, it shows a certain upward trend before 2018, which needs to be concerned about its long-term effects. In addition, the average annual change rates of the total health impacts caused by reducing PM_2.5 and ozone concentration in Beijing are −20.59% and −12.02%, respectively, while those in Tianjin are −15.34% and 3.33%, which also shows that the proportion of ozone’s impact on the health of residents in the two cities is gradually increasing.

Economic losses assessment

Based on the assessment results of health impacts in the previous section, combined with the unit economic losses of each health terminal in Table 4, the indirect economic losses caused by the failure to reduce PM_2.5 and ozone concentrations to the first-level concentration limit in Beijing and Tianjin from 2016 to 2021 are estimated using relevant formulas. The results are listed in Tables 7 and 8. And the trend of total economic losses are shown in Fig. 4 and Fig. 5. To illustrate the workflow, consider Beijing in 2016:

Table 7 Economic losses caused by PM_2.5 pollution (100 million RMB).

Table 8 Economic losses caused by ozone pollution (100 million RMB).

Fig. 4

Total economic losses caused by PM_2.5.

Fig. 5

Total economic losses caused by ozone.

PM2.5-related RSDR hospitalizations reduced: 15,871 persons (Table 5). Unit hospitalization cost (RSDR): 18,922.44656 RMB/person (Table 4). Hospitalization loss: 15,871 × 18,922.44656 ≈ 3.00 × 10⁸ RMB (Table 7). AC deaths averted: 2598 persons (Table 5). VSL (2016 Beijing): 1,197,584 RMB/person (Table 4). Mortality loss: 2598 × 1,197,584 ≈ 3.111 × 10⁹ RMB (Table 7). This demonstrates how health impacts are systematically converted into economic losses.

As for the economic losses caused by PM_2.5, the comparison with Table 5 can be found that although the health impacts of hospitalization-related health terminals account for a relatively high proportion, the corresponding economic losses account for a significantly lower proportion. In Beijing, the proportions in each year are 9.86% (95% confidence interval: 6.88%, 10.62%), 8.42% (95% confidence interval: 5.89%, 9.10%), 6.92% (95% confidence interval: 4.87%, 7.48%), 6.29% (95% confidence interval: 4.48%, 6.80%), 6.97% (95% confidence interval: 5.14%, 7.47%) and 5.49% (95% confidence interval: 4.07%, 5.88%), that is, although the total economic loss reaches a stage peak of 6.361 billion RMB (95% confidence interval: 2.948, 9.798) in 2018, the proportion of hospitalization-related economic loss to total economic loss shows a downward trend year by year, with an average annual decrease rate of 10.40%. The proportions of hospitalization-related economic losses in Tianjin are 8.15% (95% confidence interval: 5.82%, 8.73%), 7.54% (95% confidence interval: 5.45%, 8.06%), 6.36% (95% confidence interval: 4.62%, 6.81%), 5.79% (95% confidence interval: 4.27%, 6.18%), 5.48% (95% confidence interval: 4.13%, 5.82%). and 4.72% (95% confidence interval: 3.59%, 5.00%), respectively. Similarly, the total economic losses decrease in a fluctuating manner (i.e., there is a stage peak), but the proportion of hospitalization-related economic loss to total economic loss decreases year by year, with an average annual decrease rate of 10.28%, similar to that in Beijing. In terms of economic losses related to death, the economic loss of respiratory system-related death in Beijing is higher than that of cardiovascular and cerebrovascular system-related death in the same year at the early years, but in recent years, the losses of both aspects have gradually become equal. While in Tianjin, on the contrary, the economic loss of cardiovascular and cerebrovascular system-related death is higher than that of respiratory system-related death, and the multiple between the former and the latter is increasing every year, reaching as high as 2.55 times by 2021. Both in Beijing and Tianjin, the economic losses of hospitalization-related to these two diseases are very similar in each year.

In terms of total economic loss, Beijing is significantly higher than Tianjin, with the loss value remaining at more than 2 times, the main reason may be that the number of exposed population in Beijing is much higher than that in Tianjin. This finding aligns with Wang et al.⁶, who also reported higher PM_2.5-related economic losses in densely populated urban centers in the Beijing-Tianjin-Hebei region. In terms of different years, the total health impact in Beijing and Tianjin from 2016 to 2021 gradually decrease, while the total economic loss shows a wave peak, rather than a constant downward trend. However, compared with 2016, the economic loss in 2021 drop to the lowest point during the study period, mainly due to the PM_2.5 concentration has decreased significantly in recent years. And in 2021, the PM_2.5 concentration in Beijing has reached the second-level concentration limit of 35 µg/m³ in the Ambient Air Quality Standards of China. This is contrary to the conclusion reported by Zhou et al.⁴⁰ that the total economic losses caused by PM_2.5 nationwide from 2000 to 2017 exhibited a fluctuating upward trend. This discrepancy may stem from stricter local emission controls in Beijing post-2017, such as the"Clean Air Action Plan,"which prioritized coal-to-gas transitions and industrial restructuring¹⁷. In general, due to differences in PM_2.5 concentration, exposure–response coefficient of each health terminal, baseline incidence rate, unit economic loss and exposed population in the two cities in different years, the economic loss caused by failure to reach the first-level concentration limit of PM_2.5 is also different.

The economic losses caused by ozone pollution are shown in Table 8. From the table, it can be concluded that the total economic losses of various health terminals in Beijing from 2016 to 2021 are 6.358 billion RMB (95% confidence interval: 3.751, 8.925), 7.906 billion RMB (95% confidence interval: 4.652, 11.113), 9.954 billion RMB (95% confidence interval: 5.852, 13.998), 11.333 billion RMB (95% confidence interval: 6.668, 15.933), 8.837 billion RMB (95% confidence interval: 5.268, 12.367) and 7.825 billion RMB (95% confidence interval: 4.643, 10.983). From 2016 to 2019, the total economic loss increase year by year, with a maximum increase of 25.91%, and from 2019 to 2021, there is a downward trend, but the total economic loss in 2021 is still higher than in 2016. The rising trend of ozone-related losses until 2019 mirrors findings from Xie et al.¹⁰, who attributed similar patterns in the Beijing-Tianjin-Hebei region to increased photochemical reactions driven by warmer temperatures and stagnant atmospheric conditions. However, the post-2019 decline in our study contrasts with global meta-analyses (e.g., Vandyck et al.³), which suggest persistent ozone risks under climate change. This divergence could reflect the effectiveness of China’s 2018–2020"Three-Year Action Plan for Blue Skies", which targeted VOC and NO_x reductions—key precursors for ozone formation. Furthermore, death-related economic loss accounts for more than 96% of the total loss in Beijing in each year, reaching a maximum value of 98.35% in 2021. Among them, the economic loss of all-cause death health terminal is the highest, followed by cardiovascular and cerebrovascular system-related death, so controlling ozone pollution can reduce more death-related economic losses.

The total economic losses caused by ozone in Tianjin in each year are 1.885 billion RMB (95% confidence interval: 1.128, 2.636), 3.389 billion RMB (95% confidence interval: 2.042, 4.717), 4.4 billion RMB (95% confidence interval: 2.638, 6.134), 4.98 billion RMB (95% confidence interval: 2.993, 6.936), 4.922 billion RMB. (95% confidence interval: 2.948, 6.87) and 3.984 billion RMB (95% confidence interval: 2.379, 5.575), which is the same trend as that of Beijing. The total economic loss also shows a trend of first increasing and then decreasing, with a much higher increase range than that of Beijing. For example, in 2017, it increases by 79.77% compared with 2016, but the economic loss in corresponding year is still lower than that of Beijing. And the total economic loss of Beijing in each year is almost twice that of Tianjin, an important reason being that the population exposed to pollutant is less than that of Beijing. The death-related economic loss in Tianjin accounts for more than 98% in each year, even slightly higher than that in Beijing. Therefore, reducing the ozone concentration will reduce the proportion of death-related economic loss in Tianjin more than that in Beijing. In addition, it can be seen from Table 8 that the economic losses of hospitalization for cardiovascular and cerebrovascular disease-related and respiratory system disease-related in the two cities in each year are not significantly different.

Comparing the indirect economic losses caused by not achieving the national air quality first-level concentration limit for PM_2.5 and ozone concentrations, it can be found that the total economic losses caused by PM_2.5 in Beijing show a fluctuating downward trend, and the loss value in 2021 is lower than that in 2016. Although the economic losses caused by ozone have also shown a downward trend in recent years, they have been increasing in the early years, and the economic loss in 2021 is higher than that in 2016, indicating that the economic losses caused by ozone have gradually exceeded those caused by PM_2.5 in recent years. Notably, although reducing PM_2.5 yields greater health benefits, ozone pollution generates higher economic losses in Beijing. This finding underscores the urgency of prioritizing ozone emission controls alongside PM_2.5 mitigation strategies. For Tianjin, the health impacts related to PM2.5 are still more than those related to ozone, but except for 2016, the economic losses caused by ozone are higher than those caused by PM_2.5, and the difference between the two shows a increase trend in corresponding years, that is, the economic losses caused by ozone in Tianjin account for an increasingly high proportion of the economic losses caused by all air pollutants.

Conclusions

• (1) From 2016 to 2021, the health impacts of reducing PM_2.5 concentration in Beijing and Tianjin on all health terminals show a downward trend, with average annual decrease rates of 20.59% and 15.34%, respectively, while the health impacts of reducing ozone concentration on all health terminals show a fluctuating downward trend, with an increase in the middle years. Due to the differences in pollutant concentrations, exposed populations and baseline incidence rates between the two cities in different years, the impact of PM_2.5 or ozone pollution on each health terminal is significantly different. Among them, the orders of the health impacts of reducing PM_2.5 concentration on all health terminals are: respiratory system disease-related hospitalization > cardiovascular and cerebrovascular disease-related hospitalization > all-cause death > cardiovascular and cerebrovascular system-related death > respiratory system-related death. And the health impacts of reducing ozone concentration on all health terminals are ranked as follows: respiratory system disease-related hospitalization > all-cause death > cardiovascular and cerebrovascular system-related death > cardiovascular and cerebrovascular disease-related hospitalization > respiratory system-related death, and reducing the concentration of both pollutants can lead to more than 80% of hospitalization-related health impacts. Therefore, the main direction to improve the health impacts of the two cities is to reduce the number of hospitalizations for cardiovascular and cerebrovascular disease and respiratory disease. These findings align with global patterns observed in studies such as Lelieveld et al.³⁵, who emphasized hospitalization reductions as a critical metric for air pollution mitigation, and Xie et al.¹⁰, who highlighted ozone’s growing mortality burden in urban areas. However, our results diverge from European studies³⁶ where PM_2.5 remains the dominant economic driver, suggesting regional variations in pollutant synergies and governance efficacy. To address this, China could adopt integrated strategies akin to the EU’s Clean Air Programme, which combines emission caps with health impact monetization³⁰, while tailoring interventions to local ozone-PM_2.5 dynamics.

• (2) The total economic losses caused by PM_2.5 concentration that not reaching the first-level concentration limit in Beijing show a fluctuating downward trend, while the proportion of the loss value in GDP continue to decline, from 0.226% in 2016 to 0.109% in 2021, with an average annual decrease of 13.28%. The total economic losses caused by ozone first rise to a maximum of 11.333 billion RMB in 2019, and then decrease to 7.825 billion RMB in 2021. Its proportion to GDP has the same trend of change, which is higher than the proportion of economic losses caused by PM_2.5 in GDP in the same year. In addition, the economic losses caused by two pollutants in Tianjin are basically the same as that in Beijing. The proportion of total economic losses caused by PM_2.5 to GDP has decreased from a fluctuation of 0.235% in 2016 to 0.146% in 2021, with an average annual decrease of 8.57%. Like Beijing, the economic losses caused by ozone are significantly higher than that caused by PM_2.5. Although the economic loss caused by PM_2.5 and ozone pollution has been continuously decreasing in recent years, the loss value is still relatively large, and the proportion of economic loss caused by ozone in the total loss is becoming higher and higher. Therefore, it is necessary to consider the collaborative governance of PM_2.5 and ozone pollution, in order to reduce the economic loss caused by air pollution to a greater extent. This aligns with the"co-benefits"framework proposed by Vandyck et al.³, where dual-pollutant strategies yield compounded health-economic advantages. However, our ozone-driven losses contrast with findings from India¹⁴, where PM_2.5 dominates, underscoring the need for localized policy frameworks. Internationally, cities like Los Angeles and Mexico City have successfully reduced ozone through VOC-NO_x ratio adjustments⁴, offering actionable models for Beijing-Tianjin to balance industrial activity with emission controls.

• (3) Using a Poisson regression relative risk model, this study quantified PM_2.5 and ozone-related health economic losses in Beijing and Tianjin (2016–2021). A potential limitation is that hospitalization costs are estimated by scaling national averages to local income levels, which may introduce uncertainty. In addition, this study only considers five health terminals: all-cause death, cardiovascular and cerebrovascular system-related death, respiratory system-related death, cardiovascular and cerebrovascular disease-related hospitalization and respiratory system disease- related hospitalization, without fully considering other health terminals that may be affected by high concentrations of PM_2.5 and ozone, which may lead to estimated values of economic losses is lower. This study calculates the corresponding 95% confidence interval for each health terminal to better characterize the uncertainty range of economic losses caused by PM_2.5 and ozone, and providing a basis for scientific prevention and control of air pollution. Nevertheless, discrepancies persist with meta-analyses like⁴¹, who reported lower ozone mortality coefficients, suggesting future models could benefit from hybridizing concentration–response functions across climatic zones.

Future directions

• (1) This study only considers the impact of pollutant concentration values on the health economic losses, without taking into account the varying spatial distributions of pollutants in different regions. The estimated loss values can only represent the average level within the region. In the next step, we will conduct research that integrates the temporal and spatial distributions of pollutants in order to estimate more accurate economic losses.

• (2) While this study focuses on Beijing-Tianjin, future work will expand to Hebei Province to address three critical gaps: (a) Contrasting industrial emission profiles to refine the"pollutant-economic loss elasticity"framework. (b) Quantifying cross-regional spillovers using atmospheric transport models CMAQ, informed by the EU’s transnational pollution compensation mechanisms³⁶. (c) Integrating rural health data to assess urban–rural disparity—a dimension absent in current Beijing-Tianjin-Hebei studies but vital for China’s“common prosperity”goals.