Comparative Analysis of PM_2.5- and O₃-Attributable Impacts in China: Changing Trends and Driving Factors

Abstract

China’s divergent fine particulate matter (PM_2.5) and surface ozone (O₃) pollution trends pose critical threats to sustainable development. This study quantifies the spatiotemporal evolution of health burdens (premature deaths) and economic costs across 333 cities during 2015–2023, integrating the Global Exposure Mortality Model (for PM_2.5) and Log-linear Exposure-Response Model (for O₃) with income- and age-adjusted Value of Statistical Life. The results revealed an 11% decrease in PM_2.5-attributable premature deaths, but this benefit was partially offset (60%) by an 87% increase in O₃-related deaths. Furthermore, the per capita economic loss from O₃ exposure increased by 154%, far exceeding China’s 79% growth in per capita disposable income. Decomposition analysis revealed that while diverging exposure levels primarily drove differential PM_2.5- and O₃-related impacts, this disparity was significantly amplified by population aging. These findings underscore the need for air quality strategies to both sustain PM_2.5 reduction achievements and implement rigorous O₃ controls, while integrating pollution considerations into public health frameworks with special emphasis on protecting vulnerable populations.

1. Introduction

As fine particulate matter (PM_2.5) and surface ozone (O₃) are two major air pollutants sharing multiple common precursors, comparative studies of these two pollutants have been extensively conducted worldwide, examining both pollution characteristics [1,2] and their differential health consequences [3,4]. The severe PM_2.5 and O₃ pollution in China has made it a global research priority for these co-occurring pollutants [5,6].

Over the past few decades, China’s PM_2.5-prioritized air pollution control policies have yielded contrasting trends in PM_2.5 and surface O₃ concentrations [7,8]. As documented in the 2023 China Report of the Lancet Countdown on Health and Climate Change, the annual mean PM_2.5 concentration decreased by 41%, from 49 μg/m³ in 2015 to 29 μg/m³ in 2022, whereas the annual daily maximum 8-h average O₃ concentration increased from 135 μg/m³ to 145 μg/m³ [9]. Xiao et al. [10] and Lyu et al. [11] reported similar findings. Given these divergent trends, a systematic assessment of related health and economic burdens is imperative to evaluate the efficacy of past regulatory measures and inform future policy formulations advancing Sustainable Development Goals.

Both PM_2.5 and O₃ pollution pose significant threats to public health [12,13]. Epidemiological evidence has established robust associations between PM_2.5 exposure and adverse health outcomes, including lower respiratory infections (LRI) and all non-communicable diseases (NCD), and even impaired psychological well-being [14,15]. Exposure to O₃ pollution is also believed to increase the risk of death from chronic respiratory diseases (RESP) and circulatory diseases (CIRC) [16,17]. Globally, ambient air pollution accounted for approximately 9.0 million premature deaths in 2019, with 4.14 million attributable to PM_2.5 and 0.37 million to O₃ [18]. China bears a disproportionate burden, experiencing more severe health losses from these pollutants than any other nation [19,20]. For example, Geng et al. [21] estimated 2.12 million premature deaths in China in 2017 due to PM_2.5 exposure, while Yao et al. [22] reported 0.31 million premature deaths attributable to O₃ pollution.

In view of the divergent evolution of PM_2.5 and O₃ pollution, some forward-looking scholars have conducted comparative analyses of their respective health impacts. Dang and Liao [23] found that rising O₃ concentrations in eastern China during 2012–2017 contributed to 16.0 thousand additional respiratory deaths, a figure substantially outweighed by the 284.3 thousand deaths averted through PM_2.5 mitigation. Lyu et al. [11] estimated the long-term population exposure risks to PM_2.5 and O₃ in China’s major urban agglomerations from 2015 to 2021, finding that while O₃-induced premature mortality remained significantly lower than those from PM_2.5, future projections indicate that O₃-related mortality will increase in major urban areas, in contrast to the ongoing decline in PM_2.5-related mortality during 2020–2060. Similarly, Xue et al. [24] tracked spatiotemporal variations in pollution-related mortality from 2013 to 2020, observing a decline in PM_2.5-attributable deaths (from 1.31 million in 2013 to 1.06 million in 2020) alongside fluctuating O₃-attributable mortality (from 0.102 million in 2013 to 0.116 million in 2020).

Beyond health detriments, air pollution imposes substantial economic costs, encompassing medical expenditures, hospitalization outlays, and lost labor productivity [25,26]. Current research predominantly focuses on economic losses caused by either PM_2.5 or O₃ pollution in isolation. For instance, in terms of PM_2.5-related studies, Guan et al. [27] reported China’s economic losses increased marginally from 3205.1 to 3344.8 billion CNY (2015–2017), while Yang et al. [28] observed a decrease from 121.9 to 107.2 billion USD (2014–2016). Regarding O₃ impacts, Maji et al. [29] assessed seasonal variations and associated economic impacts in 338 Chinese cities (2016), and Zhang et al. [30] analyzed spatiotemporal variations in ozone-related economic losses across China (2015–2018). Only a handful of studies have conducted comparative analyses of economic impacts associated with both pollutants [31,32].

Additionally, a growing body of research has focused on identifying drivers behind long-term trends in pollution-related impacts. In terms of health effects, pollution exposure levels constitute the most direct determinant of premature mortality attributable to air pollution [33,34]. Demographic factors, particularly population size and age structure, also significantly influence these outcomes. Notably, the exacerbating effect of population aging on pollution-related health burden has gained increasing recognition in recent studies [35,36]. According to prevailing health impact assessment models based on epidemiological cohort studies, the premature mortality caused by pollution is further modulated by baseline mortality rates [32,37]. For economic impacts, additional drivers come into play beyond the aforementioned factors. Economic growth, industrial restructuring, energy intensity and mix, as well as healthcare investments have all been demonstrated to shape the long-term evolution of pollution-related economic costs [21,38]. Recent multi-driver analyses reveal that while emission control measures reduced PM_2.5 concentrations, concurrent economic expansion and aging populations partially offset these gains in economic burden reduction [21,36].

While existing studies have established a fundamental understanding of PM_2.5 and O₃ pollution variations and their health impacts in China, several critical gaps remain unresolved. First, although pre-pandemic research systematically characterized these pollutants’ distinct trends, the COVID-19 pandemic has dramatically altered pollution patterns through changes in industrial activity, transportation behaviors, and energy consumption [6,39]. This paradigm shift necessitates a fresh assessment of both pollutants’ dynamics and impacts in the post-pandemic era. Second, current research lacks comparative analysis of economic impacts between PM_2.5 and O₃ pollution. Given their differing chemical formation pathways and control requirements, quantifying their differential economic burdens—including healthcare costs, productivity losses, and mitigation expenditures—is crucial for policy prioritization [32,40]. Per capita disposable income (PDI) could be used as a key metric to evaluate these economic disparities in the comparative analyses of economic impacts between PM_2.5 and O₃ [27]. Third, while broad patterns were identified by previous national/regional analyses, critical inter-city variations may be masked. A city-scale approach could address this limitation by enabling tailored ‘one-city-one-policy’ solutions and providing data support for cross-regional coordinated governance, particularly important for ozone which exhibits stronger regional transport characteristics [41,42]. Furthermore, this multi-pollutant comparative assessment could provide new insights by emphasizing the growing health risks from population aging—the demographic factor amplifying vulnerability to both pollutants’ health effects [35,43].

In this study, these critical gaps were addressed by conducting a comprehensive, city-level analysis of PM_2.5- and O₃-related health and economic impacts across 333 Chinese cities during 2015–2023. The results provide the post-pandemic assessment of both pollutants’ evolving dynamics, while systematically comparing their differential economic burdens using income-adjusted metrics. City-scale estimates reveal previously masked spatial variations, enabling tailored pollution control strategies—particularly for regionally transported O₃. The analysis further quantifies how population aging amplifies dual-pollutant vulnerabilities, offering new insights for coordinated multi-pollutant governance under demographic transition.

2. Materials and Methods

2.1. Data Sources

City-level daily 24-h average PM_2.5, and maximum daily 8-h average (MDA8) O₃ concentrations for 333 Chinese cities during 2015–2023 were obtained from the Data Center of the Ministry of Ecology and Environment of China. In accordance with the Technical Regulation for Ambient Air Quality Assessment (HJ 633-2013) [44], the annual PM_2.5 concentrations were derived by computing the arithmetic mean of daily 24-h averages. For O₃, the annual metric was defined as the 90th percentile of MDA8 concentrations, while the arithmetic mean of the MDA8 values was used to estimate O₃-attributable premature mortality. City-specific PDI data for the study period are sourced from the National Economic and Social Development Bulletin. Age-stratified population data (for individuals aged ≥25 years, segmented into 5-year intervals: 25–29, 30–34, etc.) were provided by Zhang et al. [45]. Missing pollutant concentration and PDI data during the study period were filled using linear interpolation of adjacent-year measurements to ensure temporal continuity.

The estimation of pollution-related premature deaths requires baseline mortality rates for specific health endpoints (see Section 2.2 and Section 2.3). Given the limited availability of city-level mortality statistics, nationally aggregated data from the Global Burden of Disease (GBD) study (available at: http://ghdx.healthdata.org/gbd-results-tool, accessed on 15 December 2023) were employed. The GBD dataset provides comprehensive, age-stratified (5-year intervals for adults) baseline mortality rates in China for relevant health endpoints for the years 2015–2019 (

Table A1. Baseline mortality rates for noncommunicable diseases plus lower respiratory infections.

Age	Year
	2015	2016	2017	2018	2019	2020	2021	2022	2023
25–29	0.0002691	0.0002669	0.0002653	0.0002633	0.0002633	0.0002582	0.0002561	0.0002539	0.0002517
30–34	0.0004580	0.0004553	0.0004598	0.0004620	0.0004594	0.0004615	0.0004627	0.0004636	0.0004646
35–39	0.0007668	0.0007652	0.0007874	0.0008029	0.0007978	0.0008134	0.0008239	0.0008339	0.0008439
40–44	0.0013710	0.0013609	0.0013763	0.0013863	0.0013778	0.0013856	0.0013900	0.0013939	0.0013978
45–49	0.0022204	0.0021768	0.0020894	0.0020135	0.0019933	0.0019145	0.0018517	0.0017899	0.0017281
50–54	0.0035476	0.0034931	0.0034575	0.0034295	0.0034027	0.0033598	0.0033247	0.0032894	0.0032540
55–59	0.0059164	0.0058197	0.0057144	0.0056266	0.0055704	0.0054649	0.0053756	0.0052871	0.0051986
60–64	0.0100285	0.0098390	0.0095723	0.0093564	0.0092687	0.0090138	0.0088121	0.0086119	0.0084117
65–69	0.0166040	0.0163034	0.0158445	0.0154780	0.0153047	0.0148820	0.0145374	0.0141950	0.0138526
70–74	0.0295898	0.0291489	0.0288436	0.0285297	0.0282545	0.0278864	0.0275574	0.0272284	0.0268995
75–79	0.0508796	0.0501269	0.0492242	0.0484667	0.0479103	0.0470419	0.0462821	0.0455222	0.0447623
80+	0.1225677	0.1214715	0.1196981	0.1184987	0.1187482	0.1170132	0.1159520	0.1148908	0.1138296

Sources: Data for 2015–2019 are collected from the Global Burden of Diseases study (http://ghdx.healthdata.org/gbd-results-tool, accessed on 15 December 2023), while the rest are generated by linear extrapolation since they are not available at the time of the analysis.

,

Table A2. Baseline mortality rates for chronic respiratory diseases.

Age	Year
	2015	2016	2017	2018	2019	2020	2021	2022	2023
25–29	0.0000115	0.0000113	0.0000107	0.0000102	0.0000099	0.0000104	0.0000102	0.0000099	0.0000097
30–34	0.0000150	0.0000145	0.0000138	0.0000131	0.0000126	0.0000127	0.0000122	0.0000118	0.0000113
35–39	0.0000202	0.0000198	0.0000193	0.0000186	0.0000178	0.0000182	0.0000177	0.0000173	0.0000169
40–44	0.0000284	0.0000279	0.0000269	0.0000259	0.0000250	0.0000258	0.0000252	0.0000247	0.0000242
45–49	0.0000381	0.0000370	0.0000339	0.0000319	0.0000309	0.0000326	0.0000315	0.0000304	0.0000292
50–54	0.0000540	0.0000525	0.0000499	0.0000477	0.0000462	0.0000464	0.0000448	0.0000433	0.0000418
55–59	0.0000863	0.0000846	0.0000802	0.0000764	0.0000739	0.0000778	0.0000761	0.0000744	0.0000728
60–64	0.0001396	0.0001363	0.0001285	0.0001230	0.0001202	0.0001139	0.0001087	0.0001035	0.0000983
65–69	0.0002498	0.0002420	0.0002268	0.0002160	0.0002095	0.0001968	0.0001861	0.0001755	0.0001648
70–74	0.0005155	0.0005000	0.0004802	0.0004626	0.0004518	0.0004326	0.0004161	0.0003996	0.0003831
75–79	0.0010586	0.0010341	0.0009929	0.0009614	0.0009383	0.0009031	0.0008718	0.0008404	0.0008091
80+	0.0037541	0.0037113	0.0036217	0.0035878	0.0036177	0.0035396	0.0035000	0.0034604	0.0034208

Sources: Data for 2015–2019 are collected from the Global Burden of Diseases study (http://ghdx.healthdata.org/gbd-results-tool, accessed on 15 December 2023), while the rest are generated by linear extrapolation since they are not available at the time of the analysis.

and

Table A3. Baseline mortality rates for circulatory diseases (including diabetes).

Age	Year
	2015	2016	2017	2018	2019	2020	2021	2022	2023
25–29	0.0000856	0.0000845	0.0000832	0.0000810	0.0000796	0.0000801	0.0000789	0.0000778	0.0000767
30–34	0.0001559	0.0001545	0.0001555	0.0001540	0.0001508	0.0001492	0.0001479	0.0001466	0.0001453
35–39	0.0002707	0.0002699	0.0002787	0.0002814	0.0002758	0.0002819	0.0002840	0.0002862	0.0002884
40–44	0.0005092	0.0005038	0.0005090	0.0005083	0.0005000	0.0005019	0.0005005	0.0004991	0.0004978
45–49	0.0008510	0.0008301	0.0007863	0.0007454	0.0007287	0.0006895	0.0006565	0.0006236	0.0005907
50–54	0.0013919	0.0013630	0.0013399	0.0013142	0.0012902	0.0012642	0.0012390	0.0012138	0.0011886
55–59	0.0023602	0.0023151	0.0022591	0.0022018	0.0021594	0.0021047	0.0020532	0.0020017	0.0019502
60–64	0.0041994	0.0041091	0.0039807	0.0038695	0.0038169	0.0036938	0.0035933	0.0034929	0.0033924
65–69	0.0074869	0.0073441	0.0071137	0.0069035	0.0067880	0.0065758	0.0063919	0.0062081	0.0060243
70–74	0.0143836	0.0141570	0.0139686	0.0137667	0.0135847	0.0133757	0.0131769	0.0129781	0.0127793
75–79	0.0264086	0.0260371	0.0255336	0.0250181	0.0246173	0.0241425	0.0236823	0.0232222	0.0227620
80+	0.0691206	0.0686900	0.0677320	0.0668808	0.0668518	0.0659510	0.0653163	0.0646817	0.0640470

Sources: Data for 2015–2019 are collected from the Global Burden of Diseases study (http://ghdx.healthdata.org/gbd-results-tool, accessed on 15 December 2023), while the rest are generated by linear extrapolation since they are not available at the time of the analysis.

). For years with missing data, temporal linear interpolation was adopted to ensure temporal continuity.

Furthermore, five major urban agglomerations: 2+26 Cities, Yangtze River Delta (YRD), Pearl River Delta (PRD), Fen-Wei Plain (FWP), and Cheng-Yu Urban Agglomeration (CYU) (Figure A1), were emphasized in the analysis, given their status as China’s most pollution-burdened regions [46,47].

2.2. Estimating the Premature Deaths Attributable to PM_2.5 Exposure

The premature mortality attributable to PM_2.5 pollution was estimated using the Global Exposure Mortality Model (GEMM), the most recent integrated exposure-response function [14,48]. Given that over 99% of non-accidental deaths in the 41 cohorts (spanning 16 countries, including China) underlying GEMM were attributable to NCD and LRI, these two conditions were selected as the relevant health endpoints for adults (aged >25 years). The relative risk function from the GEMM NCD+LRI model was employed. The age-specific relative risk represents the mortality probability ratio between the observed exposure levels and a theoretical counterfactual concentration threshold below which no adverse health effects are presumed to occur. Within the GEMM NCD+LRI, the city- and year-specific ${RR}_{i,y}$ is defined as:

$${RR}_{i,y,a}=exp\left\{{\displaystyle \frac{\theta \mathit{log}(\frac{z}{\alpha }+1)}{1+exp(-\frac{z-\mu }{v})}}\right\},wherez=max(0,C_{i,y}-C_0)$$

(1)

where $i$, $y$, and $a$ refer to city, year, and age respectively; $\theta$, $\alpha$, $\mu$, and $\nu$ are the age-specific and disease-specific parameters describing the shape of the GEMM function (

Table A4. Fit parameters for the Global Exposure Mortality Model (GEMM).

Age	GEMM Parameters
	θ	Standard Error θ	α	μ	ν
25–29	0.1430	0.01807	1.6	15.5	36.8
30–34	0.1585	0.01477	1.6	15.5	36.8
35–39	0.1577	0.01470	1.6	15.5	36.8
40–44	0.1570	0.01463	1.6	15.5	36.8
45–49	0.1558	0.01450	1.6	15.5	36.8
50–54	0.1532	0.01425	1.6	15.5	36.8
55–59	0.1499	0.01394	1.6	15.5	36.8
60–64	0.1462	0.01361	1.6	15.5	36.8
65–69	0.1421	0.01325	1.6	15.5	36.8
70–74	0.1374	0.01284	1.6	15.5	36.8
75–79	0.1319	0.01234	1.6	15.5	36.8
80+	0.1253	0.01174	1.6	15.5	36.8

); $z$ denotes the difference between the annual average PM_2.5 concentration of target city in year $y$ ($C_{i,y}$) and the counterfactual concentration $C_0$ (a uniform distribution between 2.4 $\mathsf{\mu }\mathrm{g}/{\mathrm{m}}^3$and 5.9 $\mathsf{\mu }\mathrm{g}/{\mathrm{m}}^3$).

Accordingly, the PM_2.5-attributable premature death in city $i$ in year $y$ is:

$$D_{i,y}^{PM}={\sum }_{a=25}^{80+}(B_{y,a}^{NCD+LRI}\times P_{i,y,a}\times {\displaystyle \frac{{RR}_{i,y,a}-1}{{RR}_{i,y,a}}})$$

(2)

where $D_{i,y}^{PM}$ denotes the number of premature deaths attributable to PM_2.5 in city $c$ and year $t$; $P_{i,y,a}$ is the city- and year-specific exposed population at age $a$; $B_{y,a}^{NCD+LRI}$ is the year- and age-specific Chinese national baseline mortality rate for NCD plus LRI.

To estimate the 95% confidence interval (CI) of the estimated annual premature deaths, the age-specific relative risk of each city was randomly and simultaneously sampled 1000 times using Python version 3.6.2. In each iteration, the annual premature deaths of the corresponding cities were summed to represent the annual total premature deaths of China or the urban agglomerations. The range between the 2.5th and 97.5th percentiles of 1000 iterative estimates was set as the 95% CI.

2.3. Estimating the Premature Deaths Attributable to O₃ Exposure

Epidemiological evidence has identified two principal causes of O₃-attributable premature mortality: chronic respiratory diseases (RESP) and circulatory diseases (CIRC, including diabetes mellitus) [17,29]. The Log-linear Exposure-Response Model (LERM) was first established by Jerrett et al. [16] through large-scale epidemiological studies. Based on their research, Turner et al. [17] demonstrated that the annual mean of the maximum daily 8-h average (AMDA8) O₃ concentration provides a more robust metric for assessing long-term O₃ exposure effects on cause-specific mortality than the summer-only daily 1-h maximum O₃ metric employed in earlier studies. Since then, the LERM has been extensively applied in O₃-related health impact assessments [49]. Especially, it has been used for China-specific analyses at city level [29,34,50]. Accordingly, in this study, the relative risk of death from a specific disease was calculated via the AMDA8 O₃-based LERM, following the approach of Turner et al. [17]:

$${RR}_{i,y}={exp}^{\beta (C_{i,y}-C_0)}$$

(3)

where $C_0$ is counterfactual concentration of O₃, which follows a uniform distribution between 29.1 $ppb$ and 35.7 $ppb$ according to GBD [49]; $\beta$ is the exposure-response coefficient that quantifies the change in disease-specific mortality risk per unit increase in O₃ concentration, derived from the hazard ratio as follows:

$$\beta =lnHR/\mathsf{\Delta }C$$

(4)

where $HR$ is the hazard ration; $\mathsf{\Delta }C$ is the O₃ concentration increment set at 10 $ppb$ following Turner et al. [17]. $HR$ was set to be 1.12 (95% CI: 1.08–1.16) for RESP and 1.03 (95% CI: 1.01–1.05) for CIRC, respectively, which are both controlled for PM_2.5 to minimize double counting of attributable deaths [17]. To convert the monitored O₃ concentration ($\mathsf{\mu }\mathrm{g}/{\mathrm{m}}^3$) into the same metric as $\mathsf{\Delta }C$ and $C_0$, this study assumed that 1.96 $\mathsf{\mu }\mathrm{g}/{\mathrm{m}}^3=$ 1 $ppb$ according to Yin et al. [51].

City-level annul premature deaths related to O₃ pollution were estimated as:

$$D_{i,y}^{ozone}={\sum }_{a=25}^{80+}(B_{y,a}^{RESP/CIRC}\times P_{i,y,a}\times {\displaystyle \frac{{RR}_{i,y}-1}{{RR}_{i,y}}})$$

(5)

where $D_{i,y}^{ozone}$ is the city- and year-specific premature death attributable to O₃ pollution; $B_{y,a}^{RESP/CIRC}$ is the year- and age-specific Chinese national baseline mortality rate of RESP or CIRC (including diabetes).

The 95% CI estimate of annual premature deaths attributable to O₃ pollution in China and major urban agglomerations was similar to that of PM_2.5. $HR$ was randomly sampled 1000 times using Python version 3.6.2. In each iteration, the annual premature deaths of the corresponding cities were summed to represent the annual total premature deaths of China or urban agglomerations, and the range between the 2.5th and 97.5th percentiles of 1000 iterative estimates was set to be the 95% CI.

2.4. Valuing Economic Loss

The monetization of mortality has been operationalized through non-market valuation techniques because of the absence of a functional market for human life [52]. The concept of Value of Statistical Life (VSL) was introduced to establish a theoretical framework for mortality risk valuation, formally defined as the monetary equivalent of marginal mortality risk reduction derived from individuals’ revealed preferences regarding trade-offs between minor risk variations and income adjustments [53,54]. This metric has gained widespread application in quantifying the economic burden of public health deterioration caused by air pollution exposure [27,28,29]. Among the prevalent VSL estimation methodologies, the willingness-to-pay (WTP) approach has been employed extensively. WTP reflects the maximum monetary amount that individuals would expend for marginal reductions in mortality risk [55,56], exhibiting significant variation across age cohorts because of differences in remaining life expectancy, quality-adjusted life years, accumulated wealth, and socioeconomic positioning [57,58]. In contrast to conventional studies that applied uniform VSL estimates across all demographic strata to assess air pollution-related mortality costs, Yin et al. [35] pioneered an age-stratified VSL framework to evaluate the health economic burdens attributable to ambient PM_2.5. Their methodology incorporates per capita GDP as a key parameter in age-specific VSL computations. Nevertheless, in this study, GDP was replaced by disposable income metrics through the application of an income-risk valuation approach, allowing a more robust economic characterization of health loss to be obtained. This refinement aligns with the fundamental economic premise that an individual’s WTP for mortality risk mitigation exhibits a positive elasticity with respect to personal income growth [27]. By integrating the age-stratified VSL framework [35] and income-risk valuation approach [27], the economic losses attributable to PM_2.5 and O₃ pollution can be calculated as:

$${Age\_VSLY}_{i,y,a}={\displaystyle \frac{{VSL}_0\times {\left(\frac{{INCOME}_{i,y}}{{INCOME}_{2019}}\right)}^e\times \frac{{LE}_{y,a}}{{LE}_{mean}}\times \frac{w_a}{w_{mean}}}{{\sum }_{a=age}^T\left[P_{y,a}\times (1+1/{(1+\gamma )}^{T-a-1})\right]}}$$

(6)

$$EL\_{Age\_VSL}_{i,y,a}=D_{i,y,a}\times {\sum }_{t=0}^{T-a-1}{\displaystyle \frac{{Age\_VSLY}_{i,y,a+t}}{{(1+\gamma )}^t}}$$

(7)

where ${Age\_VSLY}_{i,y,a}$ represents the city-, year-, and age-specific value of a statistical life year, and ${EL\_Age\_VSL}_{i,y,a}$ denotes the corresponding economic loss calculated using the age-adjusted value of a statistical life year measure; $D_{i,y,a}$ denotes the premature deaths related to PM_2.5 or O₃ pollution;${VSL}_0$ (1.5 million CNY/person) is the baseline VSL of Chinese residents in 2019 [53]; ${INCOME}_{2019}$ (30733 CNY) is the national PDI of all Chinese residents in 2019 while ${INCOME}_{i,y}$ is the PDI of city $i$ in year $y$; ${LE}_{y,a}$ is the age- and year-specific life expectancy of Chinese and ${LE}_{mean}$ refers to all ages-average life expectancy of Chinese; $w_a$ is the average wealth at age $a$, $w_{mean}$ is the average wealth, derived from Southwestern University of Finance and Economics [59] (

Table A5. Life expectancy and wealth of Chinese.

Age	Life Expectancy (Unit: Year)									Wealth (Unit: CNY)
	2015	2016	2017	2018	2019	2020	2021	2022	2023
25–29	45.15	45.24	45.34	45.45	45.50	45.61	45.70	45.79	45.88	44,053.39
30–34	40.64	40.72	40.83	40.93	40.99	41.09	41.18	41.28	41.37	53,395.49
35–39	36.20	36.29	36.40	36.50	36.55	36.66	36.75	36.84	36.93	53,146.61
40–44	31.85	31.94	32.05	32.16	32.21	32.32	32.42	32.51	32.61	47,160.08
45–49	27.63	27.72	27.84	27.95	28.00	28.11	28.21	28.31	28.40	46,141.21
50–54	23.55	23.63	23.74	23.84	23.89	23.99	24.08	24.17	24.26	38,604.59
55–59	19.64	19.71	19.82	19.92	19.96	20.07	20.15	20.24	20.33	36,518.54
60–64	15.95	16.02	16.12	16.21	16.26	16.35	16.44	16.52	16.60	26,514.66
65–69	12.56	12.62	12.71	12.79	12.82	12.91	12.98	13.05	13.12	18,616.86
70–74	9.50	9.56	9.63	9.69	9.72	9.79	9.85	9.90	9.96	16,804.35
75–79	6.92	6.96	7.03	7.09	7.12	7.18	7.23	7.28	7.34	25,944.57
80+	4.78	4.81	4.87	4.92	4.94	4.99	5.04	5.08	5.12	13,076.67
All ages	67.99	68.12	68.28	68.42	68.53	68.68	68.82	68.96	69.09	42,990.65

Sources: Life expectancy data for 2015–2019 are collected from the Global Burden of Diseases study (http://ghdx.healthdata.org/gbd-results-tool, accessed on 15 December 2023), while the rest are generated by linear extrapolation since they are not available at the time of the analysis. Wealth data are derived from China Household Finance Survey [59].

); $T$ is the age of expected death at age $a$; $\lambda$ is the elastic coefficient of WTP, implying the change of public’s willing to pay when the PDI increases by a certain extent. Referring to Qu et al. [60], $\lambda$ was set to be 0.8 in this study.

As implemented in Equations (6) and (7), this study directly incorporates city-specific and year-specific PDI data to explicitly account for inter-city economic disparities in estimating pollution-related economic losses, thereby capturing both spatial heterogeneity and temporal dynamics in economic valuation.

2.5. The Contributions of Different Driving Factors

The decomposition analysis was used to quantify the relative contributions of key driving factors to variations in both PM_2.5- and O₃-attributable health burdens and their associated economic impacts.

For mortality impacts, four principal driving factors were examined: changes in exposure level (EXP), population aging (AGE), population growth (POP), and reduced baseline mortality rate (MOR). Referring to [61], the changes in pollution-related deaths can be expressed as:

$$∆D={\sum }_{f=1}^4{∆D}_f$$

(8)

where $∆D$ denotes changes in PM_2.5- or O₃-attributable annual premature deaths; $f$ refers to the four driving factors mentioned above; ${∆D}_f$ represents the $f$-driven death changes.

Following the GBD decomposition analysis [61], a stepwise substitution approach, transitioning each factor from its 2015 value to its 2023 counterpart, was employed to incrementally quantify the impacts of changes in each of these four factors on the change in premature mortality through sensitivity analysis. Technically, the decomposition with four factors has 24 decomposition sequences. For instance, one possible sequence to decompose changes between year 2015 to 2016 is expressed as:

$$D_{2015}={EXP}_{2015}\times {AGE}_{2015}\times {POP}_{2015}\times {MOR}_{2015}$$

(9)

$$D_{2015}^{\prime }={EXP}_{2016}\times {AGE}_{2015}\times {POP}_{2015}\times {MOR}_{2015}$$

(10)

$$D_{2015}^{″}={EXP}_{2016}\times {AGE}_{2016}\times {POP}_{2015}\times {MOR}_{2015}$$

(11)

$$D_{2015}^{‴}={EXP}_{2016}\times {AGE}_{2016}\times {POP}_{2016}\times {MOR}_{2015}$$

(12)

$$D_{2016}={EXP}_{2016}\times {AGE}_{2016}\times {POP}_{2016}\times {MOR}_{2016}$$

(13)

where $D_{2015}$ and $D_{2016}$ represent the premature deaths related to pollution exposure in year 2015 and 2016. The contribution of EXP to the change in pollution-related mortality from 2015 to 2016 can be computed by $D_{2015}^{\prime }-D_{2015}$; Similarly, contributions of AGE, POP, and MOR can be computed via $D_{2015}^{\prime \prime }-D_{2015}^{\prime }$, $D_{2015}^{‴}-D_{2015}^{″}$, and $D_{2016}-D_{2015}^{\prime }$, respectively.

For pollution-related economic impacts, the same method was used to quantify the contributions of different driving factors. Especially, an additional factor, i.e., per capita disposable income growth (PDIG), was incorporated into the analytical framework.

It should be noted that the decomposition sequence introduced potential path dependency into the results. Therefore, the mean of all sequences was taken as the final contribution estimate. This comprehensive treatment ensures robust and sequence-invariant attributions of the driving factor influences.

3. Results

3.1. PM_2.5 and O₃ Pollution Exposure and the Relevant Premature Deaths

The implementation of rigorous air quality improvement policies, particularly the Three-Year Action Plan for Winning the Blue-Sky Defense Battle (hereafter referred to as the Three-Year Plan), has yielded significant mitigation of PM_2.5 pollution across China. Empirical data demonstrate a 36% reduction in national PM_2.5 concentrations from 2015 to 2023, with annual mean levels consistently remaining below the World Health Organization (WHO) Interim Target-1 threshold (35 μg/m³) since 2020. The number of cities that achieved compliance with WHO standards exhibited a substantial increase from 74 to 207 during the study period. In marked contrast to the PM_2.5 reduction trajectory, O₃ pollution displayed an opposing temporal pattern. The AMDA8 O₃ concentration across Chinese cities increased from 135 μg/m³ in 2015 to 146 μg/m³ in 2017 and has maintained this elevated level throughout 2019. The enforcement of COVID-19 containment policies in 2020 triggered an immediate decrease in AMDA8 O₃ to 136 μg/m³, which subsequently rebounded to 145 μg/m³ in 2022 owing to the ease of mobility restrictions and industrial resumption. Notably, PM_2.5 improvements and O₃ deteriorations were concurrently recorded in approximately 60% of the monitored cities (encompassing 58% of the national population) during 2015–2023. The spatial and temporal distributions of annual PM_2.5 and AMDA8 O₃ concentrations across 333 Chinese cities are presented in Figure A2, with corresponding statistical descriptors provided in

Table A6. Statistical characteristics for the annual average PM_2.5 concentrations of 333 cities during 2015–2023.

Year	Maximum $(\mathsf{\mu }\mathbf{g}/{\mathbf{m}}^3$)	Minimum $(\mathsf{\mu }\mathbf{g}/{\mathbf{m}}^3$)	Median $(\mathsf{\mu }\mathbf{g}/{\mathbf{m}}^3$)	$\le 35\mathsf{\mu }\mathbf{g}/{\mathbf{m}}^3$ (Number of Cities)	$\le 10\mathsf{\mu }\mathbf{g}/{\mathbf{m}}^3$ (Number of Cities)
2015	118	10	49	74	0
2016	157	11	44	94	0
2017	100	10	42	102	0
2018	116	8	38	139	2
2019	110	7	35	166	6
2020	113	6	32	197	8
2021	94	7	30	216	9
2022	107	6	29	236	10
2023	89	6	31	207	9

and

Table A7. Statistical characteristics for AMDA8 O₃ concentrations of 333 cities during 2015–2023.

Year	Maximum $(\mathsf{\mu }\mathbf{g}/{\mathbf{m}}^3$)	Minimum $(\mathsf{\mu }\mathbf{g}/{\mathbf{m}}^3$)	Median $(\mathsf{\mu }\mathbf{g}/{\mathbf{m}}^3$)	$\le 100\mathsf{\mu }\mathbf{g}/{\mathbf{m}}^3$ (Number of Cities)	$>160\mathsf{\mu }\mathbf{g}/{\mathbf{m}}^3$ (Number of Cities)
2015	202	62	135	27	59
2016	200	74	138	23	59
2017	219	78	146	8	104
2018	215	74	147	6	109
2019	208	82	146	8	102
2020	194	94	136	10	56
2021	197	94	135	7	52
2022	194	91	145	6	90
2023	195	94	142	4	77

.

The divergent pollution trends were accompanied by opposing trajectories in PM_2.5- and O₃-attributable health burdens during 2015–2023. Based on the COVID-19 pandemic timeline and the assessment results, the study period was divided into three distinct phases: 2015–2019 (pre-pandemic), 2020–2021 (pandemic transition), and 2022–2023 (post-pandemic recovery). Figure 1 illustrates the temporal evolution of pollution-related premature mortality at both national and regional scales regional scales across these phases, with particular focus on five major urban agglomerations. Nationally, PM_2.5-associated premature deaths exhibited a consistent decline during the first two stages. The first stage (2015–2019) witnessed a 10% reduction from 2722.8 thousand (95% CI: 2518.4–2923.0) to 2459.4 thousand (95% CI: 2255.6–2661.9), followed by a further 6% decrease to 2308.4 thousand (95% CI: 2107.1–2510.7) in the second stage (2020–2021). Notably, two significant inflection points emerged: approximately 50% of the 2015–2019 mortality reduction occurred between 2017 and 2018, while the 2019–2020 decline coincided with marked improvements in PM_2.5, during the COVID-19 containment measures. The third stage (2022–2023) demonstrated more complex dynamics, with PM_2.5-related mortality firstly decreasing to 2276.6 thousand (95% CI: 2073.9–2481.2) in 2022 before rebounding to 2418.3 thousand (95% CI: 2205.7–2632.5) in 2023, mirroring contemporaneous fluctuations in PM_2.5.

As shown in Figure 1, the temporal evolution of O₃-attributable health burdens demonstrates an inverse relationship with PM_2.5-related impacts. During the first stage (2015–2019), O₃-induced premature mortality exhibited a substantial 60% increase from 202.2 thousand (95% CI: 148.9–256.0) to 326.0 thousand (95% CI: 261.2–388.4), effectively offsetting approximately 50% of the concurrent PM_2.5-related health gains. The pandemic-induced emission reductions during 2020–2021 (second stage) temporarily reversed this trend, yielding a 7% decline in O₃-attributable mortality to approximately 300 thousand cases. However, the post-pandemic economic recovery in 2022 precipitated a rapid resurgence of O₃ pollution, driving the associated premature mortality to 366.0 thousand (95% CI: 297.8–431.4), exceeding pre-pandemic levels. Notably, while O₃ concentrations stabilized in 2023, mortality continued to rise to 377.7 thousand (95% CI: 309.0–443.5), attributable to demographic factors including population aging. The analysis revealed that nearly 60% of the cumulative health benefits achieved through PM_2.5 reduction during 2015–2023 were negated by concurrent O₃ pollution deterioration.

The five major urban agglomerations consistently accounted for 43–49% of China’s aggregate PM_2.5- and O₃-induced premature mortality annually (see Figure 1). Among these regions, the 2+26 Cities exhibited the most severe atmospheric pollution and consequently bore the greatest health burden, representing 16–17% of nationwide pollution-attributable premature mortality, whereas the PRD demonstrated the most favorable air quality conditions and the lowest mortality impacts. With the exception of the FWP, the other four major urban agglomerations maintained a consistent downward trend in PM_2.5 concentrations through 2022, whereas the FWP displayed a marked 39% increase in PM_2.5 pollution levels accompanied by a 15% rise in associated mortality during 2015–2017. Regarding O₃ pollution, all five regions exhibited a distinct triphasic temporal pattern: during the first stage (2015–2019), FWP experienced the most dramatic O₃ pollution escalation with attributable premature mortality surging by nearly 200%, while the 2+26 Cities and PRD regions showed approximately two-fold increases in O₃-related fatalities; in the second stage (2020–2021), despite pandemic-induced O₃ concentration reductions across all regions, the corresponding mortality declines were limited, particularly in the PRD, CYU, and FWP; and during the third stage (2022–2023), CYU demonstrated the most rapid O₃ pollution rebound with a 12% increase in AMDA8 O₃ concentrations, driving a 78% surge in related mortality, while the 2+26 Cities, YRD, and FWP regions experienced increases of 25%, 25%, and 20%, respectively.

3.2. PM_2.5- and O₃-Related Economic Impacts

During the 2015–2023 period, China exhibited contrasting trends between air pollution reduction and associated economic impacts. While PM_2.5 concentrations and related health burdens decreased, the estimated economic losses attributable to PM_2.5-induced premature mortality rose from 389.1 (95% CI: 328.5–447.0) billion CNY to 504.2 (95% CI: 420.1–586.6) billion CNY, representing a 30% increase primarily driven by growth in residents’ disposable income. Although O₃-related economic losses remained substantially lower in absolute terms compared to PM_2.5, their relative increase was markedly more pronounced, with national totals escalating from 20.5 (95% CI: 15.1–25.9) billion CNY to 53.6 (95% CI: 43.9–62.9) billion CNY—an increase exceeding 150%. As illustrated in Figure 2 (top left), the temporal patterns differed significantly between pollutants: PM_2.5-related economic losses grew predominantly during the first (2015–2019, +13%) and third stages (2021–2023, +12%), whereas O₃-related losses showed the most dramatic increase (+90%) in the first stage. The pandemic-induced lockdown period (2020–2021) temporarily moderated these trends, with both PM_2.5- and O₃-associated economic impacts increasing by only 2% during this second stage, benefiting from improved air quality.

A comparative analysis of PM_2.5- and O₃-related long-term economic impacts was conducted, with differential trends relative to per capita disposable income growth visualized in Figure 2 (top right). Nationally, while PDI demonstrated a consistent annual growth of 5–9%, PM_2.5-related per capita economic losses (PELs) reached 324 CNY/person by 2022, exhibiting modest annual fluctuations (−2% to +5%) that significantly lagged PDI growth. In contrast, O₃-induced PELs displayed markedly different dynamics, increasing by over 150% nationally during the study period and substantially outpacing both PDI growth and PM_2.5-related PEL trends. The temporal evolution of O₃-related PELs followed distinct stages: a first escalation from 15 CNY/person to 28 CNY/person during 2015–2019, followed by continued growth to 35 CNY/person by 2022. Notably, both pollutant categories exhibited significant PEL increases in 2023 (PM_2.5: +10%; O₃: +8%), corresponding directly to deteriorating air quality conditions and associated health impacts. This parallel acceleration suggests emerging convergence in pollution-related economic pressures despite their historically divergent trajectories.

The 2+26 Cities and YRD regions consistently emerged as the most severely impacted, collectively representing approximately one-third of China’s total annual pollution-related economic losses. The observed reductions in PM_2.5- and O₃-related PELs across the 2+26 Cities, YRD, FWP, and CYU since 2017 (Figure 2, bottom) coincide with the implementation period of targeted pollution control policies (e.g., the Action Plan and Three-year Plan). Notably, all five urban agglomerations exhibited significantly faster growth in O₃-related PELs compared to PM_2.5-induced losses prior to this period, suggesting potential policy-driven mitigation effects that warrant further investigation. Following the relaxation of COVID-19 containment measures, worsening O₃ pollution and associated mortality increases drove renewed PEL growth in 2022 across the four regions (excluding the PRD). This resurgence was particularly pronounced in CYU, where O₃-related PELs surged by more than 87%, reflecting the region’s heightened vulnerability to the post-pandemic pollution rebound.

3.3. Effects of Driving Factors

A quantitative decomposition analysis was conducted to assess the relative contributions of five driving factors, that is, EXP, AGE, POP, MOR, and PDIG (for economic impacts exclusively), to variations in PM_2.5- and O₃-attributable health and economic burdens across the three study stages. Figure 3 and Figure 4 show the decomposition results.

EXP was the predominant determinant of pollution-associated premature mortality (Figure 3). The reduction in PM_2.5 exposure concentrations during 2015–2019 accounted for a decrease of 502 thousand PM_2.5-attributable premature deaths—nearly double the net mortality change observed during this period. Conversely, elevated O₃ exposure over the same timeframe generated incremental 96 thousand O₃-related fatalities, equivalent to approximately 48% of the 2015 baseline value. During the 2019–2021 stage, EXP maintained the dominant influence, with declining PM_2.5 and O₃ exposures producing mortality reductions of 274 thousand and 35 thousand cases. Subsequently, the resurgence of O₃ pollution exposure during 2021–2023 has contributed to an additional 69 thousand premature deaths.

Population aging represents a demographic shift toward greater proportions of vulnerable age cohorts exhibiting heightened mortality risk when exposed to air pollution, consequently increasing the annual air pollution-induced premature mortality in China. Quantitative analysis revealed that population aging contributed substantially to PM_2.5-attributable mortality, with incremental fatalities of 318 thousand (2015–2019), 195 thousand (2019–2021), and 159 thousand (2021–2023) during the respective study periods (see Figure 3). Notably, during 2021–2023, AGE surpassed EXP as the dominant driver of increasing PM_2.5-related mortality. Similarly, AGE emerged as the secondary influential factor (following EXP) for O₃-attributable mortality, accounting for additional premature deaths of approximately 36 thousand, 20 thousand, and 18 thousand across the respective study periods.

Advances in healthcare infrastructure and medical technologies have contributed to the reduction of baseline mortality rates, thereby substantially mitigating premature mortality attributable to PM_2.5 and O₃ pollution exposure. This phenomenon was quantitatively demonstrated by MOR-driven reductions of 128 thousand and 15 thousand in PM_2.5- and O₃-attributable premature deaths, respectively, during the 2015–2019 period. Furthermore, the decomposition analysis revealed that POP exerted a comparatively minimal influence on pollution-related mortality trends relative to other determinants, particularly during recent years when China’s demographic growth has experienced a marked deceleration.

As illustrated in Figure 4, PDIG emerged as the principal determinant underlying the persistent escalation of PM_2.5-related economic losses during the 2015–2023 period. In the first stage (2015–2019), a net economic loss of 93 billion CNY was generated by PDIG, offsetting 1.4-fold the 66 billion CNY gain from PM_2.5 abatement. The subsequent stage (2019–2021) witnessed PDIG-induced economic losses of 43 billion CNY, which fully neutralized the gains from continued pollution mitigation. By 2023, the PDIG further augmented PM_2.5-related economic losses by 39 billion CNY relative to 2021 levels. This pattern was similarly manifested in O₃-related economic impacts, where PDIG exerted a comparable influence on EXP. In the first stage (2015–2019), the observed 19 billion CNY rise in O₃-related losses was primarily induced by PDIG and EXP. During the second stages (2019–2021), the PDIG contributed an incremental 4 billion CNY in O₃-attributable economic losses, a magnitude precisely offsetting the contemporaneous reduction in economic losses driven by the decreased O₃ exposure level (EXP). Subsequently, the 2021–2023 period was characterized by concurrent EXP-driven (9 billion CNY) and PDIG-induced (4 billion CNY) increases in O₃-related economic losses.

4. Discussion

This study systematically examined the spatiotemporal evolution and underlying determinants of PM_2.5- and O₃-attributable health burdens and the associated economic costs across Chinese cities from 2015 to 2023. Decomposition analysis was performed to quantify the contributions of five determining factors including exposure level, population aging, population size, baseline mortality rates, and per capita disposable income to variations in PM_2.5- and O₃-attrbutable health and economic impacts.

The results demonstrate a general declining trend in annual PM_2.5-attributable premature deaths, decreasing by 11% from 2722.8 thousand cases in 2015 to 2418.3 thousand in 2023. In contrast, O₃-related premature deaths increased by 87% from 202.2 thousand to 377.7 thousand during the same period. From 2015 to 2023, the deteriorating health burden from O₃ pollution offset approximately 60% of the health gains achieved through PM_2.5 reduction. Correspondingly, further analysis revealed a different trend in economic losses between the two different pollutants. O₃-related per capita economic losses surged by over 150% during the study period, nearly doubling the 80% growth in disposable income, whereas PM_2.5-associated losses showed a modest 25% increase. Notably, if it were not for the strict COVID-19 control measures implemented in 2020–2021 which reversed the deterioration trend of ozone pollution [62], PM_2.5 and O₃, as well as the socio-economic impact of the two, may show a more significant difference trend.

In densely populated and heavily polluted urban agglomerations like 2+26 Cities, YRD, and CYU, air pollution control measures were used to prioritize PM_2.5 [63,64], leading to a significant reduction in regional PM_2.5 pollution levels. Accordingly, the annual premature deaths attributable to PM_2.5 exposure in 2+26 Cities, YRD, and CYU decreased by 17%, 17%, and 11%, respectively, from 2015 to 2023. Concurrently, 60% of the Chinese cities (encompassing 58% of the population) have experienced worsening O₃ pollution. This deterioration translated into a 176% national increase in O₃-attributable premature mortality from 2015 to 2023, with particularly severe impacts in FWP (248%), CYU (118%), and 2+26 Cities (110%). Even worse, several major urban agglomerations all exhibited accelerated growth in per capita O₃-related economic losses following the cessation of COVID-19 restrictions, continuing a pre-pandemic trend of escalating impacts.

The variations in health and economic consequences attributable to PM_2.5 and O₃ exposure are driven by multiple factors. Among these factors, exposure level and population aging have emerged as the predominant drivers shaping the temporal patterns of PM_2.5- and O₃-related health impacts. Given the paramount importance of exposure reduction, policymakers should prioritize integrated governance strategies targeting both PM_2.5 and O₃ pollution, with particular emphasis on coordinated reductions of their common precursors especially like volatile organic compounds (VOCs) and nitrogen oxides (NOx). Current projections suggest that the potential benefits of end-of-pipe control measures will be largely depleted by 2030, necessitating more fundamental interventions such as energy-climate policy integration and economic structural reforms [21,65]. Notably, comprehensive sustainable energy transitions aligned with ambitious climate mitigation targets (e.g., the 1.5 °C warming limit) could potentially reduce China’s PM_2.5 exposure levels below 10 μg/m³ (meeting WHO’s interim target-4) by 2060 [66], representing a viable pathway for long-term air quality improvement.

Effective regional air pollution mitigation requires coordinated yet tailored strategies that account for varying pollution characteristics and distinct health-economic impact profiles across cities. Previous research has tried to establish spatial variations in the multi-pollutant formation mechanisms. For instance, PRD and YRD regions were originally characterized as VOC-sensitive O₃ regimes [46,67,68]. However, emerging evidence suggested that recent VOC emission controls may have transitioned the YRD and adjacent areas toward NO_x-sensitive or mixed-sensitivity conditions [69,70]. This evolving sensitivity underscores the need for updated VOC-NO_x-O₃ sensitivity analyses to inform targeted O₃ mitigation policies. Meanwhile, for China’s most heavily polluted regions, including the 2+26 Cities and FWP, where PM_2.5 concentrations remain elevated despite recent improvements, comprehensive air quality management strategies that address both PM_2.5 and O₃ pollution remain critically important. From another perspective, implementing a cross-regional ecological compensation mechanism, whereby municipalities demonstrating concurrent deterioration in both PM_2.5 and O₃ levels provide financial transfers to those showing opposite trends, could simultaneously advance nationwide air quality improvements while addressing regional disparities.

Older individuals demonstrate disproportionate susceptibility to air pollution effects, and coupled with China’s sustained population aging process [35,71,72], this demographic transition has markedly increased both the public health burden and economic losses attributable to air pollution exposure. Results of this study reveal that, during 2021–2023, population aging exerts a driving effect on air pollution-attributable premature deaths that is much stronger than exposure level change. This underscores the urgent need for age-adaptive safeguards, such as targeted air pollution alert systems for elderly communities, to mitigate aging-driven health burdens and associated economic losses. Such interventions are particularly critical in cities where aging rates exceed 20%. Proactive policies aligning with Sustainable Development Goal 10 (Reduced Inequalities) must address the disproportionate vulnerability of aging populations to environmental risks, ensuring equitable health protection amid demographic shifts.

The findings demonstrate strong consistency with previous studies using comparable methodologies while providing new insights. For PM_2.5-attributable mortality, the 2015 estimate of 2.722 million premature deaths in China shows general agreement with Burnett et al. [14] (2.47 million) and Lelieveld et al. [73] (2.20 million), with the moderately higher estimate being attributable to the updated city-scale population data from Zhang et al. [45] incorporating China’s Seventh National Population Census. Ozone estimates exhibit greater variability across studies: the 2017 assessment (325 thousand deaths) closely matches Yao et al. [22] (310 thousand) but differs more substantially from Yang et al. [34] (245 thousand) and Maji et al. [29] (74 thousand), reflecting methodological differences in exposure-response modeling. Regarding pollution-related economic losses, the assessment results are not directly comparable with previous studies due to the novel integration of income-risk and age-adjusted VSL methods in this study—an analytical approach not employed in prior research.

Several limitations of this study should be acknowledged, as with any comparable research. First, potential overlap represents an important limitation in multi-pollutant health impact assessments. For O₃, a two-pollutant LERM adjusted for PM_2.5 was employed, ensuring the assessment results of O₃-attributable premature deaths are not affected by PM_2.5 interference. The potential overlap originates from the GEMM application for PM_2.5-related mortality, where potential O₃ interference could theoretically occur. On the one hand, different health endpoints were used in PM_2.5- and O₃-related mortality estimates. On the other hand, the GEMM, developed through rigorous cohort studies, inherently minimizes O₃ confounding with low-uncertainty parameters resistant to pollutant interference. Therefore, any overlap in health impacts between the two pollutants falls within acceptable limits. Second, while our city-scale income adjustments help mitigate regional disparities, the use of a uniform national baseline ${VSL}_0$ introduces inherent limitations. Although this approach may modestly underestimate economic burdens in high-income cities and overestimate them in less developed regions, these systematic biases preserve the relative rankings across cities. Most critically, as the same ${VSL}_0$ were consistently applied to both PM_2.5 and O₃ calculations, the comparative assessment of their differential economic impacts remains robust. The primary constraint lies in the absolute valuation of losses rather than their inter-city or inter-pollutant comparisons. Similar to ${VSL}_0$, the application of national-level baseline mortality rates introduces inherent uncertainties in city-scale assessments. This methodological approach primarily affects the spatial dimension of our analysis, potentially attenuating inter-city disparities in absolute mortality estimates. However, it does not compromise the validity of temporal trend comparisons or the relative differences between PM_2.5 and O₃ impacts, given the fact that baseline mortality rates generally remain relatively stable temporally (e.g., <7% annual variation based on GBD data used here). Third, while city-level annual average concentrations provide nationally consistent metrics, they inherently mask intra-urban exposure heterogeneity. Specifically, this approach cannot capture: (1) pollution hotspots near emission sources, (2) seasonal peaks in pollutant concentrations (particularly relevant for O₃), or (3) micro-environmental exposure differences (e.g., indoor vs. outdoor, occupational variations). These limitations likely introduce conservative bias in mortality estimates, as subpopulations experiencing higher exposures are statistically diluted in city-wide averages. Moreover, the lag between time of exposure and death is not considered since this study focuses on comparative analysis of PM_2.5- and O₃-attributable long-term impacts where influence of lag effects is negligible.

5. Conclusions

This study systematically quantified the health and economic burdens attributable to PM_2.5 and O₃ pollution across Chinese cities during 2015–2023, while elucidating their divergent trends and underlying determinants. The results revealed an 11% decrease in PM_2.5-related premature deaths, but this benefit was partially offset (60%) by an 87% increase in O₃-related premature deaths. In addition, the per capita economic loss from O₃ exposure increased by 154%, far exceeding China’s 79% growth in per capita disposable income. The findings highlight the need for future air quality management strategies to not only maintain current PM_2.5 reduction gains but also implement rigorous O₃ control measures. Given the significant variations, locally tailored yet coordinated approaches are imperative to address the distinct air quality challenges faced by different cities and regions. Moreover, the compounding effects of air pollution and population aging demand greater research and policy attention to aging-related environmental vulnerabilities.