Skip to main content
NCBI home page
UKPMC Funders Author Manuscripts logoLink to UKPMC Funders Author Manuscripts
. Author manuscript; available in PMC: 2025 Dec 8.
Published in final edited form as: One Earth. 2025 Nov;8(11):101488. doi: 10.1016/j.oneear.2025.101488

Improved global air quality health index reveals ozone and nitrogen dioxide as main drivers of air-pollution-related acute mortality

Wenzhong Huang 1, Tiantian Li 2,3, Pierre Masselot 4, Rongbin Xu 5,6, Antonio Gasparrini 4, Francesco Sera 7, Michelle L Bell 8, Masahiro Hashizume 9, Susanne Breitner 10,11, Shilu Tong 12,13, Haidong Kan 14, Zhengyu Yang 1, Yiwen Zhang 1, Wenhua Yu 1, Pei Yu 1, Shuang Zhou 1, Qinghua Sun 2,3, Jingwei Zhang 2,3, Eric Lavigne 15,16, Joana Madureira 17,18,19, Yue Leon Guo 20,21,22, Vânia Gaio 23,24,25, Shanshan Li 1,*, Yuming Guo 1,26,*; MCC Collaborative Research Network
PMCID: PMC7618329  EMSID: EMS209693  PMID: 41362416

Summary

Ambient air pollutants are leading contributors to global mortality. Despite the well-established risks, most studies have relied on single-pollutant models in limited regions, leaving the combined effects and individual contributions of pollutants unclear, particularly across countries. Here, we integrate daily mortality and air pollutant (nitrogen dioxide [NO2], ozone [O3], fine particulate matter, and sulfur dioxide) data from 482 cities in 12 countries/territories from 1998 to 2021 to assess the joint mortality risks and identify the main contributing pollutant through an air quality health index of multi-pollutant constrained groupwise additive models (AQHI-Multi). AQHI-Multi outperformed commonly used air quality indices in capturing the overall mortality risks. O3 and NO2 were the leading contributors (accounting for over 70% across countries/territories), with O3’s share increasing slightly to moderately in most countries/territories. These findings highlight the need for developing air quality indices using advanced multi-pollutant models and the emerging global significance of targeted control of O3 and NO2.

Graphical abstract.

Graphical abstract

Open in a new tab

Introduction

Air pollution is a significant public health concern, with both short- and long-term exposure leading to a wide range of diseases. It has been estimated that air pollution accounted for more than 6.5 million deaths globally each year,1 affecting cardiovascular health,2 respiratory health,3 and other conditions such as diabetes,4 obesity,5 and neurological disorders.6 Despite the substantial and well-established evidence, air pollution remains a leading contributor to the global burden of disease (BoD).7 Moreover, the air pollution-related BoD is expected to increase due to climate change and an aging population.8,9 It is crucial and urgently needed to develop effective interventions to mitigate the foreseen air-pollution-related BoD and inform the most effective air quality policies.

Accurately conveying real-time health risks of air pollutants and their relative contributions to the public and stakeholders could serve as an effective and important public health strategy in mitigating the air-pollution-related BoD.10,11 The development of an intuitive and interpretable index for non-experts that accurately represents the overall health risks of air pollution and the identification of major hazardous air pollutants for stakeholders are crucial for effective risk mitigation.12 The Air Quality Index (AQI), originally developed by the US Environmental Protection Agency (EPA) in 1976, is now the most widely applied index to inform the general public of short-term risks of ambient air pollution.13,14 However, despite the various guidelines for AQI development in different countries and regions, the AQI is threshold based and calculated as the maximum value of the segmented linear function with fixed breakpoints for each air pollutant (i.e., the maximum individual AQI of each pollutant).15,16 Therefore, AQI does not adequately and accurately account for continuous exposure-response functions and the empirical joint health effects of exposure to multiple air pollutants.16,17 Importantly, for many pollutants, such as particulate matter and ozone, research indicates the absence of a threshold or safe level below which no adverse health impacts are observed.18,19

To address these limitations, the Air Quality Health Index (AQHI), which incorporates the risk estimates of air pollution from empirical epidemiological evidence, was proposed as an improved alternative to the AQI.16 However, the currently used AQHI is constructed using the estimates from a single pollutant model that does not account for the presence of other air pollutants.20 Considering that people are typically exposed to mixtures of air pollutants rather than a single pollutant in reality, an AQHI calculated using co-effects estimated through a multi-pollutant model strategy may offer greater accuracy in representing the health risks. Additionally, employing a multi-pollutant model allows for the assessment of each pollutant’s contribution to the overall risks, which remains largely unknown.21 This will improve our understanding of the health effects of air pollutants across regions characterized by diverse population demographics and varying air-pollution profiles. Such information may also be essential for developing targeted mitigation strategies and informed policy interventions that prioritize reducing the major hazardous air pollutants, thereby effectively and affordably decreasing the air-pollution-related BoD.

Here, we integrated city-level mortality and monitored air-pollution data over the last two decades from 12 countries/territories across four continents, spanning diverse socioeconomic contexts, including the largest developed country (the United States [US]) and the largest developing country (China). We aimed to develop an improved AQHI using an advanced multi-pollutant constrained groupwise additive model (AQHI-Multi) to more accurately capture the overall mortality risks from short-term exposures to ambient air pollutants and to quantify the relative contributions of individual pollutants in different countries/territories. We observed that the AQHI-Multi exhibited a non-threshold, generally strongest and most linear relationship with mortality compared to commonly used air quality indices. Ozone (O3) and nitrogen dioxide (NO2) together contributed over 70% of the AQHI-Multi, with a slight to moderate temporal increase in the relative contribution of O3 in most countries/territories. The findings support the potentially superior accuracy of multi-pollutant model-based indices in representing and conveying the mortality risks of air-pollution mixtures on a multi-country scale and highlight the increasing global priority of implementing targeted controls for O3 and NO2 to reduce the air-pollution-related BoD.

Results

Descriptive statistics

We analyzed 19.8 million deaths from 482 cities within 12 countries/territories (Figure 1 and Table S1). The populations in these cities experienced a broad range of air-pollution levels, with city-specific mean concentration (μg/m3) ranging from 5.3 in Canada to 94.0 in Mainland China for daily average particulate matter with aerodynamic diameter ≤2.5 μm (PM2.5), from 29.2 in Mainland China to 121.2 in Taiwan for daily maximum 8-h average O3, from 4.8 in Japan to 60.8 in Mainland China for daily average NO2, and from 0.9 in Estonia to 61.2 in Mainland China for daily average sulfur dioxide (SO2) (Table S1). Large countries/territories (Mainland China and the US) included more heterogeneous air-pollution levels across cities. Detailed descriptive statistics of the deaths and air pollution for each country/territory are summarized in Table S1. The wide range of air-pollution levels and geographic distribution are illustrative of regions characterized by different climates and socioeconomic development levels, from developed regions in North America and northern Europe to developing areas in Eastern Asia (Figure 1).

Figure 1. Spatial distribution of the 482 cities in 12 countries/territories; the average daily mean concentration of PM2.5, SO2, and NO2; and the daily maximum concentration of 8-h O3 during the data-collection period.

Figure 1

Open in a new tab

NO2, nitrogen dioxide; O3, ozone; PM2.5, particulate matter with aerodynamic diameter ≤ 2.5 μm; SO2, sulfur dioxide.

Mortality risks associated with ambient air pollutants

Figure S1 shows the overall lag-response pattern of the mortality risks associated with PM2.5, SO2, NO2, and O3 in the week after exposure (lag 0–7 days). The mortality risks increased consistently after exposure to higher levels of PM2.5, SO2, NO2, or O3, generally peaking in the first 2 days after exposure (lag 0–1 day) and persisting up to 3 days (lag 0–2 days). As the elevated mortality risks tended to be statistically insignificant and minimal 3 or more days after exposure, in subsequent analysis we applied a 3-day moving average for PM2.5, SO2, NO2, or O3 concentration, which represented the average of the same and the previous 2 days (lag 0–2 days), to capture the short-term adverse effects of air pollution. Figure 2 shows overall cumulative exposure-response curves of mortality risks with each air pollutant. As the PM2.5, NO2, and SO2 exposure levels increased, the mortality risks demonstrated a supra-linear and monotonic increase. These curves remained similar after excluding cities without data on all-cause mortality (Figure S2) or after imputing missing air-pollution data (Figure S3). Nearly identical results were also observed when data after the year 2019 were excluded (Figure S4).

Figure 2. The overall exposure-response curve of mortality risks, i.e., percentage change in mortality, with each of the air pollutants PM2.5, SO2, NO2, and O3.

Figure 2

Open in a new tab

The shaded area indicates the 95% confidence interval (CI). NO2, nitrogen dioxide; O3, ozone; PM2.5, particulate matter with aerodynamic diameter ≤ 2.5 μm; SO2, sulfur dioxide.

Associations of various air quality indices with mortality

The overall and country/territory-specific utility of various types of indices of air quality is depicted in Figure 3. Overall, AQHI based on a multi-pollutant model (i.e., AQHI-Multi) represented the highest mortality risk, demonstrating the strongest association with unobserved mortality (i.e., mortality outcomes in the separate test dataset), followed by AQHI based on a single-pollutant model (i.e., AQHI-Single), the US version of AQI (AQI-USA), the China version of AQI (AQI-CHN), and the European Union version of AQI (AQI-EU). At the country level, the relatively higher utility of AQHI-Multi was also observed in most study countries/territories, especially for the US, Mainland China, Canada, Portugal, and Japan, compared to other indices of air quality. The utility of the different indices of air quality was robust to models with various specifications (Figure S5) or using different proportions of training data (Figure S6). Similarly, the index-response curves showed that AQHI-Multi had the strongest, most of the study countries/territories. Notably, the relative contribution weight of O3 showed a slight to moderate increase over time in most of the study countries/territories, with the most pronounced increases observed in Mainland China, Canada, Spain, and Estonia.

Figure 3. The overall and country/territory utility of AQI-USA, AQI-EU, AQI-CHN, AQHI-Single, and AQHI-Multi, presented as the percentage change in mortality with 95% CI, for each interquartile increase in the index.

Figure 3

Open in a new tab

The utility was examined in testing data using the parameters estimated based on training data. The early 70% of the data were selected as the training data and the remaining 30% as the testing data for each city. AQI-USA, the US Environmental Protection Agency’s air quality index; AQI-EU, the European Environment Agency’s air quality index; AQI-CHN, the Chinese Ministry of Ecology and Environment’s air quality index; AQHI-Single, the air quality health index based on the single-pollutant model; AQHI-Multi, the air quality health index based on the multi-pollutant CGAIM; CGAIM, constrained groupwise additive index model; CI, confidence interval.

Discussion

Despite the well-established elevated most linear, and non-threshold relationship with mortality risk compared to the other indices (Figure 4). As the index value increased, the air-pollution-related mortality risks rose proportionally for AQHI-Multi. In contrast, the mortality risks tended to level off or even decrease with the increase of AQHI-Single and AQIs.

Figure 4. The index-response curve of various air quality indices with mortality risks.

Figure 4

Open in a new tab

The shaded area indicates the 95% CI. The curves were fitted using the indices in testing data constructed based on the parameters derived from the training data. The early 70% of the data were selected as the training data and the remaining 30% as the testing data for each city. AQI-USA, the US Environmental Protection Agency’s air quality index; AQI-EU, the European Environment Agency’s air quality index; AQI-CHN, the Chinese Ministry of Ecology and Environment’s air quality index; AQHI-Single, the air quality health index based on the single-pollutant model; AQHI-Multi, the air quality health index based on the multipollutant CGAIM; CGAIM, constrained groupwise additive index model; CI, confidence interval.

Spatiotemporal patterns of AQHI-Multi and pollutant contributions

The final constructed index of AQHI-Multi for each country/territory during the study period is shown in Figure 5, presented as the monthly number of days in each air quality risk category based on the AQHI-Multi value. The risk parameters for calculating the AQHI-Multi in each study country/territory are shown in Table S2. The number of low-risk days (i.e., AQHI-Multi ≤3) in a month generally decreased in Switzerland and Portugal during the study period while increasing in Mainland China, Mexico, and Taiwan (Figure 5). The trends remained relatively constant over time in the US, Canada, Germany, Spain, Japan, Australia, and Estonia. The relative contributions of each air pollutant (PM2.5, O3, SO2, and NO2) to the AQHI-Multi in each country/territory are depicted in Figure 6. The specific contributions to an elevated AQHI-Multi varied across air pollutants and countries/territories. Generally, O3 and NO2 were the leading air pollutants that contributed to a higher value of AQHI-Multi in different countries/territories, followed by PM2.5 and SO2. The contributions of O3 were particularly high in Canada, Japan, Australia, Estonia, and Portugal, accounting for more than 50% of the air-pollution-related joint mortality risks represented by AQHI-Multi. In contrast, NO2 was the main air pollutant driving the increase in AQHI-Multi in Germany and Switzerland. The contributions of SO2 were consistently and comparatively limited across countries.

Figure 5. Temporal evolution of the monthly number of days of each air quality risk category based on AQHI-Multi in each study country/territory during the study period.

Figure 5

Open in a new tab

The health risks from air pollution in a day were classified into four categories based on the AQHI-Multi values: low (1–3), moderate (4–6), high (7–10), and very high (>10). The dashed lines represent the linear trend. AQHI-Multi, the air quality health index based on the multi-pollutant CGAIM; CGAIM, constrained groupwise additive index model.

Figure 6. Overall relative contribution weight in percent of PM2.5, SO2, NO2, and O3 to AQHI-Multi in each country/territory over the study period.

Figure 6

Open in a new tab

AQHI-Multi, the air quality health index based on the multi-pollutant CGAIM; CGAIM, constrained groupwise additive index model; NO2, nitrogen dioxide; O3, ozone; PM2.5, particulate matter with aerodynamic diameter ≤ 2.5 μm; SO2, sulfur dioxide.

When viewed on a temporal scale, the contributions of various air pollutants to AQHI-Multi showed seasonal variations (Figure S7). Generally, relatively higher contributions of O3 and NO2 were observed in summer and winter, respectively, in mortality risks associated with short-term exposures to air pollutants, their joint mortality risks and respective contributions remain unclear, especially across diverse geographic regions. This reduces the accuracy of quantifying and communicating overall real-world mortality risks, constrains mechanistic understanding of pollutant-specific health impacts, and impedes the formulation of evidence-based, targeted mitigation strategies. Based on multi-country data and an advanced multi-pollutant model, we developed an improved air quality index (i.e., AQHI-Multi) by incorporating empirical population risks from co-exposures to air pollutants to more accurately present and convey the overall mortality risks associated with ambient air pollution. AQHI-Multi exhibits a higher utility in representing the overall mortality risks associated with air-pollution exposure compared to the currently used AQIs and AQHI based on the single-pollutant model or considerations of multiple pollutants without accounting for their joint effects. The risk contributions of PM2.5, SO2, NO2, and O3 varied across countries/territories. Generally, O3 and NO2 were the leading air pollutants that contributed most to AQHI-Multi across countries/territories. A slight to moderate increase in the relative contribution of O3 was observed over time in most of the study countries and territories.

Air-pollution warning indicators serve as an important tool for communicating health risks to the public and mitigating the health burden from air-pollution exposure.14 However, the currently most widely used air-pollution indicator, the AQI, is calculated solely based on the air-pollution concentrations. This does not explicitly account for the population vulnerability. Specifically, populations with different characteristics exhibit varying mortality risks even when exposed to the same pollution levels. To date, studies on the development of air-pollution risk indicators remain scarce, especially on a multi-country scale.14,22 By considering both air-pollution levels and population vulnerability, the AQHI was developed as an improved alternative to the AQI through the inclusion of health risk estimates of air pollutants derived from epidemiological evidence. Further, people are exposed to multiple pollutants rather than a single air pollutant in real life. In this study, we improved the AQHI by considering the joint effects of air pollutants using an advanced multi-pollutant model. These improvements were evidenced by our validity analysis, where AQHI-Multi exhibited higher utility in presenting the air-pollution-related mortality risks, followed by AQHI-Single and AQIs. A non-threshold and approximately linear index-response relationship was exclusively observed for AQHI-Multi in the test dataset, in contrast to other indices. This suggested that AQHI-Multi was a more sensitive indicator that accurately reflected the health risks equivalent to its value at different levels of air pollution. By comparison, AQHI-Single and AQIs may lack effectiveness in accurately conveying the risks of air-pollution-related mortality, particularly during periods of low or high air-pollution levels. Our findings are consistent with previous studies on AQHI establishment that were largely based on a single-pollutant model or restricted to a single area or region in Mainland China.20,23–25 Similarly, these studies also suggested a stronger association of AQHI with health outcomes compared to AQI. This study advanced previous AQHI research by moving beyond single-pollutant models and geographically limited analyses, employing a multi-pollutant model across diverse countries and territories that could quantify both joint mortality risks and individual pollutant contributions more accurately by accounting for previously overlooked potential non-linear and combined effects. These advances provide a more comprehensive and accurate basis for health risk communication and the development of effective response strategies for ambient air pollution.

Based on the improved approach for AQHI constructions, we calculated AQHI-Multi and characterized the temporal trends in air-pollution-related joint mortality risks, as well as the contributions of different air pollutants across various countries and territories. The diverse patterns observed in these trends indicated great variations in the temporal changes of air-pollution-related mortality risks across different countries/territories. Currently, evidence on the temporal trends in the overall mortality risk from air-pollutant mixtures from these countries/territories remains limited. This constrains the ability to detect changes or improvements in public health outcomes over time, undermines the assessment of long-term effectiveness of air quality policies, and hampers timely identification of emerging risks from pollutant mixtures. We found that the monthly number of low-risk days (i.e., AQHI-Multi ≤3) exhibited the most significant increase in Mainland China. This highlights the great significance of China’s devoted efforts during the past decade since the introduction of the “Atmosphere Ten Articles” policy in 2013 to dramatically reduce the air-pollution level, which could prevent hundreds of thousands of annual attributable premature deaths.26

We further examined the relative contributions of each air pollutant to the mortality risks associated with short-term exposure to air-pollutant mixtures, as represented by the AQHI-Multi, in various countries. Despite the different magnitude of contribution to the mortality risks for various air pollutants in different countries/territories, O3 and NO2 made a prominent contribution to the AQHI-Multi in most of the study countries/territories. By comparison, the contribution of SO2 was consistently limited across different study countries/territories. While there is substantial evidence to support the adverse associations of air pollution with human health, studies with a quantitative assessment of the specific contribution of each air pollutant compared to other pollutants are scarce, especially on a multi-country scale.27,28 Previous studies on health risk assessment of air pollutants were largely based on a single- or two-pollutant model that did not consider the co-existence of multiple pollutants, partially to avoid multi-collinearity due to the high correlations among air pollutants.20,21 The indices developed by such approaches likely bias population health risk estimates, thereby reducing the accuracy and effectiveness of public health risk assessment and communication and potentially leading to suboptimal or misdirected air quality interventions. By comparison, multi-pollutant models such as constrained groupwise additive index models (CGAIMs) can account for the potential multi-collinearity and provide the joint risk estimate of co-exposures to multiple pollutants as well as the relative weight of each pollutant.29 Leveraging an advanced multi-pollutant CGAIM, the developed AQHI-Multi advances beyond prior single- or limited-pollutant risk assessments by more accurately quantifying the relative importance of each pollutant from a perspective of public health. This enhanced understanding supports the development of targeted air quality management strategies to cost-effectively reduce the substantial disease burden currently attributable to air pollution. Based on the real-time informed index of AQHI-Multi and the relative contribution weight from each air pollutant, limited resources can be prioritized based on the pollutants that currently contribute most significantly to health risks.

Our findings were generally consistent with our previous investigation in a metacity (Guangzhou) in Mainland China, which also used a multi-pollutant model and detected that O3 and NO2 were the major air pollutants contributing to elevated morbidity (outpatient visits and hospital admissions) and mortality risks.20 The current study supports this finding on a multi-country scale and highlights the leading role of O3 and NO2 in contributing to air-pollution-related acute mortality in different countries/territories with diverse geographic and sociodemographic contexts. Furthermore, compared with the previous multi-pollutant model-based AQHI establishment approaches (e.g., Bayesian multi-pollutant models),20,23,30–32 the AQHI-Multi developed using CGAIM in this study enables intuitive interpretation and flexible integration of operational requirements and prior knowledge in a computationally efficient manner. This makes it more adaptable for local health authorities and environmental agencies to develop region-, population-, and outcome-specific air quality health indices while avoiding the computationally intensive procedures (e.g., Markov chain Monte Carlo sampling). The observed dominant role of O3 and NO2 could be partly attributed to the highly reactive and potent oxidant properties as well as their typically higher concentrations in urban environments, especially in developed regions.33,34 These factors can provoke more immediate inflammatory responses compared to PM2.5 and SO2, including the triggering of asthma exacerbations and acute cardiovascular events.35–38 Furthermore, we observed that the relative contribution of O3 to AQHI-Multi and air-pollution-related premature mortality has increased over time in most of the study countries and territories. This trend could be mainly driven by the increasing level of O3 exposure, especially under a warming climate. Climate change has resulted in longer and hotter summers, which extend the ozone season and increase peak O3 concentrations.39 Targeted strategies need to be developed to reduce the pollution levels of O3 and NO2 to more effectively and affordably mitigate the heavy disease burden associated with air pollution.

This study had multiple strengths. By using a large population-based dataset across countries/territories, we were able to establish an improved AQHI that robustly captures the mortality risks associated with complex air-pollutant mixtures across regions with diverse geographic and sociodemographic contexts and to examine its utility and applicability in different settings. For outcome data, we used 19.8 million mortality observations with relatively high spatiotemporal resolution (city- and daily-level) from a well-established framework of the Multi-City Multi-Country (MCC) network that has been widely used in previous studies.40–45 This provided high statistical power and broad representativeness within a validated epidemiological framework. For exposure data, we used reliable monitoring of air-pollution concentrations from local authorities within each country/territory, thereby reducing local measurement error that could bias risk estimates, in contrast to the modeled exposure data used in most previous studies. For the modeling strategy, we used a well-established time-series model46,47 and an advanced multi-pollutant model,29 which have been illustrated above.

However, some limitations must also be acknowledged. For the US, we only had data from 1999 to 2006, which limited the generalizability of our findings to the current conditions in the US. Additionally, our data collection and analysis were performed at the city level, assuming uniform air-pollution exposures for individuals within each city. This non-differential exposure misclassification tends to systematically bias the risk estimates toward the null, which indicates our results are more likely to be conservative, and the findings on the greater utility of AQHI-Multi compared to other air quality indices are unlikely to be substantially affected.48,49 Furthermore, the AQHI-Multi in the present study was formulated based on mortality data from urban areas, limiting the applicability to inform the morbidity risks (e.g., outpatient visits) and risks for populations in suburban or rural areas. However, considering that AQHIs are commonly applied in urban contexts and mortality data are generally more readily available, this approach enhances the applicability and reproducibility of our findings. Moreover, although we quantified the varying contribution of different air pollutants to health risks across regions, we did not have data on pollutant chemical composition or emission sources, which precluded us from delineating the specific mechanisms underlying these spatial variations (e.g., potential differences in toxicity). Finally, the limited number of study locations in some countries (e.g., Mexico and Germany) restricted the statistical power to differentiate the utility of different health indices of air quality in these countries. However, results from larger countries/territories with many more study locations (e.g., the US, Mainland China, and Canada) consistently supported a higher accuracy of AQHI-Multi in representing the air-pollution-related mortality risks compared to other indices. Therefore, this limitation may be more likely to have contributed to a more conservative interpretation of our findings. Future studies could apply and extend the AQHI-Multi formulation strategy to other countries and health outcomes to validate and complement our findings.

In summary, we found that the mortality risks related to air pollution were more accurately represented by the AQHI constructed using multi-pollutant models (i.e., CGAIMs) than by currently used air quality indices on a multi-country scale. O3 and NO2 were consistently identified as the main pollutants contributing to the short-term mortality risks from ambient air pollution, with the relative contribution of O3 showing slight to moderate increases over the past two decades in most of the studied countries/territories. The results of this study emphasize the urgent need for local health authorities and environmental agencies to integrate advanced multi-pollutant models to more effectively capture and communicate air-pollution-related health risks. In addition, more mitigation efforts and enhanced targeted strategies should be devoted to responding to the emerging health impacts of O3 and NO2.

Methods

Data sources

We obtained health and environmental data from the database of the MCC Collaborative Research Network, which has been described in detail in our previous work.40,42 In brief, daily data on mortality, monitoring of ambient air pollution, temperature, and relative humidity were obtained from local authorities within each country/territory. The current analysis was limited to cities available for four major air pollutants: daily average concentration of PM2.5, SO2, NO2, and daily maximum 8-h average O3 (482 cities in 12 countries/territories, with an overall study period ranging from 1998 to 2021) (Table S1). PM10 and carbon monoxide (CO) were not included because (1) PM10 exhibits a high correlation with PM2.5, and PM2.5 already encompasses the primary adverse effects associated with particulate matter50; and (2) the adverse public health impact of ambient CO is comparatively modest as the concentrations of CO consistently remain well below the air quality standards established in various countries and territories.20,51 Daily PM2.5, SO2, NO2, and daily maximum 8-h average O3 had an overall missing rate of 4.0%, 5.3%, 1.9%, and 2.1%, respectively. Days with missing pollutant measurements were excluded during the analysis. In 23 out of the 482 cities, all-cause mortality data were not available and, instead, mortality from non-external causes (International Classification of Diseases, 9th Revision [ICD-9], codes 0–799 or ICD-10, codes A0−R99) was collected to represent all-cause mortality.40–42

Statistical analysis

Single-pollutant model

In the single-pollutant model, we adopted a standard two-stage analytical framework to estimate the associated mortality risk for each air pollutant (PM2.5, SO2, NO2, and O3).40,42,52 In the first stage, a quasi-Poisson regression distributed lag non-linear model, accounting for the possible overdispersed daily mortality counts, was applied to model the lag-response and exposure-response relationships of air-pollution concentration with mortality in each city and for each air pollutant. To account for the non-linear and delayed effects of air pollution, the air-pollutant concentration was modeled with a distributed non-linear and lag term up to 7 days after exposure, with a natural cubic spline function of two internal knots placed at the 25th and 75th concentration percentiles and two internal knots (plus an intercept) equally placed on the log scale of lag days, respectively.40,53 Natural variations in the mortality (i.e., long-term trends, seasonal and weekly patterns) were adjusted by using a natural cubic spline of time with 7 degrees of freedom (df) per year and including a dummy variable of day-of-week indicator in the model.40,54,55 Potential confounding effects by weather conditions were also controlled by including 8-day moving averages of ambient temperature and relative humidity with natural cubic splines of 6 and 3 df, respectively, in the model.40,56,57 We tested these modeling choices in the sensitivity analysis.

In the second stage, the city-specific effect estimates from the first stage were pooled using random-effects meta-analysis to characterize the overall lag-response and exposure-response relationships between mortality and air pollutants. The pooled risk estimates are presented as the percentage change (with 95% confidence intervals) in daily mortality per 10 μg/m3 increase in air-pollutant concentrations.

Multi-pollutant model

In the multi-pollutant model, we applied a two-stage analytical framework similar to the above single-pollutant analysis. In the first stage, we applied CGAIM to estimate the joint mortality risks of air-pollutant mixture (PM2.5, SO2, NO2, and O3) and to find the optimal weights for each air pollutant that represent potentially adverse effects in each city.58 The development and methodology of CGAIM has been detailed in our previous work.58 In brief, compared to the traditional single-pollutant model, CGAIM includes all exposures of interest simultaneously in the model as a group within which index weights and the potentially non-linear association between the index and health are estimated. Given the complexity of integrating multiple pollutants, CGAIM additionally allows constraints on both the weights and the association to yield meaningful and stable indices that reflect the overall health impact of all exposures of interest. The constraints in CGAIM allow the integration of additional information reflecting prior assumptions about the risk estimates as well as the operational requirements to ensure both identifiability and a better interpretability of the resulting indices. More specifically, we implemented a constraint that all relative contribution weights of air pollutants in the mixture were ≥0 and summed to 1, and that the relationship of the air-pollutant mixture with mortality was monotone increasing.58 The CGAIM was adjusted for the same confounders as those in the single-pollutant model (i.e., natural variations in the mortality and weather conditions).

In the second stage, the effect estimates generated from the CGAIM in each city were pooled using random-effects meta-analysis to obtain overall and country/territory-specific joint risk estimates of the air-pollutant mixture as well as the relative contribution weights of each air pollutant.

Air quality indices formulation

For each country/territory, we calculated the AQHI-Single and AQHI-Multi using the corresponding country/territory-specific effect estimates from the above single- and multi-pollutant models, respectively. Based on previous studies,16,20 we first derived death risk functions (DRFs) related to air pollution using the following formulas:

DRF-Single=100∑k=1p(eβkcXkti−1), (Equation 1)
DRF-Multi=βc∑k=1pWkcXkti, (Equation 2)

where DRF-Single and DRF-Multi represent the DRF calculated based on the risk parameters from the single-pollutant and multi-pollutant models, respectively. In Equation 1 for DRF-Single, βkc denotes the risk estimate of air pollutant k in country/territory c where city i is located, and Xkti denotes the concentration (μg/-m3) of air pollutant k on day t in city i. Therefore, DRF-Single can also be interpreted as the total excess mortality risks associated with air pollutant 1 to p. In Equation 2 for DRF-Multi, the air pollutants are considered as a whole, and βc denotes the risk estimate of the air-pollutant mixture in country/territory c; Wkc refers to the joint mortality risk and relative weight of air pollutant k in country/territory c where city i is located.

To create an understandable and comparable metric and facilitate the risk communications, we further scaled the DRF to index values from 1 to 10+ in a way that the index value of 3 corresponds to the DRF derived for the concentrations where the World Health Organization short-term air quality guideline values in 2021 were met.59,60 DRF-Single and DRF-Multi were scaled to AQHI-Single and AQHI-Multi, respectively. A greater AQHI value indicates higher mortality risks associated with short-term exposure to air pollutants. The health risks from air pollution could be classified into four categories based on the AQHI values: low (1–3), moderate (4–6), high (7–10), and very high (>10).61

For comparisons, we also calculated a series of currently commonly used AQIs according to the guidelines in different countries and regions. Specifically, we calculated the AQI-USA (ranging from 0 to 500+), AQI-EU (0–6), and AQI-CHN (0–300+) based on the guidelines from the US Environmental Protection Agency,62 the European Environment Agency,63 and the Chinese Ministry of Ecology and Environment,64 respectively. Though having different numerical scales compared to AQHIs (0–10+), a higher AQI value also indicates a greater health risk from air pollution.

Air quality indices utility

To check the utility and effectiveness in representing mortality risks of different air quality indices, we conducted a validation analysis based on a train-test split.20,23 Specifically, for each city, we used the early 70% of data as the training set and the subsequent 30% as the testing set. The data were partitioned into two continuous parts based on time, with one preceding the other, because, in practice, the AQHIs were developed using continuous data from a specific historical period and applied to current and future scenarios. The two-stage single- and multi-pollutant analyses were conducted using the training dataset to generate parameters, which were used to calculate AQHI-Single and AQHI-Multi with the testing dataset. AQI-USA, AQI-EU, and AQI-CHN were also calculated with the testing dataset. For each index, we then estimated and compared the percent change in mortality for each interquartile range (IQR) increase in the index, as well as the index-response relationships of various indices with mortality, using the standard two-stage time-series analysis described above.20,23 Specifically, in the first stage, for each city in the test set, we modeled the mortality risks associated with each IQR increase in the index using a quasi-Poisson regression model, with adjustments consistent with those in the main model (i.e., natural variations in mortality, temperature, and relative humidity). We used overall and country/territory-specific IQRs for each index to account for substantial variability in the numerical ranges of the indices and ensure the comparability of the effect estimates across cities and indices. To model potential non-linearity, we conducted additional city-specific models in which the linear index term was replaced with a non-linear term represented by a natural cubic spline function, with two knots positioned at the 25th and 75th percentiles of the index distribution, averaged across all cities in the test set.20,25,65 Then, in the second stage, these city-specific effect estimates were pooled using random-effects meta-analysis to derive overall and country/territory-level results. An index with a higher estimate was considered to be more strongly associated with mortality and to have greater utility in representing and predicting mortality risks.

Sensitivity analysis

To test the robustness of our results, we conducted a series of sensitivity analyses by (1) excluding the cities without all-cause mortality data; (2) imputing missing air-pollutant values using a natural spline function based on other available daily values for each city; (3) excluding the data after the year of 2019 to assess the extent of COVID-19’s potential impact; (4) employing different model specifications, which include altering the df for time (5, 6, 8, or 9 per year) to adjust for temporal trends, df for temperature (3, 4, 5, or 7), the length of the averaging window for temperature and relative humidity adjustments (3 or 21 days), additional adjustment for public holiday, and not adjusting for the relative humidity; and (5) using a different proportion (50%, 60%, 70%, or 80%) of data as the training set.

All analyses were performed using the R language and environment (version 4.1.3),66 with the R packages “dlnm”67 and “cgaim”58 for the single- and multi-pollutant model, respectively. The random-effects meta-analysis was performed using the R package “mixmeta.”47

Resource Availability

Lead contact

Requests for further information and resources should be directed to and will be fulfilled by the lead contact, Yuming Guo (yuming.guo@monash.edu).

Materials availability

This study did not provide any novel materials.

Supplementary Material

Supplemental information can be found online at https://doi.org/10.1016/j.oneear.2025.101488.

Appendix

Science For Society

Ambient air pollution is a leading threat to premature mortality in populations world-wide. However, current risk assessments and communications often overlook the combined exposure to multiple air pollutants. Most existing measures rely on single- or two-pollutant assessments, while a consistent quantification of the joint effects and individual pollutant contributions across countries and regions remains scarce.

Our study shows that air quality indices developed using multi-pollutant epidemiological models could more accurately capture the short-term mortality risks than currently used indices do. Ozone and nitrogen dioxide are the main pollutants driving these risks, with ozone’s impact growing in most of the studied countries/territories over the past two decades. This improved index can help toward better monitoring, understanding, and communication of air-pollution health risks, enabling more targeted strategies to mitigate the associated substantial disease burden.

Highlights.

  • An advanced multi-pollutant model-based air quality health index was developed

  • The index outperforms current indices in capturing acute mortality of air pollution

  • O3 and NO2 are the main pollutants driving the risk index across regions

  • O3 shows rising contributions to the index in most studied countries/territories

Acknowledgments

This work was supported by the Australian Research Council (DP210102076) and the Australian National Health and Medical Research Council (APP2000581). W.H. was supported by China Scholarship Council funds (202006380055). Z.Y. and W.Y. were supported by a Monash Graduate Scholarship and a Monash International Tuition Scholarship. Y.Z. was supported by a NHMRC e-Asia Joint Research Program grant (GNT2000581). R.X. was supported by VicHealth Postdoctoral Research Fellowships 2022. P.Y. was supported by Monash Faculty of Medicine Nursing and Health Science (FMNHS) Early Career Postdoctoral Fellowships 2023. S.L. was supported by an Emerging Leader Fellowship (GNT2009866) of the Australian National Health and Medical Research Council. Y.G. was supported by a Career Development Fellowship (GNT1163693) and a Leader Fellowship (GNT2008813) of the Australian National Health and Medical Research Council. A.G. was supported by the EU’s Horizon 2020 Project, Exhaustion (grant 820655), and Wellcome-funded project BREATHE (grant no. 308914/Z/23/Z). F.S. was supported by the Medical Research Council, United Kingdom (grant no. MR/R013349/1), the Natural Environment Research Council United Kingdom (grant no. NE/R009384/1), and the EU’s Horizon 2020 project, Exhaustion (grant no. 820655). M.H. was supported by the Japan Science and Technology Agency as part of SICORP, grant number JPMJSC20E4. S.T. was supported by the Science and Technology Commission of Shanghai Municipality (grant no. 18411951600), China. The funders had no role in study design, data collection, data analyses, data interpretation, and study report.

Footnotes

Author contributions

Y.G. and S.L. conceptualized the study and contributed to the correspondence work. W.H. and P.M. designed the methodology. W.H. conducted the statistical analysis and took the lead in drafting the manuscript and interpreting the results. T.L., P.M., R.X., A.G., F.S., M.L.B., M.H., S.B., S.T., H.K., Z.Y., Y.Z., W.Y., P.Y., S.Z., Q.S., J.Z., E.L., J.M., Y.L.G., V.G., S.L., and Y.G. contributed to providing the mortality data and revising the manuscript.

Declaration of interests

The authors declare no competing interests.

In brief, Despite well-established risks from air pollution, consistent quantification of joint health risks and individual pollutant contributions across countries/territories remains lacking. Using health surveillance data from 12 countries/territories, this study introduces an improved index that could more accurately capture and communicate the mortality risks of air pollution than currently used indices. Highlighting O3 and NO2 as emerging key drivers of the risk index across countries/territories, the findings underscore the need for refined risk metrics and more targeted air quality management policies.

Data and code availability

All data used in our study were obtained from the MCC Collaborative Research Network (https://mccstudy.lshtm.ac.uk/) under a data-sharing agreement and cannot be made publicly available. Researchers can refer to MCC participants, who are listed in the author list of our study, for information on accessing the data for each country. Demonstrative codes are available at https://github.com/PierreMasselot/cgaim. Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

References

  • 1.Fuller R, Landrigan PJ, Balakrishnan K, Bathan G, Bose-O’Reilly S, Brauer M, Caravanos J, Chiles T, Cohen A, Corra L, et al. Pollution and health: a progress update. Lancet Planet Health. 2022;6:e535–e547. doi: 10.1016/s2542-5196(22)00090-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2.de Bont J, Jaganathan S, Dahlquist M, Persson Å, Stafoggia M, Ljungman P. Ambient air pollution and cardiovascular diseases: An umbrella review of systematic reviews and meta-analyses. J Intern Med. 2022;291:779–800. doi: 10.1111/joim.13467. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3.Requia WJ, Adams MD, Arain A, Papatheodorou S, Koutrakis P, Mahmoud M. Global Association of Air Pollution and Cardiorespiratory Diseases: A Systematic Review, Meta-Analysis, and Investigation of Modifier Variables. Am J Public Health. 2018;108:S123–S130. doi: 10.2105/AJPH.2017.303839. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4.Yang BY, Fan S, Thiering E, Seissler J, Nowak D, Dong GH, Heinrich J. Ambient air pollution and diabetes: A systematic review and meta-analysis. Environ Res. 2020;180:108817. doi: 10.1016/j.envres.2019.108817. [DOI] [PubMed] [Google Scholar]
  • 5.An R, Ji M, Yan H, Guan C. Impact of ambient air pollution on obesity: a systematic review. Int J Obes. 2018;42:1112–1126. doi: 10.1038/s41366-018-0089-y. [DOI] [PubMed] [Google Scholar]
  • 6.Zundel CG, Ryan P, Brokamp C, Heeter A, Huang Y, Strawn JR, Marusak HA. Air pollution, depressive and anxiety disorders, and brain effects: A systematic review. Neurotoxicology. 2022;93:272–300. doi: 10.1016/j.neuro.2022.10.011. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7.Global burden of 87 risk factors in 204 countries and territories, 1990-2019: a systematic analysis for the Global Burden of Disease Study 2019. Lancet. 2020;396:1223–1249. doi: 10.1016/S0140-6736(20)30752-2. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Chen K, Vicedo-Cabrera AM, Dubrow R. Projections of Ambient Temperature- and Air Pollution-Related Mortality Burden Under Combined Climate Change and Population Aging Scenarios: a Review. Curr Environ Health Rep. 2020;7:243–255. doi: 10.1007/s40572-020-00281-6. [DOI] [PubMed] [Google Scholar]
  • 9.Turnock ST, Reddington CL, West JJ, O’Connor FM. The Air Pollution Human Health Burden in Different Future Scenarios That Involve the Mitigation of Near-Term Climate Forcers, Climate and Land-Use. Geohealth. 2023;7:e2023GH000812. doi: 10.1029/2023GH000812. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.Kyrkilis G, Chaloulakou A, Kassomenos PA. Development of an aggregate Air Quality Index for an urban Mediterranean agglomeration: relation to potential health effects. Environ Int. 2007;33:670–676. doi: 10.1016/j.envint.2007.01.010. [DOI] [PubMed] [Google Scholar]
  • 11.Mason TG, Mary Schooling C, Ran J, Chan KP, Tian L. Does the AQHI reduce cardiovascular hospitalization in Hong Kong’s elderly population? Environ. Int. 2020;135:105344. doi: 10.1016/j.envint.2019.105344. [DOI] [PubMed] [Google Scholar]
  • 12.K P, Kumar P. A critical evaluation of air quality index models (1960-2021) Environ Monit Assess. 2022;194:324. doi: 10.1007/s10661-022-09896-8. [DOI] [PubMed] [Google Scholar]
  • 13.Fitz-Simons T. Guideline for Reporting of Daily Air Quality: Air Quality Index (AQI) (U.S. Environmental Protection Agency) 1999 [Google Scholar]
  • 14.Mandal TK, Gorai AK. Air Quality Indices: A Literature Review. J Environ Sci Eng. 2014;56:357–362. [PubMed] [Google Scholar]
  • 15.Liblik V, Kundel H. In: Urban Ecology. Breuste J, Feldmann H, Uhlmann O, editors. Springer Berlin Heidelberg; 1998. Estimation of Air Quality by Complex Index; pp. 114–118. [DOI] [Google Scholar]
  • 16.Stieb DM, Burnett RT, Smith-Doiron M, Brion O, Shin HH, Economou V. A new multipollutant, no-threshold air quality health index based on short-term associations observed in daily time-series analyses. J Air Waste Manag Assoc. 2008;58:435–450. doi: 10.3155/1047-3289.58.3.435. [DOI] [PubMed] [Google Scholar]
  • 17.World Health Organization. Personal Interventions and Risk Communication on Air Pollution. Geneva, Switzerland: 2020. [Google Scholar]
  • 18.Bell ML, Peng RD, Dominici F. The exposure-response curve for ozone and risk of mortality and the adequacy of current ozone regulations. Environ Health Perspect. 2006;114:532–536. doi: 10.1289/ehp.8816. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.Wei Y, Feng Y, Danesh Yazdi M, Yin K, Castro E, Shtein A, Qiu X, Peralta AA, Coull BA, Dominici F, Schwartz JD. Exposure-response associations between chronic exposure to fine particulate matter and risks of hospital admission for major cardiovascular diseases: population based cohort study. BMJ. 2024;384:e076939. doi: 10.1136/bmj-2023-076939. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 20.Huang WZ, He WY, Knibbs LD, Jalaludin B, Guo YM, Morawska L, Heinrich J, Chen DH, Yu YJ, Zeng XW, et al. Improved morbidity-based air quality health index development using Bayesian multi-pollutant weighted model. Environ Res. 2022;204:112397. doi: 10.1016/j.envres.2021.112397. [DOI] [PubMed] [Google Scholar]
  • 21.Huang W, Zhou Y, Chen X, Zeng X, Knibbs LD, Zhang Y, Jalaludin B, Dharmage SC, Morawska L, Guo Y, et al. Individual and joint associations of long-term exposure to air pollutants and cardiopulmonary mortality: a 22-year cohort study in Northern China. Lancet Reg Health West Pac. 2023;36:100776. doi: 10.1016/j.lanwpc.2023.100776. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Tan X, Han L, Zhang X, Zhou W, Li W, Qian Y. A review of current air quality indexes and improvements under the multi-contaminant air pollution exposure. J Environ Manage. 2021;279:111681. doi: 10.1016/j.jenvman.2020.111681. [DOI] [PubMed] [Google Scholar]
  • 23.Cao R, Wang Y, Huang J, Zeng Q, Pan X, Li G, He T. The construction of the air quality health index (AQHI) and a validity comparison based on three different methods. Environ Res. 2021;197:110987. doi: 10.1016/j.envres.2021.110987. [DOI] [PubMed] [Google Scholar]
  • 24.Chen R, Wang X, Meng X, Hua J, Zhou Z, Chen B, Kan H. Communicating air pollution-related health risks to the public: an application of the Air Quality Health Index in Shanghai, China. Environ Int. 2013;51:168–173. doi: 10.1016/j.envint.2012.11.008. [DOI] [PubMed] [Google Scholar]
  • 25.Du X, Chen R, Meng X, Liu C, Niu Y, Wang W, Li S, Kan H, Zhou M. The establishment of National Air Quality Health Index in China. Environ Int. 2020;138:105594. doi: 10.1016/j.envint.2020.105594. [DOI] [PubMed] [Google Scholar]
  • 26.Lu X, Zhang S, Xing J, Wang Y, Chen W, Ding D, Wu Y, Wang S, Duan L, Hao J. Progress of Air Pollution Control in China and Its Challenges and Opportunities in the Ecological Civilization Era. Engineering. 2020;6:1423–1431. doi: 10.1016/j.eng.2020.03.014. [DOI] [Google Scholar]
  • 27.Huang W, Li S, Guo Y. Enduring public health challenge: air pollution and the potential dominant risks of particulate matter. Environ Health. 2023;2:1–2. doi: 10.1021/envhealth.3c00185. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28.Manisalidis I, Stavropoulou E, Stavropoulos A, Bezirtzoglou E. Environmental and Health Impacts of Air Pollution: A Review. Front Public Health. 2020;8:14. doi: 10.3389/fpubh.2020.00014. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29.Keil AP, Buckley JP, O’Brien KM, Ferguson KK, Zhao S, White AJ. A Quantile-Based g-Computation Approach to Addressing the Effects of Exposure Mixtures. Environ Health Perspect. 2020;128:47004. doi: 10.1289/EHP5838. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30.Xu H, Zeng W, Guo B, Hopke PK, Qiao X, Choi H, Luo B, Zhang W, Zhao X. Improved risk communications with a Bayesian multipollutant Air Quality Health Index. Sci Total Environ. 2020;722:137892. doi: 10.1016/j.scitotenv.2020.137892. [DOI] [PubMed] [Google Scholar]
  • 31.Zhang L, Chen Y, Dong H, Wu D, Chen S, Li X, Liang B, Yang Q. Improving the construction and prediction strategy of the Air Quality Health Index (AQHI) using machine learning: A case study in Guangzhou, China. Ecotoxicol Environ Saf. 2024;287:117287. doi: 10.1016/j.ecoenv.2024.117287. [DOI] [PubMed] [Google Scholar]
  • 32.Cao R, Liu W, Huang J, Pan X, Zeng Q, Evangelopoulos D, Yin P, Wang L, Zhou M, Li G. The establishment of Air Quality Health Index in China: A comparative analysis of methodological approaches. Environ Res. 2022;215:114264. doi: 10.1016/j.envres.2022.114264. [DOI] [PubMed] [Google Scholar]
  • 33.Sandhiya L, Kolandaivel P, Senthilkumar K. Oxidation and nitration of tyrosine by ozone and nitrogen dioxide: reaction mechanisms and biological and atmospheric implications. J Phys Chem B. 2014;118:3479–3490. doi: 10.1021/jp4106037. [DOI] [PubMed] [Google Scholar]
  • 34.Malashock DA, Delang MN, Becker JS, Serre ML, West JJ, Chang KL, Cooper OR, Anenberg SC. Global trends in ozone concentration and attributable mortality for urban, peri-urban, and rural areas between 2000 and 2019: a modelling study. Lancet Planet Health. 2022;6:e958–e967. doi: 10.1016/S2542-5196(22)00260-1. [DOI] [PubMed] [Google Scholar]
  • 35.Zheng XY, Orellano P, Lin HL, Jiang M, Guan WJ. Short-term exposure to ozone, nitrogen dioxide, and sulphur dioxide and emergency department visits and hospital admissions due to asthma: A systematic review and meta-analysis. Environ Int. 2021;150:106435. doi: 10.1016/j.envint.2021.106435. [DOI] [PubMed] [Google Scholar]
  • 36.Sun W, Lu K, Li R. Global estimates of ambient NO(2) concentrations and long-term health effects during 2000-2019. Environ Pollut. 2024;359:124562. doi: 10.1016/j.envpol.2024.124562. [DOI] [PubMed] [Google Scholar]
  • 37.Grzywa-Celińska A, Krusiński A, Milanowski J. Smoging kills’ - Effects of air pollution on human respiratory system. Ann Agric Environ Med. 2020;27:1–5. doi: 10.26444/aaem/110477. [DOI] [PubMed] [Google Scholar]
  • 38.Chen G, Guo Y, Yue X, Xu R, Yu W, Ye T, Tong S, Gasparrini A, Bell ML, Armstrong B, et al. All-cause, cardiovascular, and respiratory mortality and wildfire-related ozone: a multicountry two-stage time series analysis. Lancet Planet Health. 2024;8:e452–e462. doi: 10.1016/S2542-5196(24)00117-7. [DOI] [PubMed] [Google Scholar]
  • 39.Doherty RM, Wild O, Shindell DT, Zeng G, MacKenzie IA, Collins WJ, Fiore AM, Stevenson DS, Dentener FJ, Schultz MG, et al. Impacts of climate change on surface ozone and intercontinental ozone pollution: A multi-model study. JGR Atmospheres. 2013;118:3744–3763. doi: 10.1002/jgrd.50266. [DOI] [Google Scholar]
  • 40.Liu C, Chen R, Sera F, Vicedo-Cabrera AM, Guo Y, Tong S, Coelho MSZS, Saldiva PHN, Lavigne E, Matus P, et al. Ambient Particulate Air Pollution and Daily Mortality in 652 Cities. N Engl J Med. 2019;381:705–715. doi: 10.1056/NEJMoa1817364. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 41.Zhao Q, Guo Y, Ye T, Gasparrini A, Tong S, Overcenco A, Urban A, Schneider A, Entezari A, Vicedo-Cabrera AM, et al. Global, regional, and national burden of mortality associated with non-optimal ambient temperatures from 2000 to 2019: a three-stage modelling study. Lancet Planet Health. 2021;5:e415–e425. doi: 10.1016/S2542-5196(21)00081-4. [DOI] [PubMed] [Google Scholar]
  • 42.Gasparrini A, Guo Y, Hashizume M, Lavigne E, Zanobetti A, Schwartz J, Tobias A, Tong S, Rocklöv J, Forsberg B, et al. Mortality risk attributable to high and low ambient temperature: a multicountry observational study. Lancet. 2015;386:369–375. doi: 10.1016/S0140-6736(14)62114-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 43.Alahmad B, Khraishah H, Royé D, Vicedo-Cabrera AM, Guo Y, Papatheodorou SI, Achilleos S, Acquaotta F, Armstrong B, Bell ML, et al. Associations Between Extreme Temperatures and Cardiovascular Cause-Specific Mortality: Results From 27 Countries. Circulation. 2023;147:35–46. doi: 10.1161/CIRCULATIONAHA.122.061832. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 44.Liu C, Cai J, Chen R, Sera F, Guo Y, Tong S, Li S, Lavigne E, Correa PM, Ortega NV, et al. Coarse Particulate Air Pollution and Daily Mortality: A Global Study in 205 Cities. Am J Respir Crit Care Med. 2022;206:999–1007. doi: 10.1164/rccm.202111-2657OC. [DOI] [PubMed] [Google Scholar]
  • 45.Liu C, Chen R, Sera F, Vicedo-Cabrera AM, Guo Y, Tong S, Lavigne E, Correa PM, Ortega NV, Achilleos S, et al. Interactive effects of ambient fine particulate matter and ozone on daily mortality in 372 cities: two stage time series analysis. BMJ. 2023;383:e075203. doi: 10.1136/bmj-2023-075203. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 46.Gasparrini A. Distributed Lag Linear and Non-Linear Models in R: The Package dlnm. J Stat Softw. 2011;43:1–20. [PMC free article] [PubMed] [Google Scholar]
  • 47.Sera F, Gasparrini A. Extended two-stage designs for environmental research. Environ Health. 2022;21:41. doi: 10.1186/s12940-022-00853-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 48.Hutcheon JA, Chiolero A, Hanley JA. Random measurement error and regression dilution bias. BMJ. 2010;340:c2289. doi: 10.1136/bmj.c2289. [DOI] [PubMed] [Google Scholar]
  • 49.Wei Y, Qiu X, Yazdi MD, Shtein A, Shi L, Yang J, Peralta AA, Coull BA, Schwartz JD. The Impact of Exposure Measurement Error on the Estimated Concentration-Response Relationship between Long-Term Exposure to PM2.5 and Mortality. Environ Health Perspect. 2022;130:77006. doi: 10.1289/EHP10389. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 50.Janssen NAH, Fischer P, Marra M, Ameling C, Cassee FR. Short-term effects of PM2.5, PM10 and PM2.5-10 on daily mortality in The Netherlands. Sci Total Environ. 2013;463–464:20–26. doi: 10.1016/j.scitotenv.2013.05.062. [DOI] [PubMed] [Google Scholar]
  • 51.Chen MJ, Leon Guo Y, Lin P, Chiang HC, Chen PC, Chen YC. Air quality health index (AQHI) based on multiple air pollutants and mortality risks in Taiwan: Construction and validation. Environ Res. 2023;231:116214. doi: 10.1016/j.envres.2023.116214. [DOI] [PubMed] [Google Scholar]
  • 52.Vicedo-Cabrera AM, Sera F, Liu C, Armstrong B, Milojevic A, Guo Y, Tong S, Lavigne E, Kyselý J, Urban A, et al. Short term association between ozone and mortality: global two stage time series study in 406 locations in 20 countries. BMJ. 2020;368:m108. doi: 10.1136/bmj.m108. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 53.Gasparrini A. Modelling Lagged Associations in Environmental Time Series Data: A Simulation Study. Epidemiology. 2016;27:835–842. doi: 10.1097/EDE.0000000000000533. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 54.Chen R, Yin P, Meng X, Wang L, Liu C, Niu Y, Liu Y, Liu J, Qi J, You J, et al. Associations between Coarse Particulate Matter Air Pollution and Cause-Specific Mortality: A Nationwide Analysis in 272 Chinese Cities. Environ Health Perspect. 2019;127:17008. doi: 10.1289/EHP2711. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 55.Yin P, Chen R, Wang L, Meng X, Liu C, Niu Y, Lin Z, Liu Y, Liu J, Qi J, et al. Ambient Ozone Pollution and Daily Mortality: A Nationwide Study in 272 Chinese Cities. Environ Health Perspect. 2017;125:117006. doi: 10.1289/EHP1849. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 56.Guo Y, Gasparrini A, Armstrong BG, Tawatsupa B, Tobias A, Lavigne E, Coelho MdSZS, Pan X, Kim H, Hashizume M, et al. Heat Wave and Mortality: A Multicountry, Multicommunity Study. Environ Health Perspect. 2017;125:087006. doi: 10.1289/EHP1026. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 57.Chen R, Yin P, Wang L, Liu C, Niu Y, Wang W, Jiang Y, Liu Y, Liu J, Qi J, et al. Association between ambient temperature and mortality risk and burden: time series study in 272 main Chinese cities. BMJ. 2018;363:k4306. doi: 10.1136/bmj.k4306. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 58.Masselot P, Chebana F, Campagna C, Lavigne É, Ouarda TBMJ, Gosselin P. Constrained groupwise additive index models. Biostatistics. 2023;24:1066–1084. doi: 10.1093/biostatistics/kxac023. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 59.Wong TW, Tam WWS, Yu ITS, Lau AKH, Pang SW, Wong AHS. Developing a risk-based air quality health index. Atmos Environ. 2013;76:52–58. doi: 10.1016/j.atmosenv.2012.06.071. [DOI] [Google Scholar]
  • 60.Adebayo-Ojo TC, Wichmann J, Arowosegbe OO, Probst-Hensch N, Schindler C, Künzli N. A New Global Air Quality Health Index Based on the WHO Air Quality Guideline Values With Application in Cape Town. Int J Public Health. 2023;68:1606349. doi: 10.3389/ijph.2023.1606349. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 61.Gutenberg S. Demystifying the air quality health index. Can Pharm J. 2014;147:332–334. doi: 10.1177/1715163514552560. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 62.U.S. Environmental Protection Agency. Technical Assistance Document for the Reporting of Daily Air Quality – the Air Quality Index (AQI) 2018. https://www.airnow.gov/sites/default/files/2020-05/aqi-technical-assistance-document-sept2018.pdf .
  • 63.European Environment Agency. European Air Quality Index Calculation. 2021. https://ecmwf-projects.github.io/copernicus-training-cams/proc-aq-index.html .
  • 64.Chinese Ministry of Ecology and Environment. Technical Regulation on Ambient Air Quality Index (on trial) 2012. https://www.mee.gov.cn/ywgz/fgbz/bz/bzwb/jcffbz/201203/W020120410332725219541.pdf .
  • 65.Chen R, Yin P, Meng X, Liu C, Wang L, Xu X, Ross JA, Tse LA, Zhao Z, Kan H, Zhou M. Fine Particulate Air Pollution and Daily Mortality. A Nationwide Analysis in 272 Chinese Cities. Am J Respir Crit Care Med. 2017;196:73–81. doi: 10.1164/rccm.201609-1862OC. [DOI] [PubMed] [Google Scholar]
  • 66.R Core Team. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing; 2024. [Google Scholar]
  • 67.Gasparrini A, Leone M. Attributable risk from distributed lag models. BMC Med Res Methodol. 2014;14:55. doi: 10.1186/1471-2288-14-55. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Appendix
EMS209693-supplement-Appendix.pdf (1.3MB, pdf)

Data Availability Statement

All data used in our study were obtained from the MCC Collaborative Research Network (https://mccstudy.lshtm.ac.uk/) under a data-sharing agreement and cannot be made publicly available. Researchers can refer to MCC participants, who are listed in the author list of our study, for information on accessing the data for each country. Demonstrative codes are available at https://github.com/PierreMasselot/cgaim. Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

RESOURCES