Spatio-temporal health benefits attributable to PM2.5 reduction in an Indian city

ABSTRACT Fine particulate matter (PM2.5) is linked with a wide spectrum of human health effects and has the highest contribution to total air pollution mortality. This study aims to quantify health benefits of reducing PM2.5 concentration to World Health Organization standard (annual mean = 10 µg m−3) for various health endpoints during 2011–2019 period using AirQ+ and BenMAP-CE software packages. Intraurban assessment in Vellore city, India was done by estimating health benefits at ward level. Both software packages estimated annual average all-cause, ischemic heart disease, stroke, and chronic obstructive pulmonary disease health benefits in the range of 919–945, 175–234, 70–152, and 99–175 cases at city level and 15–16, 3–4, 1–3, and 2–3 cases at ward level, respectively. Sensitivity analysis showed that relative risk had a large influence on health benefit estimates. Present study results will play a crucial role in the future air quality and public health policies of Vellore city.


Introduction
Air pollution is found to cause 6.67 million global deaths in 2019 and became the leading risk factor for mortality. Fine particulate matter (PM 2.5 ), a major air pollutant, is related to 4.14 million deaths accounting for 62% of total air pollution mortality (Health Effects Institute 2020). Most of the PM 2.5 mortality is accounted for both non-communicable (chronic obstructive pulmonary disease (COPD), ischemic heart disease (IHD), lung cancer, and stroke), and communicable (lowerrespiratory infection) diseases. Asian countries such as India and China are estimated with the highest PM 2.5 related premature mortality cases (Health Effects Institute 2018).
India is one of the populous countries and it accounted for almost one-fourth of PM 2.5 related global mortality (Health Effects Institute 2020). This can be because entire India's population is residing in areas exceeding World Health Organization (WHO) standard (annual mean PM 2.5 = 10 µg m −3 ) (Health Effects Institute 2019). Spatial and temporal trends of health benefits (avoidable premature mortality) especially disease-specific analysis is crucial for improving the healthcare system and thus reducing related expenditure (Yang and Zhang 2018). WHO and the U.S. Environmental Protection Agency developed AirQ+ and Environmental Benefits Mapping and Analysis Program -Community Edition (BenMAP-CE), respectively, to estimate health benefits of achieving PM 2.5 standards. Several recent studies have adopted these software packages (Kermani et al. 2020;Hajizadeh et al. 2020Hajizadeh et al. , 2021Manojkumar and Srimuruganandam 2021a). So far only two studies compared the results of AirQ+ and BenMAP-CE (Sacks et al. 2020;Mirzaei et al. 2021 Few Indian studies addressed the national, state, and district-level health benefits of reducing PM 2.5 concentrations (Upadhyay et al. 2018;Manojkumar and Srimuruganandam 2021a). Also, city-level premature mortality is estimated in Agra, Delhi, Chennai, Lucknow, and Mumbai (Maji et al. 2017a(Maji et al. , 2017bAfghan and Patidar 2020;Manojkumar and Srimuruganandam 2021b;Markandeya et al. 2021). Indian cities are usually divided into various wards that represent the administrative boundaries. Thus, estimating ward-level health benefits will help to assess the intraurban spatial heterogeneity. However, detailed intraurban and temporal assessments of PM 2.5 concentration and health benefits were not done earlier. Further, results from two software packages were not investigated in India. Hence, the present study aims to address these research gaps by (1) analysing the trend of PM 2.5 concentration, (2) estimating health benefits attributable to all-cause, COPD, IHD and stroke endpoints, (3) investigate and compare the results from AirQ+ and BenMAP-CE software packages. Novelty in the present study include (1) assessing the intraurban spatial heterogeneity of PM 2.5 concentration and health benefits, (2) comparing AirQ + (IER function) with BenMAP-CE (relative risks from cohorts), (3) calculating the sensitivity of input parameters viz., concentration, incidence, relative risks, and population in both software packages, and (4) creating ward-level health benefit map.

Study area
Vellore city located in Tamil Nadu state, India is selected for the present study. Vellore has 60 wards under 4 zones and their administrative boundaries are shown in Figure 1. The city is partially surrounded by the Eastern Ghats and the weather is usually dry and hot. Since Vellore is the corridor for two megacities (Bangalore and Chennai) and nearby states, it experiences severe traffic. Major air pollution sources are vehicular traffic, waste burning, resuspended road dust, and construction dust. Vellore is selected by the Indian government and World Bank to develop various sectors such as E-governance, infrastructure, cleanliness, and health.

Data
Input data required for both software packages are population, PM 2.5 concentration data, relative risks, and baseline incidence. Monthly PM 2.5 concentration from satellite data is used in the study due to the unavailability of ground-level data. Recently Dey et al. (2020) developed a 1 km ambient PM 2.5 database for India using aerosol optical depth and scaling factor from Moderate Resolution Imaging Spectroradiometer (MODIS) and Modern-Era Retrospective analysis for Research and Applications (MERRA-2) data, respectively. PM 2.5 Database is available for the last two decades (2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015)(2016)(2017)(2018)(2019) in the SAANS (satellite-based application for air quality monitoring and management at a national scale) web portal (http://www.saans.co.in/home.html). Raw dataset is downloaded through this portal and the PM 2.5 concentration values for Vellore city are extracted using R studio software. Ward-level PM 2.5 data used in this study are shown in Supplementary Figure. The Indian census is decennial and the population data with growth rate are made available by the Vellore corporation. Since the population in 60 wards of Vellore city is available only from the 2011 census, the starting year of study period is set to 2011. Thus, with available PM 2.5 and population data, the study period is considered from 2011 to 2019. Population in other years (2012-2019) are projected using the following expression.
Where P n = projected population for the n th year (n = 2012-2019), P 2011 = 2011 census population, r = annual growth rate, t = years between P 2011 and P n . This expression was used in our earlier study (Manojkumar and Srimuruganandam 2021a). Ward-level population for all study years is given in Supplementary Table S1. Year-wise baseline incidence rate of India is available from Institute for Health Metrics and Evaluation. Hence, baseline incidence rate of all-cause, COPD, IHD, and stroke endpoints are downloaded via Institute for Health Metrics and Evaluation website (http://ghdx.healthdata.org/gbd-results-tool). Baseline incidence values used in this study are given in Supplementary Table S2. So far, no Indian cohorts related long-term PM 2.5 exposure concentration with all-cause, COPD, IHD, and stroke endpoints. Also, it is found that PM 2.5 concentrations in Chinese cohorts are similar to Indian exposure levels. Thus, relative risk values from foreign cohorts (see Table 1) are adopted in this study due to the unavailability of Indian cohorts (Krewski et al. 2009;Yin et al. 2017). Most relative risk values are available for the age group above 30 years. Hence, health benefits are estimated for the population above 30 years.

AirQ+ and BenMAP-CE
Health benefits of achieving WHO standard (10 µg m −3 ) at city and ward levels are estimated for all-cause and disease-specific (COPD, IHD, and stroke) endpoints using two software packages viz., BenMAP-CE (version 1.5) and AirQ+ (version 2.0). These packages are available at https://www.epa.gov/benmap and https://www.euro.who.int/en/health-topics/environmentand-health/air-quality/activities/airq-software-tool-for-health-risk-assessment-of-air-pollution, respectively. BenMAP-CE is developed by U.S. Environmental Protection Agency for estimating the health benefits and economic value of air quality change with respect to PM 2.5 (Sacks et al. 2020). Following functions are used by BenMAP-CE to estimate the health benefits.
Here, HB = health benefits, Y o = baseline incidence rate, β = concentration-response coefficient, ΔPM = difference in PM 2.5 concentration, P = population, RR = relative risk. Initially, PM 2.5 concentration at 1 Km spatial resolution is loaded in BenMAP-CE. Model data, monitor data, and monitor rollback options are available for creation of PM 2.5 grid. Among these options, monitor rollback is suitable for the present study as it has roll back to a standard option that can create a baseline grid and control grid based on PM 2.5 data and standard value, respectively. Change in PM 2.5 concentration is estimated by the difference of baseline and control grids. Voronoi Neighborhood Averaging method is adopted for interpolating PM 2.5 concentration in unmonitored locations. Year wise changes in PM 2.5 concentration grids are created and the other input data such as population, baseline incidence rate, and the concentration-response coefficient (estimated from RR values given in Table 1) are entered in the software. Finally, health benefits are estimated using Equation (2). AirQ+ software is developed by WHO Regional Office for Europe and can be used for impact assessment, disease burden, and risk analysis of various air pollutants viz., PM 2.5 , PM 10 , ozone, and nitrogen dioxide. Impact assessment of long-term ambient PM 2.5 is chosen for the present study. As there are 60 wards in Vellore, the multiple area analysis in AirQ+ is selected. Ward wise baseline PM 2.5 concentration estimated from BenMAP-CE is entered under air quality data option. The integrated Exposure Response (IER) function can be used for quantifying relative risks of COPD, IHD, and stroke endpoints. Burnett et al. (2014) developed this function by integrating relative risks from outdoor and non-outdoor (household solid fuel, active, and second-hand smoking) studies. IER functions used in various Global burden of disease assessments are included in AirQ+ (Lim et al. 2012;Forouzanfar et al. , 2016. In addition to these, the functions with pre-established cut-off values of WHO guideline value (10 µg m −3 ) and interim-targets (15, 25, and 35 µg m −3 ) are also made available. IER function with pre-established cut-off values of WHO standard is selected for present study. Due to the unavailability of IER function for all-cause endpoint, the relative risk values are adopted from Krewski et al. (2009) study. AirQ+ initially quantifies the attributable proportion using relative risk values (Equation (4)). Further, health benefits of all endpoints are estimated using Equation (5).
Where, AP = attributable proportion, RR(c) = relative risk in category c of exposure, p(c) = proportion of exposed population in category c of exposure, HB = health benefits, P = population, BI = baseline incidence. Ward-level PM 2.5 concentration and health benefits results are aggregated at the city level for discussion.
Sensitivity of BenMAP-CE and AirQ+ results at city level were evaluated by (1) modifying the achievement of the WHO standard (10 µg m −3 ) to WHO interim target-3 (15 µg m −3 ), (2) using the country-level all-cause baseline incidence and state-level disease-specific baseline incidence from earlier studies (Chowdhury and Dey 2016;Shi et al. 2018), (3) adopting population growth rate given by Census of India (Census of India 2019), and (4) using lower and higher beta values from non-Chinese studies (refer Table 1) (Naess et al. 2007(Naess et al. , 2007Lipsett et al. 2011;Carey et

PM 2.5 concentration
Average PM 2.5 concentration refers to mean values of whole study period (2011-2019), whereas annual PM 2.5 concentration represents the year-specific values. The average ± standard deviation and range of annual PM 2.5 concentration in Vellore city during the study period were 49.58 ± 2.67 µg m −3 and 45.23 ± 1.83 to 53.17 ± 1.33 µg m −3 , respectively. Ward-wise average PM 2.5 concentration and annual PM 2.5 percentage change from 2011 are shown in Figure 2. All wards were found with highest annual PM 2.5 concentration in 2016 and the lowest in 2015. A total of 23 wards were found to be exceeding annual PM 2.5 concentration of 50 µg m −3 whereas all other wards were in the range of 45-50 µg m −3 . Among all wards, the 4 th ward had a maximum average PM 2.5 concentration of 51.89 ± 2.17 µg m −3 while the lowest was observed in 28 th ward (47.97 ± 2.46 µg m −3 ). Ward-level annual PM 2.5 percentage (2012-2019) change from 2011 was between 2.86 and 7.87%. Except for 2015, all other years had a positive annual PM 2.5 percentage change from 2011 (i.e. increase in concentration). A total of 36 wards reported with average PM 2.5 percentage change in the range of 3-6% while the remaining wards had <3% (9 wards) and >6% (15 wards).

Health benefits
Average and cumulative health benefits refer to mean and cumulative values during 2011-2019. Further, annual health benefits represent the year-specific values. City-level annual health benefits of all health endpoints are shown in Table 2. Also, the cumulative benefits at ward level are presented in Figure 3.

All-cause
BenMAP-CE and AirQ+ estimated 8271 and 8506 all-cause cumulative health benefits, respectively at city level during the study period. Likewise, the average health benefit estimates were 919 and 945 cases, respectively. The highest and lowest health benefit estimates at Vellore city were observed

Ischemic heart disease
Cumulative health benefits were observed from 1573 to 2103 cases in Vellore. Annual estimates of the city varied between 198 and 269 cases in BenMAP-CE and 150 to 196 cases in AirQ+ software. The cumulative and average IHD health benefits in both software packages were higher in 2016- Table 2. City-level annual health benefits estimated by AirQ+ and BenMAP-CE.

Software
Year All cause IHD Stroke COPD BenMAP-CE 2019 period when compared to 2011-2015. Also, the health benefits estimates were found to be maximum and closer to each other in the last two study years (2018 and 2019). Ward-level cumulative health benefit estimate ranges from 22 to 53 cases in BenMAP-CE and 18 to 40 cases in AirQ+. BenMAP-CE results showed that 43, and 17% of wards were observed with cumulative health benefits in the range of > 30 and 20-30 cases, respectively, whereas in AirQ+ about 25%, 67%, and 8% of wards were in >30, 20-30, and <20 cases, respectively.

Chronic obstructive pulmonary disease
BenMAP-CE estimated cumulative COPD health benefits of 1572 cases and average health benefits of 175 cases in Vellore. The annual health benefit estimates of city varied between 146 and 201 cases. AirQ+ showed lower estimates of 890 and 99 cases under cumulative and average health benefits. Also, the annual estimate in AirQ+ ranges from 78 to 118 cases. City-level annual health benefits in AirQ+ (BenMAP-CE) were more than 100 (180) cases from 2016 to 2019 while other study years were reported in the range of 78-95 (146-174) cases. As per BenMAP-CE, about 92% of total wards had cumulative health benefits >20 cases while remaining wards were estimated to be below 20 cases. Similarly, AirQ+ results showed that 3%, 85%, and 12% of wards had cumulative health benefits in the range of >20, 10-20, and <10 cases, respectively.

Stroke
Stroke had the lowest health benefits among all endpoints considered in the study. Cumulative health benefits of 1368 and 633 cases were estimated in Vellore by BenMAP-CE and AirQ+, respectively. Similarly, the average health benefits at the city level were 152 and 70 cases, respectively. Annual health benefits of Vellore were in the range of 133 to 173 cases in BenMAP-CE and 62-78 cases in AirQ+. Ward-level cumulative health benefits range from 14 to 34 cases in BenMAP-CE and 9-18 cases in AirQ+. As per BenMAP-CE, about 95 and 5% of wards had >15 and <15 cumulative health benefits, respectively while AirQ+ results showed that 13%, 17%, and 70% of wards were in the range of >15, 10-15, and <10 cumulative health benefits, respectively.

Sensitivity analysis
The year 2018 was considered for sensitivity analysis due to its highest benefits and the results are summarized in Table 3. Achieving interim target-3 in BenMAP-CE resulted in 933, 243, 157, and 182 cumulative health benefits of all-causes, IHD, stroke, and COPD, respectively. These results were 9-11% less than achieving WHO standard of 10 µg m −3 . Usage of baseline incidence from other studies resulted in reduced health benefits in all endpoints except stroke. The sensitivity of population in BenMAP-CE was observed to be similar across all endpoints. Adopting higher beta values showed 23%, 64%, 31%, and 56% higher estimates for all-cause, IHD, stroke, and COPD, respectively whereas the use of lower beta values resulted in the reduction of 49%, 77%, 39%, and 54%, respectively from main analysis.
Since the option for changing relative risk was available only for all-cause in AirQ+, the higher and lower relative risk estimates were not performed for disease-specific endpoints. AirQ+ estimates of all-cause, IHD, stroke and COPD varied from −48% to 29%, −9% to −32%, −16% to 7%, and −13% to −22%, respectively, across all sensitivity analysis. Similar to BenMAP-CE, the AirQ+ estimates of all-cause health benefits showed highest change while using highest and lowest relative risk values. Achieving interim target-3 had highest change in stroke endpoint while baseline incidence caused maximum change in IHD and COPD endpoints. Population parameter was found with lowest change in health benefit estimates among IHD and COPD. Both AirQ+ and BenMAP-CE estimated a similar decrease in all-cause health benefits while adopting WHO interim target-3.

Discussion
The annual mean PM concentration from 2011 to 2019 was higher than stipulated WHO standard. Present study results were compared with Indian megacities and it showed that PM 2.5 concentration at Vellore city was higher than Chennai (37 ± 17 µg m −3 ) city but lower than Delhi (114 ± 86 µg m −3 ), Kolkata (80 ± 67 µg m −3 ), Mumbai (54 ± 36 µg m −3 ), and Hyderabad (51 ± 23 µg m −3 ) (Sreekanth et al. 2018;Chen et al. 2020). Apart from megacities, a total of 76 Indian cities had a lower PM 2.5 concentration than the present study estimate (Pal et al. 2018). Further, Vellore had a significantly higher PM 2.5 concentration than global megacities viz., Chicago, London, Moscow, New York, Paris Sao Paulo, and Tokyo during 2013 (Cheng et al. 2016). Also, the trend clearly shows that annual average PM 2.5 concentration in recent years (2016-2019) was 6% higher than earlier years (2011)(2012)(2013)(2014)(2015). Vellore received smart city grant during 2016 and thereafter many projects such as improving road network, water distribution system, and sewer system are being carried out throughout the city. Traffic congestion also became regular due to these projects. Thus, these are the major reasons for increased PM 2.5 concentration during 2016-2019 at Vellore city.
Health benefit estimates of present study were lower when compared to other Indian studies 2017a, 2017bManojkumar and Srimuruganandam 2021b). This can be because Vellore had lower PM 2.5 concentration and population when compared to other cities. IHD, stroke, andCOPD attributed 354-641, 274-413, 18-42 cases, respectively, during 2012-2016 at Karaj city of Iran. (Kermani et al. 2020). Except COPD, all other endpoints in Karaj, Iran had higher cases than the present study. Also, Isfahan, a megacity of Iran had higher IHD cases and lower COPD and stroke cases when compared to present study (Hajizadeh et al. 2021). Kahraman and Sivri (2022) estimated all-cause attributable cases in six cities of Turkey. Results showed that Bursa and Istanbul had higher estimate of 1823 and 5207 cases during 2019 when compared with present study while Tekirdağ, Kocaeli, Sakarya, and Balıkesir, had lower estimate of 393, 495, 624, and 887 cases. These variations in estimates between studies can be related to city population and use of different baseline incidence and relative risk values.
City and ward-level AirQ+ and BenMAP-CE estimates of all-cause health endpoints were nearer to each other with variation between 2% and 3%. However, IHD, stroke, and COPD estimates in BenMAP-CE were 24-27%, 53-55%, 41-47%, respectively, higher than AirQ+ city-level estimates. Except relative risk, all other input data were same in both software packages for disease-specific health benefit estimates. Relative risk values were directly adopted from cohort studies in BenMAP-CE while AirQ+ estimated relative risks through the IER function. It could be a reason for variation in disease-specific health benefit estimates. This contrasted with earlier studies because the relative risk from same cohort was used in both software and thus estimated similar results (Sacks et al. 2020;Mirzaei et al. 2021). Limitations in the study are as follows. Health benefits estimated were based on PM 2.5 concentration alone and thus eliminating the influence of chemical properties. It was assumed that population was exposed to only one pollutant i.e. PM 2.5 that excludes synergetic effects of other air pollutants. Also, relative risks were adopted from foreign cohorts due to unavailability of Indian cohorts. Thus, results can be improved by using local epidemiological data.

Conclusion
The trend of PM 2.5 concentration and the health benefits of achieving WHO standard were assessed in Vellore city for the period 2011-2019. High resolution (1 km) satellite data was used for estimating PM 2.5 concentration. Results showed that annual PM 2.5 concentration was higher in recent years (2016-2019) when compared to earlier (2011)(2012)(2013)(2014)(2015). Also, the city and ward-level annual PM 2.5 concentrations were found to be exceeding WHO standard during all the study years. Vellore had cumulative all-cause, IHD, stroke, and COPD benefits in the range of 8271-8506, 1573-2103, 633-1368, and 8,901,572 cases, respectively. Cumulative health benefits at ward-level range from 86-208, 22-53, 14-34, 16-39 cases in all-cause, IHD, stroke, and COPD endpoints. Although estimates were different, the trend of health benefits among the endpoints (All-cause> IHD> COPD> stroke) were same in both software packages. This trend follows the baseline incidence. Also, the highest PM 2.5 concentration was observed during 2016 whereas maximum health benefits were achieved in 2018. This was because the year 2018 had higher population than 2016. Health benefit map of Vellore highlighted the importance of reducing PM 2.5 concentration at the city and ward levels. Further, our study demonstrated that health benefit results may vary based on the methodology used in BenMAP-CE and AirQ+.