Valuing burden of premature mortality attributable to air pollution in major million-plus non-attainment cities of India

Accelerating growth due to industrialization and urbanization has improved the Indian economy but simultaneously has deteriorated human health, environment, and ecosystem. In the present study, the associated health risk mortality (age > 25) and welfare loss for the year 2017 due to excess PM2.5 concentration in ambient air for 31 major million-plus non-attainment cities (NACs) in India is assessed. The cities for the assessment are prioritised based on population and are classified as ‘X’ (> 5 million population) and ‘Y’ (1–5 million population) class cities. Ground-level PM2.5 concentration retrieved from air quality monitoring stations for the NACs ranged from 33 to 194 µg/m3. Total PM2.5 attributable premature mortality cases estimated using global exposure mortality model was 80,447 [95% CI 70,094–89,581]. Ischemic health disease was the leading cause of death accounting for 47% of total mortality, followed by chronic obstructive pulmonary disease (COPD-17%), stroke (14.7%), lower respiratory infection (LRI-9.9%) and lung cancer (LC-1.9%). 9.3% of total mortality is due to other non-communicable diseases (NCD-others). 7.3–18.4% of total premature mortality for the NACs is attributed to excess PM2.5 exposure. The total economic loss of 90,185.6 [95% CI 88,016.4–92,411] million US$ (as of 2017) was assessed due to PM2.5 mortality using the value of statistical life approach. The highest mortality (economic burden) share of 61.3% (72.7%) and 30.1% (42.7%) was reported for ‘X’ class cities and North India zone respectively. Compared to the base year 2017, an improvement of 1.01% and 0.7% is observed in premature mortality and economic loss respectively for the year 2024 as a result of policy intervention through National Clean Air Action Programme. The improvement among 31 NACs was found inconsistent, which may be due to a uniform targeted policy, which neglects other socio-economic factors such as population, the standard of living, etc. The study highlights the need for these parameters to be incorporated in the action plans to bring in a tailored solution for each NACs for better applicability and improved results of the programme facilitating solutions for the complex problem of air pollution in India.

Air pollution has globally become a leading reason accounting for 22-53% of all deaths from cardiovascular diseases (CVD), ischemic heart diseases (IHD), stroke, chronic obstructive pulmonary disease (COPD) and lung cancer (LC) 1 . World Health Organisation 86 reported that India has the highest of total polluted cities and was the major contributor to annual particulate concentration at a global level. More than 90% of people in India breathe air that exceeds the World Health Organisation (WHO) interim target-1 (35 µg/m 3 ) 2 . Balakrishnan et al. 63 reported total mortality of 1.24 million in India due to air pollution (Ambient + Household) for the year 2017. The report claims that the estimated figure is an underestimation as additional diseases attributable to air pollution were unaccounted for. Welfare loss due to air pollution for the south Asian region, 2013 was reported to be 7.4% of the Gross Domestic Product (GDP) 3 . The welfare loss comprises of negative externalities due to ambient particulate, household particulate and ambient ozone pollution, whereas monetary loss due to other harmful pollutants such as black carbon, organic carbon, SO 2 , NO 2 , etc. was not considered. Pandey et al. 4 estimated a total of 1.67 million premature mortality due to air pollution resulting in a total labour output productivity loss of 28.8 billion US$ for the year 2019 in India. Air pollution has almost topped the list of risk factors that cause mortalityin the country just below high blood pressure, tobacco and dietary risks. 84 . Increased Urbanisation . Annual PM 2.5 concentration data was retrieved for the year 2017 from the Central Pollution Control Board (CPCB) and respective State Pollution Control Board (SPCB) websites (Detailed in Supplementary Table S2). Consistency in data retrieval was maintained by considering only those stations which have data available for more than 104 days 25 but was made limited only for continuous monitoring stations and not manual stations in the current study due to the limited availability in manual monitored data. The retrieved data was manually screened by eliminating daily averaged concentrations deviating away from 2 to 1000 µg/m 3 range (Saini & Sharma 2019). Linear correlation developed between PM 10 -PM 2.5 was used to account for the missing PM 2.5 concentrations from PM 10 . It was observed that PM 10 -PM 2.5 relation couldn't be just developed using 2017 data for certain stations due to large data gaps, hence data from 2018 was considered for developing the correla-Scientific Reports | (2021) 11:22771 | https://doi.org/10.1038/s41598-021-02232-z www.nature.com/scientificreports/ tion with an assumption that clean air action plan for all the NACs being initiated with the base year 2019 [ 22 ] and had no substantial improvement in air quality for the immediately previous year. Thus developed correlation values were utilised to predict the respective PM 2. 5 for the year 2017. Due to the lack of either PM 10 or PM 2.5 monitoring by the government-operated instruments in cities like Kanpur, Dhanbad, Patna and Dehradun have necessitated for direct conversion of PM 10 to PM 2.5 using the ratio factor analysed from literature [26][27][28][29][30] for developing the relationship. 'RR' estimated using Eq. (1) was incorporated in Eq. (2) to estimate the excess mortality attributable to PM 2.5 ( E Mortality ). ' (RR − 1)/RR ' is attributable fractions defined as the proportion of mortality disease burden among exposed populations attributable to risk factors 34 . B i is the baseline mortality for the year 2017 specific to individual diseases (i) and city for age > 25 years. All-cause baseline mortality (age > 25) for each city was adjusted from percentage share of age-wise death rates and total mortality retrieved at the state and district urban level respectively from the Civil Registration System (CRS) India 35 using the population distribution data at the state, district and city levels being projected to the study period (https:// censu sindia. gov. in/). The percentage shares of age-wise death rates were assumed to be the same for both state and district urban regions. The percentage shares of cause-specific mortality (age > 25) for individual states were gathered from the Global Burden of Disease (GBD) India Compare Data Visualization interface 36 . The percentage share from GBD was then directly incorporated into city shares to estimate cause-specific PM 2.5 mortality cases (age > 25). www.nature.com/scientificreports/ Economic losses due to PM 2.5 mortality. Due to the unavailability of market value for human lives 37 , the monetary burden due to health risk was calculated using the method of Value of Statistical Life (VSL). VSL is an individual's willingness to pay (WTP) to avoid the risk of mortality 38,39 . The 2 most common methods to estimate WTP are contingent valuation (CV) using a questionnaire [40][41][42] and Compensating Wage Differential (CWD) using the Hedonic wage function approach 43 . The hedonic wage approach is widely used by various researchers 37,[44][45][46][47][48] 51,52 . ' β ' is the income elasticity and is recommended to be 0.8 53,54 .
Total economic loss (US$) due to PM 2.5 mortality for state ' k ' of NACs were calculated using Eq. (4). ' MC i ' is disease ('i') specific mortality cases.
Scenario modelling on policy intervention. NCAP was released by the Government of India with an overall national target of 20-30% reduction in PM by 2024 keeping 2017 as the base year. Scenario modelling for the suggested target was attempted using Eq. (1)-(4) to estimate the improvement in terms of monetary benefit (US$) with an optimum reduction in PM 2.5 by 30%. The baseline incidence of disease-specific mortality was assumed to be the same as that of 2017. The urban population (age > 25) statistics for the year 2024 provided by (https:// censu sindia. gov. in/) was used in the assessment.

Results and discussion
Annual average of PM 2.5 concentration. Data availability at continuous and manual monitoring stations for each NACs was shown in Supplementary Table S2. Missing data/data gaps of the NACs were completed using linear regression relation developed between the retrieved PM 2.5 and PM 10 data with previous studies as references. The detailed linear regression models and equation developed/retrieved were shown in Supplementary Fig. S1 and Table S3 respectively. The Pearson's correlation coefficient (r) ranged from 0.65 to 0.97 showing a strong linear correlation between the two pollutants of size 2.5 and 10 µm. Figure 3 shows the annual PM 2.5 average for 31 NACs for the year 2017. PM 2.5 concentration ranged from 33 to 194 µg/m 3 . The maximum concentration was found to be approximately 5 times the NAAQS-India prescribed annual average (ie., 40 µg/m 3 ) 25  Health risk assessment. The total mortality from IHD, Stroke, COPD, and LC is categorised under NCD 60 . Further, these causes along with LRI are represented as GEMM 5-COD (Cause of Death) in the study. To eliminate double-counting, mortality due to GEMM 5-COD was segregated from GEMM (NCD + LRI) and are reported as NCD-other. PM 2.5 -All-cause mortality. Estimated PM 2.5 all-cause mortality cases for the year 2017 is shown in Fig. 4 concentration. This is due to the difference in baseline incidence cases and exposed population of those cities (Details in Supplementary Sheet: Table S4, Table S5 and Fig. S2). Similar observations were also reported for a study in China 49 . Burnet et al. 12 reported total mortality for India using GEMM ie., NCD + LRI for the year 2015 to be 2219 thousand. Recently a study carried out by Maji 31 for china reported a total mortality GEMM (NCD + LRI) of 1930 thousand for the year 2017 and then a decreases by 9% for the year 2019. David et al. 11 showed that 49% of total mortality (Estimated using the IER    51 for Mumbai and Delhi reported total premature mortality (5-COD) due to PM 2.5 for the year 2015 to be 13,196 and 14,844 respectively based on the IER method. The reported estimates were found higher than the current study for two reasons (a) High PM 2.5 value; (b) Assumption of India-level constant baseline mortality rate for both 'X' class cities. Total mortality reported at the country level considers mortality due to both urban and rural styles of living 60,62 . 63 reported that 38.8% of total air pollution mortality in the country is attributable to household air pollution using filthy cooking fuel. The same study also reported that the figure doesn't hold true for 'X' class cities like Delhi whose percentage share towards household air pollution is approximately 0.4%. Hence considering the country level baseline incidence at the city level may result in overestimated premature mortality cases. Maji et al. 16 11.6% and 11.3% of total NCD-other respectively. IHD showed the highest percentage share for most of the NACs ranging from 20 to 72% and LC being the lowest ranging from 1 to 3%. Stroke and COPD almost showed similar variation ranging from 8 to 31% and 8 to 33% respectively. Stroke constituted the highest percentage share at Bhubaneswar and Guwahati. Whereas, for Jaipur and Jodhpur, 33% of the share is constituted of COPD cases. LRI ranged from 7 to 19% and NCD-other from 0 to 25%. Cities like Ludhiana and Amritsar exhibited no NCD-Other cases.   Figure 7 shows the percentage share of death estimated due to PM 2.5 pollution. The percentage share of mortality due to PM 2.5 ranged from 7.2 to 18.4%. For the year 2015 2 reported that 10.6% of the total premature mortality in the country is attributable to Particulate Matter (PM). The highest range of mortality was observed for North, Central and East zone cities (Table 1) reported that 76% of total premature deaths within the IGP region is due to indigenous emission sources and the rest due to cross-boundary transport and natural resources. He also reported that transboundary movement of pollutants from IGP anthropogenic sources to the North, Central, East, West and South zone accounts for 8% of total premature mortality. Cities like Jaipur (14.9%), Jodhpur (15.3%) in North and 'X' class cities like Mumbai (11.8%), Pune (10.6%) and Ahmedabad (12.8%) which showed higher percentage share was due to contribution from anthropogenic sources, outside India and natural sources (David et al. 2019) 64,65 . Central and South Zone cities showed the lowest percentage share of the range 7.6-9.5%. Bangalore being an 'X' class city showed a comparatively lower percentage share of 8.8% probably due to low influence from cross-boundary transport of pollutants within and outside Indian regions compared to others (David et al. 2019) 64 and improved meteorological conditions compared to North India cities 66 . Guttikunda et al. 67 reported that Central Zone cities like Raipur and Bihali Durg together contributed 17% and 12% of PM 2.5 from transportation and domestic cooking respectively to ambient air. This is minimal compared to other megacities. Bhubaneswar reported the lowest mortality share of 7.3%. The overall mortality due to air pollution is not just limited to PM 2.5 concentration but also depend on the population being exposed.

Monetary estimate of damages due to premature mortality attributable to air pollution
Economic damage (Million US$) associated with estimated cause-specific PM 2.5 mortality is shown in Fig. 8 69 using the benefits transfer approach and is ≈ 24.2% of our value used in the present study resulting in a higher difference in total

Scenario setting using existing policy interventions
The set of studies for the year 2017 was repeated for 2024. Policy intervention to reduce the impact of air pollution was found beneficial in reducing total economic loss but was not consistent for all NACs in the study. The inconsistency observed was due to the rise in exposed population (Age > 25) ( Supplementary Fig. S2 Figure 9a-c shows a potential economic loss (Million US$ as of 2017) for the year 2024 on 31 NACs due to a 30% reduction in PM 2.5 concentration setting 2017 as the base year. Total Mortality attributable to PM 2.5 for the year 2024 was estimated to be 79,633 [95% CI 70,859] which is 0.9% less than that of the base year 2017 and is shown in (Supplementary Fig. S3). GEMM 5-COD was found to capture a higher percentage of GEMM (NCD + LRI) for lower exposure reduction 12 . The estimated PM 2.5 cause-specific economic loss for the year 2024 is shown in (Supplementary Fig. S4). 50.6% of total economic loss for the year 2024 was attributed to IHD followed by COPD (15%), stroke(12.7%), LRI (9.9%), NCD-others (9.5%) and LC (2.1%). Percentage change in economic loss and total mortality cases are directly proportional ranging from − 12.4 to + 14.6% and is shown in Fig. 9d. Table 1 shows cities located in the West (− 4.1%) and South (− 2.3%) zone showed significant average change compared to that of North (+ 0.03%), East (+ 0.1%) and Central (+ 4.6%) India. All 'X' class cities showed significant reduction except Ahmedabad (+ 0.5%) and Delhi (+ 5.4%). Cities majorly residing in IGP states of Delhi, Bihar, and Uttar Pradesh showed the highest economic damage ranging from 5.3 to 14.2% excess from the base year 2017. Dehradun on the other hand also showed 14.6% excess monetary loss in the year 2024. Soni et al. 70 reported that soil, road dust, industrial activities, transportation activities, and anthropogenic burning to be dominant polluting sources in the Dehradun region along with influence from neighbouring polluted IGP regions. A comparative geographical representation of city-specific economic loss due to estimated PM 2.5 all-cause premature mortality for the years 2017 and 2024 is shown in Supplementary Figure S5. Maji et al. 16 carried out scenario modelling for Delhi, Mumbai, Kolkata, Bangalore and reported an increase in mortality by 20% for the year 2024 while implying Best Practise for Emission Control (BPEC) to reduce PM 2.5 concentration by 45% compared to the base year 2010. He also reported that maximum potential health benefits for all cities will be availed upon reaching IT-3 (15 µg/m 3 ) and AQG (10 µg/m 3 ) scenarios by 2040. In the current study for the same megacities, a percentage change in mortality by − 2.5% with NCAP policy interventions was observed. This improvement can be largely attributable to higher PM 2.5 reduction by NCAP policy interventions over BPEC. Higher reduction in mortality can be observed at lower concentration scenarios (such as IT-3, AQG) due to a sharp rise in RR as this function in GEMM follows a supralinear behaviour like IER which flattens at higher PM 2.5 concentration without resulting in a substantial mortality reduction (Saini and Sharma 2019).

Study assumptions and limitations
Health risk assessments studies have evolved over the years reducing the uncertainties as these can be the critical factors questioning the viability of any such studies. Several assumptions and limitations are involved in our study are (A) PM 2.5 concentrations retrieved from ground-based stations are considered as a representative value for the entire city. There may be some degree of misclassification while averaging the PM 2.5 value due to the spatial dependencies of pollutants 71 . (B) Study do not estimate economic loss associated with PM 2.5 morbidity, synergic effects, mortality associated with other external events such as accidents due to episodic events like haze. (C) Cause-specific baseline mortality share for each of the NACs used for calculations were specific to their respective state. This is due to the reason that mortality share details were unavailable at city levels. The verbatim share values transferred at the city level irrespective of state may result in underestimating/overestimating premature deaths. (D) The total death reported 35 can have a certain degree of variation as reported by 75 . An accuracy test in comparison with Sample Registration Survey (SRS) 72 value (Detailed in Supplementary Table S5) was carried out. The urban mortality rate was found below the lower limit for Uttar Pradesh and Uttarakhand. The mortality rates reported in 72 were specific to state-level (Rural and urban class) and not at the district/city level. Any modification in city-level mortality based on values reported in 72 can escalate the chances of further uncertainty. (E) The change in baseline mortality associated with policy interventions is not taken into account in the current study. The increase in NCD + LRI for some cities for the year 2024 was attributed to increasing urban sprawl, change in lifestyle and ageing 10,16 . (F) Study does not differentiate the associated mortality cases concerning indoor air quality. it is assumed that mortality associated with indoor quality for these 'X' and 'Y' class cities was minimal due to the improved lifestyle compared to the 'Z' class cities of India. (G) In this study, the Chemical composition of PM 2.5 species with an adverse effect on health 73,74 was not demarcated and assumes equivalent toxicity same as that of PM 2.5 mass concentration upon simple addition. (H) Due to paucity in city-specific data, the study assesses the VSL using state-specific GDP and PCI assuming the value to remain uniform for individual states irrespective of the city. However, there exists no agreed method or value for assessing the cost of mortality due to air pollution and are subjective based on the time and level of uncertainty.

Conclusion
The current study quantifies the economic burden due to premature mortality attributable to PM 2.5 exposure in 31 NACs for the year 2017. Additionally, the study also quantifies the potential monetary benefits based on target scenarios suggested in National Clean Air Programme(NCAP) for the year 2024. PM 2.5 attributable total premature mortality cases were estimated to be 80,447, resulting in a total economic loss of about 90,185 million US$ for the year 2017. IHD (47%) contributed the highest mortality followed by Stroke (14.7%), COPD (17.0%), LRI (9.9%) and LC (1.9%). NCD-others which were neglected in previous studies has accounted for 9.3% of total premature mortality resulting in a total economic loss of 9106.8 million US$. 7.6-18.4% of total mortality was attributable to excess PM 2.5 exposure. The highest PM 2.5 total premature mortality (Economic burden) of 49,212 (65,621 million US$) and 24,227 (38,511 million US$) was observed for class 'X' cities and North India Zone respectively. Despite being subjected to low PM concentration, some major cities reported high mortality cases due to a larger portion of the population being exposed. A decrease of 0.9% to 79,709 cases and 0.7% to 89,558 million US$ was observed on premature mortality and economic loss respectively for the year 2024. However, it is vital to note that the improvement assessed are limited to 31 of the 122 NACs (ie., 25% of the total NACs). The overall improvement in the air quality of 122 NACs through NCAP is anticipated to facilitate a substantial economic benefit attributable to premature mortality reduction. A reduction by 0.6% to 6.9% in cause-specific mortality cases except for LRI (increase by 5.6%) was observed. The overall improvement observed as a result of policy intervention seems insignificant and needs detailed analysis and insights for improvement. This is probably because no city-specific social indicators such as rising population & baseline mortality rates along with economic indicators such as Per Capita Income & Consumer Price Index were considered in deciding the target criteria, which seems to be the need of the hour. It is clear, that for a country like India with 122 NACs, a policy with immediate effects to alleviate pollution levels in achieving NAAQS was mandatory. Subsequently, there exists a robust requirement of necessitating these previously mentioned indicators in setting target scenarios via policy intervention to ensure maximum health and economic benefits in the identified NAC. It is important to emphasize the India specific cohort studies considering heterogeneous topographic conditions, as the lung capacity and dose-response vary largely subjected to the level of pollution and exposure at which people are inhabited. Such studies can favour in prioritising the cities/district/state that requires stringent abatement rules to improve the well-being of human. It is vital to strengthen the available database such as continuous PM 2.5 monitoring and subsequent expansion of the monitoring network for improved PM representation, minimising the data gaps in mortality report, Cause-specific mortality reporting at state/district level, Availability of economic indicators at district/city level, VSL survey at state/district/city level. Tuning up these vital requirements can help researchers arrive at an estimate with lower uncertainties benefiting policy makers in decision making. Contemporarily, successful implementation and progress of NCAP actions and targets shall be assessed at ground level for all the NACs to aid the anticipated benefits of improved air quality in the country.

Data availability
All data generated or analysed during this study are included in this published article and its Supplementary Information files. The raw datasets and corresponding codes generated during and/or analysed during the current study are available in the Economic-Assessment repository, https:// github. com/ moort hynair/ Econo mic-Asses sment. git.