Methodology & input data<\/strong><\/p>\n\n The Cost Estimator is based on methodology used for the Greenpeace Southeast Asia & CREA 2020 Toxic Air<\/u><\/em> report, which estimated the global health impact and economic cost resulting from exposure to fossil fuel air pollution.3 <\/sup>The Cost Estimator uses real-time ground-level PM2.5 <\/sub>measurements collated in IQAir\u2019s database.4 <\/sup>The algorithm applies scientific risk models in combination with population and public health data to estimate the health and economic costs of air pollution exposure. Mortality and cost estimates represent the total impact attributable to PM2.5<\/sub>.<\/p>\n\n Accuracy<\/strong><\/p>\n\n The Cost Estimator results represent a useful estimate of the real world cost based on currently available scientific knowledge. Like all research, the data and method used by the Cost Estimator contain uncertainty. This uncertainty relates to the precision and representativeness of pollution measurement data, demographic and epidemiological data and information about the economic cost of health impacts. Scientific understanding of the relation between exposure to air pollution and the associated risks for health is an active area of research; the latest available knowledge has been used. In case of missing data, conservative assumptions have been adopted, such that there is an inbuilt tendency towards underestimation of the cost of air pollution. Specifically the Cost Estimator accounts only for the impact of PM2.5<\/sub>, and not for other air pollutants which are also known to have negative health impacts.<\/p>\n\n 1) Air Pollution Data<\/strong><\/p>\n\n Sensor types<\/strong><\/p>\n\n The Cost Estimator uses measurements of PM2.5<\/sub>from ground-based outdoor pollution monitors in cities around the world. The devices are deployed and operated by governments, organizations such as PurpleAir, and private individuals. The data is sourced and aggregated continuously by IQAir.5<\/sup><\/p>\n\n City-wide value<\/strong><\/p>\n\n Most cities have several monitors, but the Cost Estimator uses one air pollution value for the entire city. Where there are multiple monitors within a city, their median value is used. Pollution is recorded hourly. The interval for computing 24-hour averages starts at 0:00 UTC in all cities around the world.<\/p>\n\n Quality controls<\/strong><\/p>\n\n Monitors with non-governmental operators undergo quality checks by IQAir before being incorporated into the database. The operator demonstrates by photo submission that the sensor is in a suitable location, which means outdoors and sheltered from precipitation, wind and direct sunlight. In addition, IQAir continuously checks that the carbon-dioxide (CO2<\/sub>) concentration, humidity and temperature readings reflect outdoor conditions during operation.<\/p>\n\n IQAir performs quality assurance checks on the aggregated data, for example by removing outlying data points when compared with neighbouring monitoring stations. The measurements for each monitoring station are grouped according to the city in which they are located.<\/p>\n\n Indirect PM2.5 <\/sub>data<\/strong><\/p>\n\n Some monitors only measure coarse particulate matter (PM10<\/sub>) and do not measure PM2.5<\/sub>. For these monitors, the PM2.5 <\/sub>values are derived from the PM10 <\/sub>measurement using a machine learning system. This system introduced an additional uncertainty; therefore only those cities where the share of native PM2.5 <\/sub>monitors is greater than 90% are included.<\/p>\n\n Representativeness \/ city filtering<\/strong><\/p>\n\n It is assumed by the Cost Estimator that the median of annual mean PM2.5 <\/sub>concentrations measured by ground level monitoring sites is representative of exposure in each city. However, cities are heterogeneous and this assumption introduces a significant uncertainty. Monitors in each city were assessed to identify obvious bias that could lead to overestimation of PM2.5 <\/sub>exposure. The station locations were overlaid on satellite imagery and manually checked through visual inspection, including the use of street level photography where available. Cities are only included where (a) the monitors show a reasonable geographic spread across the metropolitan area (no clustering, even distribution<\/p>\n\n Data completeness<\/strong><\/p>\n\n Real-world air pollution datasets are rarely complete. Outages during maintenance and technical faults result in missing data. On days without pollution measurement for a city, the previous day\u2019s measurement is used as a best guess. Only cities with >75% data availability over the year are included.<\/p>\n\n Other pollutants<\/strong><\/p>\n\n Besides PM2.5<\/sub>, there are other pollutants which are known to cause damage to human health, such as nitrogen dioxide (NO2<\/sub>), ozone (O3<\/sub>) and sulfur-dioxide (SO2<\/sub>). Although many of the cities included in the Cost Estimator also have monitors for these pollutants, data availability is generally much poorer. Furthermore, health impacts from NO2 <\/sub>are not included in our results because NO2 <\/sub>has a high spatial variability and thus the data from point measurements are less representative for the whole city.<\/p>\n\n <\/p>\n\n 2) Demographic and Epidemiological Data<\/strong><\/p>\n\n To assess the health impacts, the Cost Estimator uses population and epidemiological data for the metropolitan area of each city. The Cost Estimator assumes that the measured pollutant concentrations aggregated by IQAir provide a reasonable representation of pollution exposure across the metropolitan area of each city (see previous section). Population numbers for most world cities are taken from the United Nations World Population Prospects 2019,6 <\/sup>except for Taiwanese cities (Dept. of Household Registration, Ministry of the Interior, Taiwan7<\/sup>), Nairobi (Kenya Bureau of Statistics8<\/sup>), Delhi and Patna (India Population 20209 <\/sup>and India Population 201910<\/sup>) and Turkish cities (Turkish Statistical Institute11<\/sup>).<\/p>\n\n The data sources of the cities\u2019 incidence rates of each health impact as well as the risk functions used to compute the fraction of the health impacts that is attributable to air pollution are listed in Table 1.<\/p>\n\n Table 2. Relative risk at a 10 \u00b5g\/m<\/strong>3<\/sup><\/strong>increase in concentration for various health impacts and data source of the incidence rates.<\/strong>12,13,14,15,16,17,18,19,20,21,22<\/sup><\/strong><\/strong><\/p>\n\n 3) Health Impact Assessment<\/strong><\/p>\n\n The Cost Estimator links specific health impacts to air pollution using scientific risk models from the literature for each impact. The health impacts listed in Table 1 are included in the estimator.<\/p>\n\n The calculation applies concentration-response functions to the pollutant concentrations of PM2.5<\/sub>to determine the portion of the city\u2019s incidence\/prevalence rates of different health impacts caused by air pollution. A concentration-response function relates a pollutant concentration to the impact of that pollutant on a population.<\/p>\n\n Deaths, years of life lost and years lived with disability due to PM2.5 <\/sub>exposure are calculated using non-linear, age-specific risk, which gives the increase in risk of different health effects as a function of pollutant concentration, compared with clean air (see Table 1 for references).<\/p>\n\n For these health impacts, the annual incidence m <\/em>of a health impact in a specific city is related to the city\u2019s population P<\/em>, the cause-specific death rate in the country m<\/em>0<\/sub>, the risk ratio r <\/em>and the annual average pollutant concentration x<\/em>a <\/sub><\/em>by<\/p>\n\n m <\/em>= P <\/em>. m<\/em>0<\/sub> . (1 \/ r<\/em>(xa<\/sub><\/em>)),<\/p>\n\n where r<\/em>(x<\/em>a<\/sub><\/em>) is an empirical function (adopted from the literature). For most health impacts, r<\/em>(x<\/em>a<\/sub><\/em>) is modelled by a log-linear risk function, meaning that r <\/em>increases in a multiplicative manner per incremental increase in pollutant concentration:<\/p>\n\n r<\/em>(x<\/em>a<\/sub><\/em>) = exp[c <\/em>. (x<\/em>a<\/sub><\/em>–x<\/em>0<\/sub>)] if x<\/em>a<\/sub><\/em> <\/em>> x<\/em>0<\/sub><\/em>r <\/em>= 1 otherwise,<\/p>\n\n with c <\/em>being an empirical constant (also called \u201cconcentration-response factor\u201d) and x<\/em>0<\/sub><\/em>the detection<\/em> threshold <\/em>(also called \u201cno-risk threshold\u201d, Table 1 and Figure 1), which is usually the lowest concentration found in the study population – it is based on absence of direct evidence, not evidence that there is no risk. Excluding impacts below this threshold is a very conservative assumption.<\/p>\n\n For presentation of real-time data on the Cost Estimator website these annual costs are then apportioned linearly to the period of the calendar year that has elapsed in proportion to the average pollutant concentrations of the year to date, such that time periods with higher pollution incur greater costs. The year-to-date incidence m<\/em>d<\/sub><\/em>is calculated from the past-365-day incidence by<\/p>\n\n md<\/sub> = m . <\/em>[(xd<\/sub> – x<\/em>0<\/sub>) \/ <\/em>(xa<\/sub> <\/em>– x<\/em>0<\/sub>)] . t <\/em>\/ 365d if xa<\/sub>, xd<\/sub> > x<\/em>0<\/sub>; 0 otherwise,<\/p>\n\n where xd<\/sub><\/em>is the year-to-date average pollutant concentration and t <\/em>is the elapsed time in the current calendar year.<\/p>\n\n Figure 1: <\/strong>\u200bRelative risk function \u200br<\/em>\u200b(\u200bx<\/em>\u200ba<\/sub><\/em>\u200b) for preterm births due to PM\u200b2.5<\/sub> pollution (Trasande et al. 2016). Note the discontinuity of the central\/best estimate at the detection threshold \u200bx<\/em>0<\/em>\u200b<\/sub> = 8.8 \u200b\u03bcg\/m\u200b3<\/sup> as a result of the conservative assumption \u200br<\/em>\u200b(\u200bx<\/em>\u200ba<\/em>\u200b<\/sub> <\u200bx<\/em>\u200b0\u200b<\/em>\u200b<\/sub>) = 1.<\/p>\n\n Cost Assessment<\/strong><\/p>\n\n Health impacts associated with air pollution exposure have economic consequences. Principal among these are direct health care costs, (for example the cost of asthma treatment,) and the cost of reduced economic productivity. Economic productivity may be reduced by absence from or reduced capacity for work.<\/p>\n\n In addition, a higher risk of death or disability affects people\u2019s well-being; this welfare cost is measured in money by converting it to the amount that people are willing to pay in order to avoid the negative health outcome (\u201cwillingness-to-pay\u201d).<\/p>\n\n These two different types of cost are included where possible and are combined to a total cost. The health impacts listed in Table 2 are included in the calculation as described below:<\/p>\n\n Lost life years<\/strong><\/p>\n\n The economic losses from these air pollution-related deaths are assessed based on the resulting reduction in life expectancy, with one year of life lost equating to economic loss of EUR 56,000 (USD 67,000) in the European Union, following the EEA cost-benefit methodology27<\/sup>, and adjusted by Gross<\/p>\n\n Diabetes, asthma and other chronic respiratory diseases, and disabilities caused by stroke<\/strong><\/p>\n\n The Global Burden of Disease project29 <\/sup>has quantified the degree of disability caused by each disease into a \u201cdisability weight\u201d that can be used to compare the costs of different illnesses.<\/p>\n\n The economic cost of disability and reduced quality of life caused by diabetes and chronic bronchitis is assessed based on these disability weights, combined with the economic valuation of disability used by the UK environmental regulator DEFRA30<\/sup>. The valuation is adjusted by GNI PPP for other countries. For example, type 2 diabetes without complications has a disability weight of 4.9%, meaning that the cost of one year lived with diabetes is estimated at 4.9% of the cost of one year lived with disability, or USD 4,000 in the UK and USD 1,600 at world average income level.<\/p>\n\n The economic cost of asthma related to air pollution is assessed based on two indicators: new cases of asthma linked to NO2 <\/sub>exposure, and emergency room visits related to PM2.5 <\/sub>exposure. An assessment of the direct and indirect cost per year associated with childhood asthma, including medical costs and loss of income to the child\u2019s caregiver, found a cost of USD 3,800 and USD 4,000 in two different communities in California, United States31<\/sup>. The midpoint of these two valuations is used for the estimates, adjusted by the ratio of California\u2019s Gross Regional Product to the US national average, and by GNI PPP for other countries. The cost of an emergency room visit is taken from the same study.<\/p>\n\n Exposure to PM2.5<\/sub>is very likely linked to an even larger number of new asthma cases globally than exposure to NO2<\/sub>, but uncertainty in the estimates is large32<\/sup>, so this effect is not included. Instead, only the economic cost of emergency room visits for asthma linked to PM2.5 <\/sub>exposure are included. This is only a small part of the overall cost of the burden of asthma linked to PM2.5<\/sub>.<\/p>\n\n Preterm Birth and Low Birth Weight<\/strong><\/p>\n\n PMStep by step methodology<\/h2>\n\n
Health impact<\/strong><\/td> Pollutant<\/strong><\/td> Metric<\/strong><\/td> Relative risk<\/strong> r(x-x<\/em>0<\/sub><\/em>=10 \u00b5g\/m3<\/sup>)<\/td> Detection<\/strong> threshold<\/strong> x<\/em>0<\/sub><\/em>(\u00b5g\/m3<\/sup>)<\/td> Relative Risk<\/strong> Source<\/strong><\/td> Incidence Rate Source<\/strong>23<\/sup><\/strong><\/td><\/tr> Child asthma emergency room visits24<\/sup><\/td> PM2.5<\/sub><\/td> Annual mean<\/td> 1.025<\/td> 6<\/td> Zheng et al. 2015<\/td> Anenberg et al 2018<\/td><\/tr> Adult asthma emergency room visits25<\/sup><\/td> PM2.5<\/sub><\/td> Annual mean<\/td> 1.023<\/td> 6<\/td> Zheng et al. 2015<\/td> Anenberg et al 2018<\/td><\/tr> Preterm births<\/td> PM2.5<\/sub><\/td> Annual mean<\/td> 1.15<\/td> 8.8<\/td> Trasande et al 2016<\/td> Chawanpaiboon et al 2019<\/td><\/tr> Days of work absence<\/td> PM2.5<\/sub><\/td> Annual mean<\/td> 1.039<\/td> 0<\/td> WHO HRAPIE 2013<\/td> EEA 2014<\/td><\/tr> Adult deaths, years of life lost and years of life with disability<\/td> PM2.5<\/sub><\/td> Annual mean<\/td> non-linear<\/td> 2.4<\/td> Burnett et al26<\/sup><\/td> GBD 2017<\/td><\/tr><\/tbody><\/table><\/figure>\n\n ![\"\"](\"https:\/\/www.greenpeace.org\/static\/planet4-southeastasia-stateless\/2021\/02\/97840e32-screen-shot-2021-02-10-at-14.43.27.png\")
<\/figcaption><\/figure>\n\n