This is the foundation for the number of deaths between 29,000 - 40,000 a year 'caused' by pollution.
Basically, you aggregate the damage caused by pollution to equivalent lives lost through aggravation of their diseases and the general reduction in their life expectancy.
Source: government sponsored study.
http://webarchive.nationalarchives.gov.uk/20140505111410/http://www.comeap.org.uk/images/stories/Documents/Statements/FINAL_Local_mortality_burden_statement_August_2012.pdf#page=3
ANNEX 2: Methods Calculating the attributable fraction 1. For a given relative risk, RR, associated with a ubiquitous exposure such as outdoor air pollution, the proportion of disease (or deaths) that is attributable to that exposure (the population attributable risk fraction, or attributable fraction) is calculated by a simple formula: AF = (RR-1)/RR. This is often expressed as a percentage.6 2. For example, the proportion of deaths attributable to 10 μg/m3 of PM2.5 air pollution, assuming an associated relative risk of 1.06, would be 100 × 0.06/1.06 = 5.7%. 3. Estimates of mortality burden in a local area need to use a relative risk (and associated attributable fraction) reflecting the risk associated with the local population-weighted7 annual average PM2.5 concentrations under consideration8 . The RR applicable locally can be approximated by linear scaling (i.e. by assuming that if 10 µg/m3 leads to a 6% change in risk, then concentrations which differ by 1 µg/m3 should lead to differences in RRs of 0.6%. From this, the local attributable fraction can be derived. Linear scaling is inexact9 but this approach is unlikely to lead to practically important differences when estimating local RR and attributable fraction, particularly as the PM2.5 concentrations under evaluation are not likely to be hugely different from 10 µg/m3 . 6 The formula above is a special case (for universal exposures) of the more general formula: AF = p(RR-1) / [1 + p(RR-1)], where p is the prevalence of exposure to the cause of disease (or deaths) in the population under consideration. 7 The population-weighted mean is a useful summary statistic, which greatly simplifies the calculation of human health impacts if the concentration–response function used is linear with no threshold. In our estimation of the national mortality burden of air pollution (COMEAP, 2010) the populationweighted mean was calculated by multiplying the 1 km x1 km concentration values by 1 km x 1 km population statistics from the 2001 census. The values for all of the grid squares were summed and then divided by the total population to calculate the population-weighted mean. 8 Our national estimates (COMEAP, 2010) were of the burden associated with PM2.5 from anthropogenic sources. Published data on the contribution of different sources to background (i.e. not roadside or kerbside) PM2.5 concentrations were used to estimate background PM2.5 concentrations originating from anthropogenic sources. 9 The way of translating the RR to other PM2.5 concentrations that best corresponds to the concentration response function from which it derives (based on a proportional hazards model) is through the power function: RRc = 1.06^(c/10). In the case of a burden estimate, c is the PM2.5 concentration. This approach differs increasingly from linearity for higher relative risks and higher concentration increments. (This specific formula is applicable to coefficients - such as this one linking PM2.5 concentrations with mortality risk - that are expressed in terms of RR per 10 units (here 10 µg/m3 ). The denominator in the power term would be different for RRs expressed in terms of a different increment.) 12 Calculating attributable deaths 4. An estimate of the number of deaths attributable to long-term exposure to air pollution in a local area is given by multiplying the attributable fraction by the number of deaths annually in the local area. 5. To reflect the study from which the concentration response coefficient (relative risk) was reported, we used the number of deaths at ages 30 years or more in this calculation when estimating the national mortality burdens. However, the Office for National Statistics (ONS) in England and Wales publishes data on adult mortality in 10-year age groups of 25-34, 35-44 etc, so a figure of deaths at ages 30+ at the local level might not be easy to obtain. Similar considerations apply in Scotland and Northern Ireland. An estimate could be made by combining one half of the deaths in age group 25-34 with those for 35-44. However, such an adjustment seems unnecessary: the numbers of deaths below age 35 are a small proportion of the total, and the ‘cut-off’ at age 30 is based on lack of evidence at lower ages – it is possible and indeed plausible that long-term exposure to air pollution affects mortality risks in younger people also. We consider that, even if deaths below age 25 were included in the calculation (i.e. total number of local deaths), the difference between total deaths and those at ages 30+ would make only a small difference to the burden estimate. 6. Because of the variability and instability in small datasets, the reliability of local burden estimates can be improved by using death statistics from a number of years combined (e.g. 3 or 5 years) rather than basing the calculation on the number of deaths reported locally in a single year, and we recommend that this be done unless the year-on-year variation in annual deaths is small, in percentage terms. Calculating years of life lost to the local population 7. The years of life lost to the population can be estimated by summing the years of age-specific remaining life expectancy associated with each of the attributable deaths. This is the approach we took when estimating the national burden of air pollution (COMEAP, 2010). 8. As this method requires the use of complex life-table analysis, we suggest a simpler approach be used to generate local burden estimates: multiplying the calculated number of attributable deaths by the average loss of age-specific lifeexpectancy associated with attributable deaths in our national estimates, of approximately 12 years10 10 This should not be regarded as the loss of life likely to be associated with each death affected by air pollution. A figure of 11½ years was calculated (COMEAP, 2010) as being the average loss of life if 29,000 deaths were affected by air pollution. (COMEAP, 2010). In recommending this approach we reemphasise an important issue of interpretation. We look on this calculation - using the number of attributable deaths and the associated average loss of age-specific life-expectancy - as a computationally convenient way of estimating the total mortality burden, in terms of life-years lost in a given year aggregated over the whole population. As emphasised in COMEAP (2010) and noted again in Para 8 above, the number of attributable deaths should not be interpreted as the number of 13 individuals affected; and whatever the number of deaths affected and the average loss of life, the actual amount of life lost would vary between individuals.