carbon and tropospheric ozone
4.5 Impacts on human health
4.5.5 Estimating the impact of changes in particulate matter and
ozone on human health for the reference scenario
Changes in annual deaths due to changes in outdoor air pollution are estimated using a health impact function, taking into account the exposed population (Pop), baseline mortality rates (Y0), the CRF (β) defined by the epidemiology literature, and the change in PM2.5 and O3 (ΔX), which are examined separately; this analyses is based on a data integration methods similar to that used by Anenberg et al. (2010):
Δ Deaths = Pop * Y0 * (1-e-βΔX) A number of recent risk assessments have chosen the CRF (β) from the ACS study (Co- hen et al., 2004; Liu et al., 2009c; Anenberg et al., 2010) as that to use in quantitatively estimating the effect of particle reductions. The primary argument has been that this is a large study. This Assessment has rejected that approach for a more integrative one for a number of reasons. First, to choose one study is to implicitly claim that the other studies have no information to offer on the size of the effect, they only help justify the plausibility. This is believed to be untenable. Some of the other studies (e.g. the Nurses’ Health Study (Puett et al., 2009)) were also quite large, oth- ers, such as the Six City Study (Ferris et al., 1983) had advantages in random selection of subjects and less exposure error, and there are many studies whose cumulative impact on a judgment of the best CRF can hardly be zero. Moreover, the ACS study found effect modi-
fication by education (lower effects in college
graduates) but substantially oversampled col- lege graduates. Secondly, the main ACS anal- yses assign subjects to air pollution monitors in the same metropolitan area, even if they are in a different county. Some of the counties are very large, so that distances of more than 150 km between subject and monitor can oc- cur. This induces error in the exposure, which, as discussed previously, also biases the results. Of course, integrating the literature on air pollution and mortality, and incorporating supporting literature on effects on intermedi- ary biomarkers is a daunting task; fortunately this had already been done by the US EPA. They recruited a panel of 12 experts to re- view the literature with the aim of estimat- ing the CRF for all-cause mortality, and the uncertainty around it. Interestingly, the over- selection of college-educated participants and higher than average measurement errors were cited by many experts as the reason that all but one expert (including two authors of the ACS papers) had mean estimates greater than the ACS slope. The average of the 12 estimates reported as part of the expert elici- tation gives a mean function of 1.06 per cent decrease in deaths per 1 μg/m3 decrease in PM2.5, about 1.8 times higher than the esti- mate from the ACS study.
Studies of the acute effects of particles at high dose have shown a logarithmic CRF, with lower marginal impacts at higher concentrations. This suggests the concen- tration-response estimate above is likely too high for developing countries, where BOX 4.7: Metrics of health impacts
Concentration-response function (CRF) – the relationship between ambient concentration and a
health outcome.
Premature deaths – the number of deaths occurring earlier due to a risk factor than in the ab- sence of that risk factor.
Years of life lost (YLL) – the average number of additional years a person would have lived if he or
she would not have died prematurely.
Value of a statistical life (VSL) – The maximum amount an individual would be willing to pay to reduce his or her chance of dying by a small amount in a specified time period.
123 exposure is higher. However, due to emis-
sion reduction measures to which develop- ing countries are already committed, this Assessment’s baseline projections for PM2.5 concentrations in these regions for 2030 are much lower. For example, projections show anthropogenic concentrations (which exclude dust and sea salt) in the 30’s and 40’s of μg/m3 for China and India. This is within the range where the CRF was found to be linear by Schwartz et al. (2008), and so this Assessment has assumed a linear relationship with the 1.06% per μg/m3 of PM2.5 slope above.
The methodology used in this Assessment has previously been used by the WHO in 2004 to estimate the then current global burden of outdoor air pollution on human mortality (Cohen et al., 2004). This WHO comparative risk assessment estimated that urban PM2.5 was associated with about 800 000 annual deaths globally (Cohen et al., 2004). A more recent study, again for current conditions, used a global chemical transport model to es- timate the burden of outdoor anthropogenic PM2.5 and O3 in both urban and rural areas,
finding about 3.7 million annual premature
deaths due to PM2.5 and about 700 000 due to O3 (Anenberg et al., 2010), with about 75 per cent of these occurring in Asia. These estimates will be revised again by the ongoing Global Burden of Disease 2010 Study (http://www.globalburden.org/). No study to date has examined the burden of outdoor air pollution on mortality in 2030,
which would be influenced in opposite di- rections by several factors. Expected popu- lation growth, the epidemiological shift in developing countries from infectious disease to chronic disease, and increased emissions associated with economic development would lead to a larger burden; however, air pollution mitigation policies planned in North America and Europe along with expected future programmes in Asia would reduce the overall burden.
This Assessment examines the health impacts in 2030 due to the change in PM2.5 and O3 concentrations projected in 2030 relative to 2005 concentrations based on the reference
scenario. Because the PM studies primarily show effects of exposure on deaths from car- diopulmonary disease and lung cancer, and the O3 study effects on respiratory mortality, the estimates in this Assessment are restricted
to changes in those specific causes of deaths.
The equation above produces estimates of how many fewer deaths per year in 2030 would occur were the air pollutants lowered. In addition, YLL have been calculated, based on current life expectancies.
Regional changes in mortality for the 2030 population are estimated using the change in surface PM2.5 and O3 concentrations be- tween 2005 and 2030 simulated by the two models, GISS-PUCCINI and ECHAM- HAMMOZ (see Appendix A.4). Globally, emission changes in 2030 relative to 2005 substantially impact air pollution-related mortality. For each region, changes in health impacts from PM2.5 are larger than those from O3, an effect of both the relative chang- es in concentration of the two pollutants and the stronger relationship of PM2.5 with mortality (Figure 4.7). ECHAM-HAMMOZ simulates much larger PM2.5 concentration changes and smaller O3 changes compared with GISS-PUCCINI, though both models show the same directional change for each region (Figure 4.7). Both models also show similar spatial patterns, except for small ar- eas in northern China where surface PM2.5 increases in GISS-PUCCINI but decreases in ECHAM-HAMMOZ (Figure 4.8).
The impact of ambient PM is significant
compared to other causes of death.
Implementation of tight emission regulations in North America and Europe lead to 0.1-0.8 million avoided PM2.5-related deaths annually (primarily in Europe), corresponding to 0.5-4.8 million avoided YLL. These ranges include estimates from both models and the associated 95 per cent
confidence intervals. Expected regulations in East Asia, Southeast Asia and the Pacific
are estimated to avoid 0.1-1.1 million annual PM2.5-related deaths (0.4-7.7 million YLL), but O3-related mortality is estimated to increase by 0-0.2 million (0.1- 1.4 million YLL). Continued rapid emissions
124
Figure 4.7. Regional change in annual PM2.5 cardiopulmonary and lung cancer and O3 respiratory mortality (in
millions of lives) in 2030 relative to 2005 for the reference scenario, using simulated concentrations from the two models. Confidence intervals of 95 per cent are based on uncertainty in the CRF only.
-1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0
Africa East Asia, South East
Asia and Pacific Latin America andCaribbean North America andEurope South, West andCentral Asia
Change in mortality in 2030 (millions)
(
GISS-PUCCINI PM2.5 ECHAM-HAMMOZ PM2.5 GISS-PUCCINI O3 ECHAM-HAMMOZ O3
growth in South, West and Central Asia is estimated to increase PM2.5- and O3- related deaths by 0.1-1.8 million (1.2-15.9 million YLL) and 0-0.2 million (0.1-2.4 million YLL), respectively. Only modest changes in air pollution-related mortality are expected in Africa and Latin America and the Caribbean.
The 95 per cent confidence intervals shown by
the error bars in Figure 4.7 only represent the uncertainty in the CRF and exclude additional
uncertainties that are more difficult to quantify.
Some uncertainties, such as in emissions and the representation of atmospheric processes in the models used to simulate concentrations, occur upstream of the health impact
calculation – the comparison between the two models gives some indication of the importance of the latter of these two factors. Uncertainties are also associated with extrapolating CRFs found in the developed world to the rest of the world. PM2.5 concentrations simulated by ECHAM-HAMMOZ approach 95 μg/ m3 in East and Southeast Asia and the
Pacific, which is larger than the maximum
of 60 μg/m3 simulated by GISS-PUCCINI. The true CRF at these high concentrations is unknown since no long-term epidemiology studies have yet included such high exposure
levels. The CRFs used in this Assessment are assumed to be linear; if the marginal impact of PM2.5 is smaller at high concentrations, these results would be overestimates, particularly for results based on concentrations simulated by ECHAM-HAMMOZ.
These estimates are also based on a conser- vative population projection to 2030 and are limited to a population aged 30 and older, consistent with the ACS study. The sub- stantial respiratory mortality in developing countries is likely to be impacted by air pol- lution (Carbajal-Arroyo et al., 2011) in ways our model does not capture. These results also focus only on mortality and do not ac- count for substantial changes in morbidity that would result from these changes in air pollution. The estimated changes in mortal- ity may therefore be underestimates of the true health impacts of the projected changes in concentrations.
Indoor health estimate result
Using the CRF from Ezzati et al. (2004) and the exposure changes estimated by IIASA, we estimate 220 000 deaths and 6 million disabil- ity-adjusted years of life lost could be avoided in India in 2030 due to reduced indoor air pollution if the portfolio were implemented.
125 In China, a reduction of 153 000 deaths in
2030 and 1.9 million disability-adjusted years of life lost is estimated.
4.5.6 Economic valuation of