Health benefits from reducing indoor air pollution from household solid fuel use in China — Three abatement scenarios

Environment International 33 (2007) 831 – 840 www.elsevier.com/locate/envint

Health benefits from reducing indoor air pollution from household solid fuel use in China — Three abatement scenarios Heidi Elizabeth Staff Mestl a,⁎, Kristin Aunan a , Hans Martin Seip a,b a

Center for International Climate and Environmental Research, P.O. Box 1129 Blindern, N-0318 Oslo, Norway b Department of Chemistry, University of Oslo, P.O. Box 1033 Blindern, 0315 Oslo, Norway Received 6 February 2007; accepted 26 March 2007 Available online 1 May 2007

Abstract According to the World Health Organization (WHO), indoor air pollution (IAP) from the use of solid fuels in households in the developing world is responsible for more than 1.6 million premature deaths each year, whereof 0.42 million occur in China alone. We argue that the methodology applied by WHO – the so-called fuel-based approach – underestimates the health effects, and suggest an alternative method. Combining exposure–response functions and current mortality and morbidity rates, we estimate the burden of disease of IAP in China and the impacts of three abatement scenarios. Using linear exposure–response functions, we find that 3.5 [0.8–14.7 95% CI] million people die prematurely due to IAP in China each year. The central estimate constitutes 47% of all deaths in China. We find that modest changes in the use of cooking fuels in rural households might have a large health impact, reducing annual mortality by 0.63 [0.1–3. 2 95% CI] million. If the indoor air quality (IAQ) standard set by the Chinese government (150 μg PM10/m3) was met in all households, we estimate that 0.9 [0.2–4.8] million premature deaths would be avoided in urban areas and 2.8 [0.7–12.4] million in rural areas. However, in urban areas this would require improvements to the outdoor air quality in addition to a complete fuel switch to clean fuels in households. We estimate that a fuel switch in urban China could prevent 0.7 [0.2–4.8] million premature deaths. The methodology for exposure assessment applied here is probably more realistic than the fuel-based approach; however, the use of linear exposure–response relationships most likely tends to overestimate the effects. The discrepancies between our results and the WHO estimates is probably also explained by our use of “all-cause mortality” which includes important causes of death like cardiovascular diseases, conditions known to be closely associated with exposure to particulate pollution, whereas the WHO estimate is limited to respiratory diseases. © 2007 Elsevier Ltd. All rights reserved. Keywords: Indoor air pollution; Rural; Urban; Exposure; China; Solid fuels; Health

1. Introduction Indoor air pollution (IAP) from solid fuels (biomass and coal) is known to pose a major health risk, leading to such serious illnesses as acute lower respiratory infections (ALRI) in small children, and chronic obstructive pulmonary disease (COPD) in adults. There is also evidence that lung cancer is associated with household coal combustion (Zhao et al., 2006). Other conditions like asthma, adverse pregnancy outcomes, loss of eye sight and cardiovascular diseases may also be associated

⁎ Corresponding author. Tel.: +47 22 85 87 67; fax: +47 22 85 87 51. E-mail address: [email protected] (H.E.S. Mestl). 0160-4120/$ - see front matter © 2007 Elsevier Ltd. All rights reserved. doi:10.1016/j.envint.2007.03.012

with indoor air pollution, adding to population morbidity and mortality (Smith et al., 2005). The World Health Organization (WHO) estimates that IAP is responsible for more than 1.6 million premature deaths each year in the developing world (WHO, 2002). In China alone, WHO estimates that about 420,000 die each year from the effects of IAP (Zhang and Smith, 2005). These estimates, however, were made using a method known as the fuel-based approach. The fuel-based approach uses the prevalence of fuel as an exposure surrogate and odds ratios of diseases combined with disease specific mortality and morbidity rates. This approach tends to underestimate the total disease burden due to both exposure misclassification and to limiting the estimates to selected diseases and population groups such as

832

H.E.S. Mestl et al. / Environment International 33 (2007) 831–840

children under five and adults — the most susceptible population groups. Smith and Mehta (2003) argue that with the current data availability the fuel-based approach is the most reliable. An alternative method is the pollutant-based approach where exposure assessment and exposure–response functions are used in conjunction with current rates of mortality and morbidity. The main criticism of pollutant-based approaches in general is that they tend to overestimate the health impact. Generally, the exposure–response functions originate in urban studies of developed countries. These functions are developed for populations with quite different socio-economic characteristics, pollutant mix and exposure levels than what is found in developing countries. There are indications that the slope of the exposure– response function is reduced at high concentrations. However, we do not know the true form of these functions, and adjusting the slope at high exposure levels is arbitrary. In order to estimate the burden of disease using this method, a lower exposure limit has to be defined where the pollutant concentration is assumed to no longer pose a health threat. In the case of particulates, no limit concentration has been identified (WHO, 2005), and thus setting a limit becomes subjective. In addition, we know little about the actual exposure experienced for large population groups subject to IAP from solid fuel burning. In earlier attempts to use the pollutant-based approach, simple exposure assessment has been used. In WHO (1997), the distribution of time spent by the world population in the eight most important environmental settings is combined with the mean particulate level in those settings to estimate population exposure. This results in a crude exposure assessment for developed and developing country populations divided in rural and urban areas. Our approach deviates from those applied in previous pollutant-based studies in several ways. We have earlier developed a Monte Carlo based method to estimate detailed exposure patterns for a large population based on indoor and outdoor air pollution and time activity tables (Mestl et al., 2007). Our approach divides the population in age, sex and whether people live in urban or rural settlements, and the pollution levels are estimated for several indoor and outdoor environments based on fuel use and geographic location. This approach makes large-scale exposure assessment possible revealing the finer patterns of exposure in the population. It also makes it possible to estimate exposure for people using ‘clean fuels’ such as gas and electricity, and to apply this as the lower limit, representing no exposure due to solid fuels. We argue that with the better exposure assessment this approach can be useful, at least for the Chinese population where several exposure–response studies have been conducted. The question we address here is thus, what is the burden of disease in China if we use a pollutant-based approach instead of a fuel-based approach? Because this approach enables estimation of continuous risk reduction, it allows for a more detailed impact assessment of possible interventions than what would be possible from the fuel-based approach. We look at how IAP affects different population groups and argue that the burden of disease in China may be even more substantial than previous estimates indicate. We then consider what exposure reduction and subsequent health improvement can be expected as a result

of some manageable interventions with respect to fuel-switching in Chinese households. We consider two different scenarios in this respect: clean fuels in all urban households and partial use of clean fuels in rural households. We also look at a scenario of meeting China's existing national indoor air quality (IAQ) standards, and ask whether these standards provide a useful goal, and if so, how they might be met. By looking at the exposure reduction from these scenarios in relation to the baseline mortality and morbidity, we estimate the potential health impact of the scenarios. Based on the findings we construct a fourth scenario to allow a more direct comparison with WHO estimates. The fourth scenario is a combination of the original three, and represents the ‘impact of solid fuels’ in China. We find that the WHO estimates may greatly underestimate the health risks associated with solid fuel use. We conclude that large health improvements can be achieved in China through manageable interventions in the households, and that through better exposure assessment the pollutant-based approach may be a useful supplement to the fuel-based approach in estimating burden of disease in developing countries. 2. Materials and methods The method used in this study is described in detail in a previously published exposure assessment methodology paper (Mestl et al., 2007). Here, we use the method to estimate exposure reduction for the three different scenarios described below. More specifically, for each scenario, the reduction in population weighted exposure (ΔPWE) is combined with dose–response functions and disease specific incidence rates for China to estimate change in burden of disease for each of the scenarios compared to current exposure levels.

2.1. Estimating exposure reduction The method described by Mestl et al. (2007) is based on published indoor air pollution (IAP) data and population time activity tables. The estimates are made using two-dimensional Monte Carlo simulations (2D-MC) to account for variability in large populations and uncertainties associated with the measured values. In our study, a total of 45 publications on IAP reporting measurements in China of particulate levels in kitchens, bedrooms, living rooms and workplaces were selected. The measurements were classified according to fuel used in the study households, the time of year for measurement, and geographic location (urban/rural and province). Urban outdoor particulate levels were based on monitoring in 45 Chinese cities in 2002 (Sinton et al., 2004a). The rural outdoor particulate levels were estimated based on local emissions and background levels (Mestl et al., 2007). The time activity patterns were based on two Chinese surveys, one from Chongqing (Wang et al., submitted of publication), and one from Hong Kong (Chau et al., 2002). The Chongqing survey includes both urban and rural populations. However, the survey was made in winter and did not include small children, an important group when estimating health impact. For the urban population we included small children by also using the Hong Kong study. For the rural population we had to turn to a study from Bangladesh (Dasgupta et al., 2004). There are, of course, cultural differences between China and Bangladesh, leading to different time use. However, the Chongqing and the Bangladesh time activity results show similar age and gender patterns. Where the patterns differed, we modified the Bangladesh study according to the Chongqing survey. For instance, the Bangladesh study lacks school attendance for children, and we therefore added the time that children attend school from the Chongqing study, and subtracted those hours partly from the time they spent indoors at home, and partly from the time spent outdoors in the Bangladesh survey. In the Bangladesh study we found that the time activity patterns for small children (both sexes) and for the elderly women were quite similar. The small children group (0–5 years) is missing in the Chongqing study, and we added a time activity for this age

H.E.S. Mestl et al. / Environment International 33 (2007) 831–840 group that was equal to the time activity of the female elderly in accordance with the findings in Bangladesh. The relevance and usage is discussed in more detail in (Mestl et al., 2007). The population was thus divided into groups by age and sex, and whether they lived in urban or rural areas, as well as north or south China. Using this detailed data, we estimated exposure levels and uncertainty for each of the population groups described by the time activity patterns. In this paper we use the exposure estimates from Mestl et al. (2007) as the current total exposure experienced by the Chinese population. We then apply the method to estimate reductions in population-weighted exposure in the three scenarios described below. DPWE ¼ PWEc  PWEa

ð1Þ

where PWEc is current population weighted exposure, and PWEa is the exposure after implementation of the measure.

2.2. Abatement scenarios China's economy is expanding at a very fast rate with large-scale industrialization and consequently an increasing energy demand. Fuelling the growth puts a large amount of pressure on the energy sector, and supplying energy to the household sector becomes secondary. As a result, the energy situation of Chinese households is not improving at the same rate as the general economic development in China. According to the 2000 census data (ACMR, 2004), 80–90% of the rural households still use solid fuels for cooking. In the urban areas the situation is better, with 40–50% of households using solid fuels. The most commonly used fuels for cooking in Chinese households (rural and urban) are biomass (45%), coal (27%) and gas (27%). Electricity is used by only 1% of the population. Some efforts have been made to improve the household energy situation in China. Several large-scale programs to promote improved stoves have been carried out (Smith et al., 1993). By the early 1990s, more than 130 million improved biomass stoves with a chimney and grate were installed. The programs were initially an answer to biomass shortages in the 1980s, and air pollution and health were not addressed. In a recent assessment of the stove replacement programs it was found that the improved biomass stoves have an average energy efficiency of 14%. The traditional stoves were found to have an average energy efficiency of 9%. Variation within each category was large, with many traditional stoves outperforming the improved ones (Sinton et al., 2004b). Where only improved stoves were used, the IAP level was found to be improved; however, little improvement was found when unimproved stoves were present in the same kitchen, which is quite common. Mean indoor concentrations of PM4 were found to generally exceed 150 μg PM10/m3, the level set by the Chinese health-based national indoor air quality (IAQ) standard (Edwards et al., in press). Since the target of the program was an answer to a biomass shortage in the 1980s, and the environmental aspect was not addressed, only minimal efforts were made to improve coal stoves. By 2001, it was estimated that 1.7 million households in Hubei, Sichuan, Chongqing, Shanxi and Shaanxi were still using traditional unimproved coal stoves. The three scenarios we consider in this study are identified in the succeeding sections. 2.2.1. Scenario 1: clean fuels in all urban households In the UK and several other Western countries where coal has been an important household fuel, ‘clean coal’ was promoted at some stage for smallscale use, but it was found that this was not compatible with health-based pollution standards. Ultimately, household coal use was banned in many urban areas. In China, a few cities have started prohibiting use of household coal in central urban areas. Clean fuel use in urban households is expanding, and the remaining 40–50% of the urban population that now uses solid fuels will likely gain access to clean fuels in the foreseeable future. Banning of solid fuels in urban households will reduce indoor pollution levels for urban households to the current level for the gas users, or actually even lower since this also leads to an improvement of the outdoor air. In Mestl et al. (2006), we estimated a 12% reduction in annual average outdoor PM10 concentration in Taiyuan (Shanxi) if the coal-using 30% of the urban population switched to gas. The contribution of household emissions to the urban outdoor concentration depends, among other things, on the share of households using solid fuels, and on the industrial activity con-

833

tributing to the degradation of the local air quality. Taiyuan is an industrial city, and thus receives large amounts of local pollution from the non-domestic sector. The contribution of combustion of household fuels to local air pollution is therefore probably relatively small in Taiyuan compared to cities with less heavy industry. However, for lack of other estimates we use the Taiyuan reduction ratio as the basis for estimating the reduction in outdoor pollution levels in all Chinese cities as a result of an urban fuel switch. We estimate the new exposure levels for the urban population by applying IAP levels for urban gas users for all urban residents, and a 10% reduction to the urban outdoor environment. The exposure reduction of this measure is estimated conservatively, since the indoor air quality for gas users probably will be better than it is today as a consequence of the improvement in outdoor air quality. 2.2.2. Scenario 2: partial use of clean fuels in rural households Most rural and all urban households in China have access to electricity (Sinton et al., 2004b), and targeted efforts to promote high-efficiency electric appliances like water and rice boilers could reduce the demand for solid fuels, and thus reduce both indoor and outdoor pollution levels. However, with China's fast-growing industrialization this might be a poor alternative because of current and foreseeable electricity shortages. Regional and periodic cut-offs are still common, although the Chinese government foresees an improvement due to new installed production capacity (Xinhua, 2006). Another possibility would be to promote partial use of gas (LPG or local biomass gasifiers) in rural households, although securing gas supplies might also be a problem. Targeted efforts to promote more improved stoves and inform the population of the health hazards of open fires might also be an option, perhaps the most feasible one in many areas in the near future. However, a modest fuel switch in some rural households combined with a stove improvement in others might be feasible and give reductions in IAP levels. According to Edwards et al. (in press), a partial use of clean fuels may have a limited impact. They found no significant PM4 reduction in households that sometimes used gas or electricity. They suggest that this may be due to several factors: • Both gas and solid fuel stoves would often be lit simultaneously. • The number of initial lighting events is unaffected by the usage of other fuels due to the simultaneous use of multiple stoves. Since solid fuels pollute the most at the initial stage of firing, the peak pollution events are not avoided through a partial fuel switch. • The solid fuel stove would keep smoldering long after the energy requirement was fulfilled, burning up the remaining fuel. However, they also note that the sample size was small and do not discard that there might be an effect. In this paper, we assume reduction of the IAP levels in the households using solid fuels following a partial fuel switch or stove improvement related only to cooking. We assume that replacing 30% of the need for cooking energy with clean fuels would give a conservative 15% reduction in IAP. The application of separate heating stoves in the northern provinces in winter reduces the impact of this intervention. In the north, we therefore assume no alleviation in winter and a 15% reduction in summer. In the south, we assume a 15% reduction throughout the whole year. 2.2.3. Scenario 3: meeting the national indoor air quality standard in all residences China is one of the first governments in the world to define a national health based indoor air quality (IAQ) standard for residences, set at 150 μg PM10/m3 (Edwards et al., in press). There is no epidemiological evidence that there actually is a threshold value below which the concentration of particulate matter is harmless. WHO recently published new air quality guidelines recommending that the annual average ambient concentration of PM10 should not exceed 20 μg/ m3 (WHO, 2005). This value was chosen because it has been shown feasible to achieve in highly developed urban areas. For the Chinese government, setting a more stringent standard would, however, probably not be politically feasible at the moment. In this third scenario, we estimate the effect on population exposure if all households met the IAQ standard. We do this by setting all IAP levels currently exceeding the recommended levels to 150 μg/m3.

834

H.E.S. Mestl et al. / Environment International 33 (2007) 831–840

Such a reduction in indoor air pollution would also imply an improvement in the outdoor air quality. Meeting this standard would most probably require a large-scale fuel switch to cleaner fuels. Thus for the urban areas we assume a 10% reduction in outdoor pollution levels, as is the case for the clean fuel measure. In the rural areas we assume that local air pollution from household fuel use will be eliminated, and thus set the outdoor level to the ambient background level.

2.3. Estimating health impact A range of studies in developing countries have shown that there is an enhanced risk for respiratory diseases associated with the use of solid household fuels. Most of these studies report the relative risk associated with different fuels. The actual exposure to indoor air pollution has usually not been monitored, nor has the total daily exposure resulting from the sum of exposure sources for the involved populations been estimated. In the following we take an alternative approach and use exposure–response relationships to estimate health impact. The relationships are from epidemiological studies of outdoor air and health effects, with the exception of a study from Kenya where the relationships are for indoor air pollution and health effects (Ezzati and Kammen, 2001). While, in principle, the health effects of exposure to air pollution should not depend on whether it occurs outdoors or indoors, there are major obstacles to applying outdoor air studies in the context of this study. One important issue is that we often need to extrapolate the exposure–response functions far beyond the levels occurring in the original studies. Also the pattern of exposure may be very different when exposure is partly related to indoor sources, because cooking and heating typically entail large variations over the course of a day, probably implying different amplitudes than those occurring in outdoor air. In the estimation of health effects we assume linear exposure–response functions. Extending the functions as we do is, as previously mentioned, uncertain when the range of exposure under consideration is large. It is probable that this leads to an overestimation of the effect, but we are fairly confident that we do not underestimate. If we were to use a scaling factor to reduce the slope of the relationship, the choice of that scaling factor would be arbitrary. We would no longer be certain that we are not underestimating the effect, but could still be overestimating. Thus using the relationships as we do can be seen as an upper limit estimate where the degree of overestimation is unknown. For lack of continuous exposure–response functions relevant for assessing health impacts of solid fuels, we apply the results from Chinese studies focusing on outdoor air (Aunan and Pan, 2004). Most of these are carried out in heavily polluted areas, where coal consumption is high. Hence for the coal-using

population in China, the pollutant mix experienced in the epidemiological studies might be representative of the pollutant mix experienced indoors. For the biomassusing population, the relationship might be different due to other pollutants in the emissions. If the pollutants of coal combustion are more harmful than the pollutants of biomass combustion, the burden of disease for the biomass-using population might be overestimated. This is a point for further investigations. Pope et al. (2002) describe a study from the US estimating impacts on mortality rates from long-term exposure to outdoor air pollution (in the following denoted “long-term mortality”). There are large socio-economic differences between US and Chinese populations, and the pollutant mix in China and the United States might also be substantially different, which may in turn affect vulnerability. However, there are no studies of the long-term effects in China, and since this is an important end-point, we choose to use the US study, keeping the limitations in mind. The acute respiratory infection (ARI) and ALRI coefficients are from an exposure–response study regarding health effects from indoor air pollution in Kenya (Ezzati and Kammen, 2001). Unfortunately, to our knowledge this is the only study providing continuous exposure–response functions for IAP and health effects, and more knowledge is needed. The study looks at a Kenyan population exposed to IAP from biomass combustion. Again, there are large socio-economic differences between China and Kenya. Moreover, this study was carried out in a region where biomass is the main household fuel, and the pollutant mix in China with large-scale use of other fuels might be quite different and lead to other health impacts. However, the study focuses on IAP, and the exposure–response functions are derived for exposure levels within the same range as in the present paper. Thus the limitations due to assumed linearity from the outdoor pollution level studies do not apply to the ARI and ALRI end-points. The estimated exposure reductions are combined with exposure–response functions to estimate the relative change in health outcomes in the following way: 100R ¼ ci dDPWE

ð2Þ

where R is relative change (100R in %) and ci is the exposure–response coefficient (in %) for health end-point i. To estimate the uncertainty of the percentage change in health outcome, we apply a Monte Carlo simulation assuming the exposure–response coefficient to be normally distributed, and multiply by the distributed ΔPWE from the twodimensional Monte Carlo (2D-MC) estimates. When multiplying two approximately normal distributions we get an approximately log-normal distribution and proceed to estimate median and geometric standard deviation.

Fig. 1. Exposure to PM10 (μg/m3) for the different fuel-user categories in urban and rural China. The bars show the exposure explained by outdoor air pollution as the lower portion of each bar and the exposure explained by indoor air pollution as the upper portion (estimates from Mestl et al. (2007)). The error bars show one standard deviation of the estimated total exposure.

H.E.S. Mestl et al. / Environment International 33 (2007) 831–840 A certain number of people fall ill or die each year irrespective of exposure to air pollution. We call this the baseline incidence rate (B). We assume that the observed incidence rate I is the sum of the baseline incidence rate B and the incidence rate due to the given level of air pollution exposure, H. The change in incidence rate due to reduction in air pollution is therefore estimated relative to the baseline rate: BþH ¼I H ¼ Rd B Rd Id P h ¼ Hd P ¼ 1þR

ð3Þ

where P is population and h is the health impact, or number of cases avoided. We do not actually know the baseline, and with this method we estimate a different baseline for each abatement scenario. However, by assuming a baseline, we assure that the health impact attributed to the scenario does not exceed the actual rate of disease observed. With approximately log-normally distributed relative change, the estimated number of cases avoided also becomes log-normal. We therefore estimate the 95% confidence interval by [μg / σ2g,μg ⁎ σ2g], where μg is the median of the avoided cases and σg is the geometric standard deviation.

3. Results 3.1. Estimated exposure reductions Fig. 1 shows total exposure as estimated in Mestl et al. (2007) and the share of exposure explained by indoor air pollution. The category

835

‘Total’ means all fuel user categories in the geographic area. We see that for all solid fuel user groups the major part of the exposure can be attributed to IAP. The influence of confounders such as smoking, cooking oil fumes, mosquito coils and others on the estimated total exposure is not known, as discussed in Mestl et al. (2007). Sinton et al. (2004b) found that despite high smoking rates in China, they measured no significant difference in IAP for households with or without smokers when the households used solid fuels. They explain that with the small amount of pollutants produced per cigarette compared to that from solid fuels in gas-using households there might be an increase in IAP due to environmental tobacco smoke (ETS), which might explain the extra exposure due to IAP experienced by gas users as shown in Fig. 1. When we look at the impact on exposure following the scenarios, we estimate relative change in exposure due to interventions. It is likely that smoking and other confounders cancel each other out when estimating change in exposure since the same ‘error’ may be present in both the ‘before’ and ‘after’ exposure. ΔPWE on a county level for the three different abatement scenarios is shown in Fig. 2. Not surprisingly, we find that large exposure reductions can be anticipated at a population level when measures are implemented to address the household energy situation. ΔPWE for the entire population, along with regional differences for the various scenarios, is shown in Table 1. For the urban population, switching to clean fuels would give an estimated 166 μg/m3 annual average exposure reduction. The exposure reduction achieved through a complete fuel switch is the same as the

Fig. 2. ΔPWE (μg/m3) for three different abatement options in mainland China: 1: Clean fuels in urban residences, 2: partial fuel switch in rural residences, and 3: IAQ standard met in all indoor environments.

836

H.E.S. Mestl et al. / Environment International 33 (2007) 831–840

Table 1 Reduction in population weighted exposure ΔPWE estimated for the three scenarios

North South Urban Rural Total

Scenario 1: fuel switch in urban areas

Scenario 2: partial fuel switch in rural areas

Scenario 3: meeting IAQ standard

Current PWE

161 (38) 172 (50) – – 166 (30)

42 (5) 102 (16) – – 65 (7)

– – 245 (34) 565 (54) 441 (35)

591 (59) 579 (87) 382 (47) 715 (80) 585 (72)

Column 4 shows current PWE for comparison (S.D.). IAQ: indoor air quality, standard set at 150 µg PM10/m3.

extra exposure burden experienced by the population due to solid fuels. Thus for the urban population this scenario is comparable with the WHO burden of disease estimates (WHO, 2002). Promoting a partial fuel switch in rural households is a measure giving, in this context, modest exposure reductions, while affecting a large population. The measure gives an estimated ΔPWE at 65 μg/m3. It is found to be most potent in the south, since in the north the measure was modeled to only have an effect in the non-heating season, whereas the southern population would have an exposure reduction throughout the year. The fact that we modeled this measure to only be effective outside the heating season probably underestimates the effect. However, as discussed above, we chose this approach because the share of indoor air pollution caused by heating and cooking in winter is unclear. ΔPWE for meeting the IAQ standard in all households is estimated to be 441 μg/m3 with 245 μg/m3 and 565 μg/m3 in the urban and rural areas respectively. Note that this scenario does not represent an actual intervention, but rather looks at the impact of meeting the IAP level defined as acceptable by the Chinese government. Currently, all northern urban household groups fail to meet the IAQ standard, implying that in order to meet this standard, more is needed than simply switching to clean fuels. The annual average outdoor air pollution in the northern urban areas in 2002 was 235 μg/m3 PM10. Switching to clean fuels and reducing this by 10% would still mean an annual

average outdoor concentration of 212 μg/m3. Since the outdoor air pollution influences the indoor air pollution level, a reduction in outdoor industrial emissions is needed in addition to a fuel switch to meet the IAQ standard in these areas. In the south, the annual average outdoor air pollution level is lower (130 μg/m3 PM10), and gas users manage to meet the IAQ standard, except in the kitchen. The reason they fail to meet the IAQ standard in the kitchen may be due to cooking style, with cooking oil fumes influencing the IAP level. For the rural population the situation is different, and a full fuel switch would give a larger exposure reduction than simply meeting the IAQ standard. As for the urban population, the exposure reduction following a full fuel switch, would be the same as the extra exposure burden that this population experiences from the combustion of solid fuels. However, we do not actually model a full fuel switch scenario for the rural population. Therefore we use the exposure reduction from the IAQ scenario as the extra exposure associated with solid fuels experienced by the rural population. Thus for the rural population, the exposure from the combustion of solid fuels is modeled conservatively. Based on the above findings we develop a fourth ‘the impact of solid fuels’ scenario that combines elements of two of the original three scenarios. We use results from the IAQ scenario to represent the impact of solid fuels for the rural population, and results from the fuel-switch scenario to represent the impact of solid fuels for the urban population. The estimated health impact of this fourth scenario is thus as analogous a set of results that we are able to develop for the burden of disease due to solid fuels compared to the estimates by the WHO (2002). 3.2. Health effects of abatement measures In the following, we estimate the health impact of all four scenarios. Each of the scenarios represents significant exposure reductions associated with reduced usage of solid fuels in the households. The dose–response coefficients from the literature and their standard errors are listed in Table 2 along with our estimated percentage change in health outcome (increase from baseline, 100R in Eqs. (2) and (3)) following the intervention scenarios 1–3. Meeting the IAQ standard in all indoor environments leads to very large reductions in exposures.

Table 2 Exposure–response coefficients and relative change in health outcome (increase from baseline, 100R in Eqs. (2) and (3)) for the Chinese population estimated for scenarios 1–3 End-point

All-cause mortality (long-term effect) a HA-CVD b HA-RD b Chronic RI in children b Chronic RI in adults b ARI in children c ALRI in children c ARI in adults c ALRI in adults c

Exposure– response coefficient ci (S.E.)

Urban Scenario 1: fuel switch

Scenario 3: meeting IAQ standard

Scenario 2: partial fuel switch

Scenario 3: meeting IAQ standard

0.24 (0.12) 0.07 (0.02) 0.12 (0.02) 0.44 (0.02) 0.31 (0.01) 0.07 (0.01) 0.04 (0.01) 0.04 (0.004) 0.03 (0.01)

38.1 (2.3) 11.2 (1.4) 19.7 (1.3) 64.4 (1.3) 52.6 (1.3) 12.1 (1.2) 6.9 (1.4) 6.7 (1.3) 5.0 (1.4)

56.9 (2.2) 16.7 (1.4) 29.3 (1.3) 107.0 (1.2) 75.8 (1.2) 17.6 (1.2) 10.0 (1.3) 9.7 (1.2) 7.2 (1.3)

15.1 (2.2) 4.4 (1.4) 7.7 (1.2) 26.2 (1.1) 20.5 (1.2) 5.4 (1.2) 3.1 (1.4) 2.6 (1.2) 2.0 (1.3)

132.2 (2.2) 39.0 (1.4) 67.9 (1.2) 239.2 (1.1) 177.4 (1.1) 44.6 (1.2) 25.3 (1.4) 22.8 (1.2) 16.9 (1.3)

% change in health outcome, median and geometric S.D. (σg) Rural

Scenario 4 is the combination of scenario 1 and the rural part of scenario 3. S.E.: standard error; IAQ: indoor air quality; CVD: cardiovascular diseases; RD: respiratory diseases; RI: respiratory illness; ALRI: acute lower respiratory illness; ARI: acute respiratory illness; HA: hospital admissions. a There are no Chinese cohort studies of the long-term impact of air pollution on mortality rates. Since the effect is large and important we choose to use the function from a study in the US by Pope et al. (2002). b Aunan and Pan (2004). c Exposure–response study based on indoor air pollution from biomass combustion in Kenya. Children under five and adults (Ezzati and Kammen, 2001).

H.E.S. Mestl et al. / Environment International 33 (2007) 831–840

Assuming linearity in the exposure–response functions is uncertain, and it is probable that the percentage change calculated as shown in Table 2 may be overestimated. For the rural population the estimated exposure reduction of the IAQ scenario is smaller than what could be expected from a complete fuel switch. Thus, using this to estimate the impact of IAP from solid fuels in the rural areas is conservative and may to some extent counteract the presumed overestimation due to the linear exposure–response functions. On the other hand, as will become clear below, the obtained results regarding the estimated attributable burden of disease for the rural population may still be higher than it is in reality. As mentioned, for the ARI and ALRI end-points the problem of extrapolating to large exposure values is less serious. We see that also for these end-points the impact is substantial, indicating a high importance of IAP for population health in China. Chinese crude mortality by age, sex and county in 2000 is listed in ACMR (2004). Rates of hospital admissions (HA) for cardiovascular diseases (CVD) and respiratory diseases (RD), as well as the annual rate of new cases (incidence rate) of chronic bronchitis (CB), are listed in Wan (2005) (Table 3). Combined with the percentage change in health effects for each abatement option and the population in the urban and rural areas, respectively, we estimate the number of avoidable cases, see Table 3. Unfortunately we have no reliable incidence rates for ARI or ALRI, and therefore the number of cases avoided for these end-points is not estimated. As expected, the 95% CIs are broad, and in some estimates the upper limit exceeds the actual number of cases observed. This is an artifact of the estimation method where the upper limit is estimated as μg ⁎ σ2g, where μg is the median of the avoided cases and σg is the geometric standard deviation. However, here it is mainly the lower 95% confidence limit that is of interest signifying a 2.5% probability that the true benefit of an abatement measure is below the lower bound. In scenario 1 we find that annually on average 696,000 premature deaths in the urban area could be avoided if households switch from solid fuels to clean household fuels (constituting about 28% of the annual deaths), with a 2.5% probability that the number of premature deaths is below 160,000. For the rural population, scenario 2, we find that the assumed 15% reduction in IAP would save 632,000 people from premature deaths annually, with a lower 2.5% bound at 126,000. From scenario 3, if IAQ standards were met in all indoor microenvironments, we estimate that annually 3.7 million people in China could be saved from dying prematurely each year, with a lower 2.5% bound at 838,000.

837

From Table 3, we find that hospital admissions for cardiovascular diseases could be reduced by 10% in the urban areas following a fuel switch (scenario 1). In the rural areas, scenario 2, a partial fuel switch could reduce the hospital admissions for cardiovascular diseases by 4%, while meeting the IAQ standard, scenario 3, could give a 14% and 28% reduction in the urban and rural areas respectively. Hospital admissions for respiratory diseases is estimated to reduce by 17% following scenario 1, a fuel switch in the urban areas, and 7% following scenario 2, a partial fuel switch in the rural areas. Meeting the IAQ standard, scenario 3, would give a 23% and 40% reduction for this endpoint in the urban and rural areas respectively. The number of new cases of chronic bronchitis (CB) could be reduced by 35% and 17% following a fuel switch in the urban areas, or a partial fuel switch in the rural areas (scenarios 1 and 2 respectively). Meeting the IAQ standard, scenario 3, could reduce the number of new CB cases by 43% and 64% in the urban and rural areas respectively. When looking at scenario 4, the impact of solid fuels in China, we find that 3.47 million people die prematurely due to solid fuels in the households each year, with a lower 2.5% bound at 820,000. The use of solid fuels also leads to 876,000 hospital admissions for cardiovascular diseases each year, with lower 2.5% bound at 451,000. We estimate that 2.3 million hospital admissions for respiratory diseases are due to the use of solid fuels in the households, with 2.5% bound at 1.5 million, and 8.4 million new cases of CB each year with a 2.5% probability that the true number of new cases due to solid fuels is below 6.1 million. 3.3. Comparison of our estimated mortality impacts of IAP with those of WHO Our central estimate of the long-term effect on the annual number of deaths from IAP (3.47 million) constitutes 47% of all deaths in China, which may seem unrealistically high. At first glance it does not seem probable that half of all deaths in China occur prematurely because of solid fuel smoke. On the other hand, that a large share of mortality and morbidity in China is linked to IAP seems probable, but direct causality in half the death cases might be to exaggerate the importance. It would be more informative to estimate life years lost from various causes. However, at present this may be overambitious partly because revealing which of the many possible causes lead to the premature death may not be possible. Our central estimate differs from the WHO estimate by a factor of eight. Our lower 2.5% confidence bound is

Table 3 Number of cases avoided per year when reducing population exposure to PM10 for the different abatement scenarios 1–3 End-point

Urban Incidence rate

All-cause mortality (long-term) a HA-CVD c HA-RD c CB c

Rural Cases/ year

Scenario 1: effect of fuel switch, cases avoided [95% CI]

Scenario 3: effect of meeting IAQ standard, cases avoided [95% CI]

5.14

2483

696

[160–NA b]

901

[178–NA b]

4.99 6.3 19

2410 3050 9040

244 503 3117

[118–500] [300–845] [1964–4945]

346 691 3902

[176–678] [439–1088] [2746–5544]

Incidence rate

6.4 3 6 11

Cases/ year

Scenario 2: effect of partial fuel switch, cases avoided [95% CI]

Scenario 3: effect of meeting IAQ standard, cases avoided [95% CI]

4834

632

[126–3165]

2778

[660–NA b]

2260 4350 8310

96 312 1414

[50–183] [205–475] [1059–1889]

632 1757 5312

[333–1200] [1173–2630] [4123–6845]

Scenario 4 is the combination of scenario 1 and the rural part of scenario 3. All numbers are in thousands. Urban and rural populations are 483 and 760 million respectively. CVD: cardiovascular diseases; RD: respiratory diseases; CB: chronic bronchitis; HA: hospital admissions. a Mortality rates (ACMR, 2004). b The estimated upper 95% confidence limit exceeds the actual mortality in China and is thus not applicable. c Annual incidences (new cases) per 1000 people (Wan, 2005).

838

H.E.S. Mestl et al. / Environment International 33 (2007) 831–840

almost double the WHO estimate (WHO, 2002). The size of the central estimate supports the notion that the exposure–response relationships actually level out at high exposure values. However, when comparing to the WHO estimate, some basic differences and assumptions are of importance. Our estimate refers to all-cause mortality estimated for the entire population, whereas the WHO fuel-based estimate (Smith et al., 2005) was made only for ALRI in children under five and COPD and lung cancer in adults above 30. All-cause mortality includes also mortality due to cardiovascular diseases (CVD), which are end-points shown to be strongly affected by air pollution (e.g., Pope et al., 2002). In China, CVD deaths constituted 33% of total deaths in 2002, respiratory tract cancers in total 3.5%, while COPD and ALRI (all ages) in total constituted 17% of all deaths (WHO, 2004). Thus, the fact that the end-point in our mortality estimation is all-cause mortality implies that a higher number of attributable cases is to be expected. Moreover, the fuel-based approach requires that the population is classified as exposed/not exposed based on whether they use solid fuels or not. The WHO (2002) estimate assumes lower odds ratios for the male population than the female (e.g. 1.8 and 3.2 respectively for COPD), thereby implying considerably lower exposure for the male population. It is assumed that the improved stove program has given significant IAP improvements by applying a “ventilation factor”. For children this “ventilation factor” is set to 0.25; i.e. assuming that 3/4 of children in households using solid fuels are not exposed to health-impairing IAP levels as a result of improved ventilated stoves. To account for previous long-term exposure in adults leading to chronic infections, they set the “ventilation factor” for the adults higher, at 0.5. They thus find that in China 52,000 children under five die prematurely from ALRI each year, and that 371,000 adults above 30, whereof 284,000 women, die prematurely from COPD or lung cancer each year. It has been shown that although the national improved stove program was successful in terms of promoting and installing improved stoves, it has had only a limited effect on the IAP level (Edwards et al., in press). According to the studies we apply here to estimate exposure, the use of solid fuels still results in elevated IAP and thus exposure, implying that the application of ventilation factors 0.25 and 0.5 are likely to be too optimistic and may lead to an underestimation of the proportion of people that are exposed. In Mestl et al. (2007) we found that there are only small gender differences in exposure. This is because the time spent indoors for men and women in China is quite similar, the difference is that the women spend more time in the kitchen, and the men more in the living room. With comparable average indoor air quality in these rooms this leads to comparable exposure for men and women. In Mestl et al. (2007) we argue that female exposure might be underestimated due to the use of average pollution levels and not peak levels as experienced during cooking in the exposure estimates. However, it is found that the men are exposed at severely elevated levels, thus implying that the use of quite different odds-ratios for men and women may underestimate the effect in the male population. Smoking is an important risk factor for the diseases associated with IAP. In our estimates, the use of exposure change, where smoking probably is present in both before and after intervention, it is likely that smoking cancels and does not influence our estimated health impact. In the WHO estimates they removed the fractions of lung cancer and COPD attributed to smoking before estimating cases attributable to IAP. This is conservative and may lead to an underestimation of the importance of IAP since some of the effect attributable to smoking could also be attributed to solid fuels. To make a more reasonable comparison between the WHO estimates and our own, we calculate the mortality effect in the same age groups as those included in the WHO estimates. We use age-specific mortality rates from ACMR (2004). For children under five, they are 3.8 and 7.4

per thousand in urban and rural areas respectively. For adults over 30, the respective urban and rural rates are 6.8 and 12.4 per thousand. If we estimate the total all-cause mortality for children under five attributed to solid household fuels (scenario 4), using age-specific mortality rates and mortality risk reduction (from DPWE and the exposure–response function by Pope et al. as above), we estimate that 219,000 children die prematurely due to IAP exposure each year. This corresponds to 51% of all deaths in this age group. The corresponding number obtained from the calculation procedure applied in the WHO estimates (Smith et al., 2005) if we remove the ventilation factor is 208,000 premature deaths in children. In the adult population above 30, we get approximately 3.1 million premature deaths due to IAP. For comparison, without the ventilation factor, Smith et al. (2005) would give 736,000 deaths. If Smith et al. had assumed equal incidence rates for men and women their estimation method gives approximately 1.1 million adult premature deaths each year. We see that for children under five the two estimation approaches give comparable results when assumptions regarding exposure pattern in the WHO estimates are adjusted. This is probably because the main cause of ill health in small children associated with air pollution is ALRI. Unfortunately, we have no age-and-cause specific death statistics for China to evaluate whether it is reasonable that 50% of all deaths in this age group are of causes that can be attributed to air pollution. For the adults, our central estimate is higher than the WHO estimate by a factor of three when adjusted for sex and exposure assumptions. The discrepancy is probably explained by the above-mentioned difference in using statistics for all-cause mortality as we do, or limiting to COPD and lung cancer as was done by the WHO. The assumed linear exposure–response functions probably also leads to an overestimation. Fig. 3 shows estimated premature deaths based on our approach (1), the WHO estimate (2), and estimated premature deaths using the two methods but adjusting the data to be more comparable as outlined above (3 and 4). The bars showing the estimates based on our approach also show the lower 95% confidence limit. We see that bar 4, i.e. the

Fig. 3. Estimated premature deaths attributable to solid fuel use in Chinese households. The estimates are based on the following: 1. The pollutant-based approach of this paper. 2. The WHO fuel-based approach. 3. Our approach limited to population groups children under five and adults over 30 years of age as in the WHO estimates. 4. Based on the WHO approach and data, but adjusted to no health benefit of the improved stove program (ventilation factor = 1) and equal exposure for men and women. The error bars in 1 and 3 show the lower 95% confidence limit of our estimates.

H.E.S. Mestl et al. / Environment International 33 (2007) 831–840

WHO method without the major assumptions regarding exposure pattern as discussed above, is within the confidence range of our estimates.

4. Discussion and conclusions The exposure assessment in Mestl et al. (2007) reveals that the often assumed gender differences in exposure do not seem to hold for China. Assumptions about the effect of improved stoves and gender differences are important in the WHO estimates, and with the fuel-based approach they are not possible to evaluate. Our central estimate of burden of disease from solid fuels is most likely too high, and much higher than the fuel-based WHO estimates. However, when the assumptions regarding exposure patterns in the WHO estimates are adjusted, the methods give results with overlapping confidence intervals. Our results indicate that the WHO estimates are likely to be conservative and that IAP may pose a greater threat to health in China, and possibly worldwide, than previously estimated. Our method allows for including uncertainty in all stages of the calculation, and when comparing to the WHO estimate we think it likely that the true burden of disease in China due to IAP is within our confidence range, probably in the lower end. The results also underline the need for more research both on population exposure and on exposure–response relationships related to IAP. Another source of uncertainty is the available health statistics in China. Rao et al. (2005) discuss an evaluation procedure for national cause-of-death statistics and apply the method to Chinese statistics. They conclude that there are large uncertainties in the available statistics, and that the statistics provided by the Ministry of Health (MoH) most probably underreports the total number of deaths. In the estimates discussed above, we used the mortality reported for 2000 in ACMR (2004). For comparison we estimate mortality for scenario 4 using mortality statistics from the Ministry of Health (MoH, 2004) and from China Statistics Press (CPSY, 2003). The estimated mortality based on the different statistics is shown in Table 4. The mortality rate for the entire population for ACMR and MoH are quite similar, and renders similar results (3.47 and 3.43 million deaths respectively). The CPSY statistic has a lower urban mortality rate and a higher rural mortality rate which results in

Table 4 Estimated premature mortality for scenario 4 based on three different statistics Statistic

Population group

ACMR (2004) Whole population Children b 5 Adults N 30 MoH (2004) Whole population CPSY (2003) Whole population Children b 5 Adults N 30 All numbers are in thousands.

Urban Rural Estimated incidence incidence mortality due rate rate to solid fuels [95% CI] 5.14 3.79 8.85 5.64 4.34 2.44 6.79

6.36 7.39 10.94 6.10 7.30 5.52 12.41

3474 219 3063 3434 3807 162 3268

[820–14,726] [52–924] [722–12,986] [780–15,116] [871–16,649] [37–708] [747–14,291]

839

a higher total estimated mortality of 3.8 million. These differences are well within the uncertainty of the estimates, but nevertheless point to the importance of providing reliable death statistics in the future. The main problem as discussed by Rao et al. (2005), however, is not the underreporting of total deaths, but rather the uncertainty with cause allocation. With reliable cause-of-death statistics, better estimates of the importance of solid fuels in the households could be made. We encountered similar difficulties and uncertainties when it comes to the statistical data for the annual incidence rates for chronic bronchitis. For instance, figures differed greatly between different data sources and also between the different points of time disease rates were surveyed. Moreover, it was difficult to interpret the figures in terms of the metrics annual prevalence rate and annual incidence rate. Although ‘excess premature deaths’ is a widely used metric for evaluating the impact of air pollution, it says little about the number of years by which lives are shortened in the average premature death attributed to the cause in question, which is important for estimating the life-years lost. Although the ‘life expectancy at birth’ (LEB) in a population is reduced by less that one year by a 10 μg/m3 reduction in long-term exposure to PM10 (Aunan et al., 2004 and references therein), the weighted expected remaining life years of those who die prematurely may be considerably higher (weighted according to the number of premature deaths in different age groups). Generally, premature deaths in younger persons have a much larger impact on LEB and life-years lost. The fact that IAP affects young children and most likely affects adults at all ages indicates that the high death toll estimated for this pollution source probably is accompanied by a profound impact on LEB and life-years lost in China. However, more studies on these issues are badly needed. There are differences in particulates emitted from biomass and coal, and most Chinese cohort studies focus on the effect on ill health from coal (Zhang and Smith, 2005). In this study we use the dose–response coefficients for PM10 regardless of the origin of particles. Being able to refine the estimates with different impacts for exposure to biomass IAP and IAP from coal could significantly improve the health assessment. Coalinduced IAP may lead to more severe health impacts, especially the risk of developing lung cancer (Zhao et al., 2006). This is an issue for further investigations. Achieving modest air quality improvements in rural households should be possible through a whole range of interventions as outlined above. Given the large health impact that can be anticipated, even small efforts in improving the household fuel situation in rural China should be encouraged. Acknowledgements The authors wish to thank Lynn P. Nygaard for valuable comments. We are also grateful for valuable suggestions from three anonymous referees on a previous version of the paper. The first author likes to thank Oslo University College; Department of Engineering for the PhD scholarship leading up to this work. This paper was funded in part by the Norwegian Ministry of Foreign Affairs.

840

H.E.S. Mestl et al. / Environment International 33 (2007) 831–840

References ACMR (All China Marketing Research.). 2000 China county population census data with county maps (Version III): All China Marketing Research Co. Ltd.; 2004. http://chinadatacenter.org/newcdc/. Aunan K, Pan X. Exposure–response functions for health effects of ambient air pollution for China — a meta-analysis. Sci Total Environ 2004;329:3-16. Aunan K, Fang J, Vennemo H, Oye K, Seip HM. Co-benefits of climate policy — lessons learned from a study in Shanxi, China. Energy Policy 2004;32(4):567–81. Chau CK, Tu EY, Chan DWT, Burnett CJ. Estimating the total exposure to air pollutants for different population age groups in Hong Kong. Environ Int 2002;27(8):617–30. CPSY. China population statistics yearbook. Beijing: China Statistics Press; 2003. Dasgupta S, Huq M, Khaliquzzaman M, Pandey K, Wheeler D. Who suffers from indoor air pollution? Evidence from Bangladesh. World Bank Policy Research Working Paper, vol. 3428. Development research Group World Bank; 2004. Edwards, R.D., Liu, Y., He, G., Yin, Z., Sinton, J., Peabody, J., Smith, K.R., Household CO and PM measured as part of a review of China's National Improved Stove Program, Indoor Air in Press. Ezzati M, Kammen DM. Indoor air pollution from biomass combustion and acute respiratory infections in Kenya: an exposure–response study. Lancet 2001;358(9282):619–24. Mestl HES, Aunan K, Seip HM. Potential health benefit of reducing household solid fuel use in Shanxi province, China. Sci Total Environ 2006;372:120–32. Mestl HES, Aunan K, Seip HM, Wang S, Zhao Y, Zhang D. Household energy classified by main cooking fuel and exposure to indoor air pollution in china. Sci Total Environ 2007;377:12–26. MoH (Ministry of Health). Annual health statistic in China, Ministry of Health, Peoples Republic of China, Beijing; 2004. Pope III CA, Burnett CJ, Thun MJ, Calle EE, Krewski D, Ito K, Thurston GD. Lung cancer, cardiopulmonary mortality, and long-term exposure to fine particulate air pollution. J Am Med Assoc 2002;287(9):1132–41. Rao C, Lopez AD, Yang G, Begg S, Ma J. Evaluating national cause-of-death statistics: principles and application to the case of China. Bull World Health Organ 2005;83(8) August. Sinton JE, Fridley DG, Lewis JI, Lin J, Chen Y, Zhou N. China energy databook. Environmental Energy Technologies Division; Lawrence Berkeley National Laboratory; 2004a. http://china.lbl.gov.

Sinton JE, Smith KR, Peabody JW, Yaping L, Xiliang Z, Edwards R, Quan G. An assessment of programs to promote improved household stoves in China. Energy Sustain Dev 2004b;VIII(3). Smith KR, Gu SH, Huang K, Qiu DX. 100-Million improved cookstoves in China — how was it done. World Dev 1993;21(6):941. Smith KR, Mehta S. The burden of disease from indoor air pollution in developing countries: comparison of estimates. Int J Hyg Environ Health 2003;206:279–89. Smith KR, Mehta S, Maeusezahl-Feuz M. Indoor air pollution from household use of solid fuels. Environmental and occupational risk factors. World Health Organization; 2005. p. 1435. http://www.who.int/publications/cra/ chapters/volume2/part2/en/index.html. Wan, Y., Integrated Assessment of China's Air Pollution-Induced Health Effects and Their Impacts on National Economy, Department of Social engineering, Graduate School of Decision Science and Technology, Tokyo Institute of Technology, PhD Thesis; 2005. Wang, S.X., Zhao, Y., Chen, G., Wang, F., Aunan, K., Hao, J., Assessment of population exposure to particulte matter air pollution in Chongqing, China. Submitted for publication. WHO (World Health Organization). Health and environment in sustainable development. Geneva: World Health Organization; 1997. WHO (World Health Organization). The world health report 2002. World Health Organization; 2002. http://www.who.int/whr/2002/en/. WHO (World Health Organization). Death and DALY estimates by cause for WHO member states; 2004. http://www3.who.int/whosis/menu.cfm? path=whosis,burden,burden_estimates,burden_estimates_2002N, burden_estimates_2002N_2002Rev_country&language=english. WHO (World Health Organization). WHO air quality guidelines global update 2005. World Health Organization; 2005. http://www.cepis.ops-oms.org/ bvsea/fulltext/guidelines05.pdf. Xinhua News Agency. China's power generation capacity to exceed 600 Mln Kw, Chinagate.com.cn; 2006. http://www.chinagate.com.cn/english/develop/ 49153.htm. Zhang J, Smith KR. Indoor air pollution from household fuel combustion in China: a review. The 10th International Conference on Indoor Air Quality and Climate; 2005. http://ehs.sph.berkeley.edu/krsmith/page.asp?id=1. Zhao Y, Wang S, Aunan K, Seip HM, Hao J. Air pollution and lung cancer risks in China — a meta-analysis. Sci Total Environ 2006;366:500–13.