smithkr_2003

7/27/2019 SmithKR_2003

1/11

Int. J. Hyg. Environ. Health 206, 279 289 (2003) Urban & Fischer Verlaghttp://www.urbanfischer.de/journals/intjhyg

International Journalof Hygiene andEnvironmental Health

The burden of disease from indoor air pollution in developingcountries: comparison of estimates

Kirk R. Smith, Sumi Mehta

Environmental Health Sciences, School of Public Health, University of California Berkeley, California, USA

Received December 16, 2002 Revision received January 10, 2003 Accepted January 13, 2003

Abstract

Four different methods have been applied to estimate the burden of disease due to indoor airpollution from household solid fuel use in developing countries (LDCs). The largest numberof estimates involves applying exposure-response information from urban ambient airpollution studies to estimate indoor exposure concentrations of particulate air pollution.Another approach is to construct child survival curves using the results of large-scalehousehold surveys, as has been done for India. A third approach involves cross-nationalanalyses of child survival and household fuel use. The fourth method, referred to as the`fuel based' approach, which is explored in more depth here, involves applying relative riskestimates from epidemiological studies that use exposure surrogates, such as fuel type, toestimates of household solid fuel use to determine population attributable fractions by

disease and age group. With this method and conservative assumptions about relative risks,4 5 percent of the global LDC totals for both deaths and DALYs (disability adjusted lifeyears) from acute respiratory infections, chronic obstructive pulmonary disease, tuberculosis,asthma, lung cancer, ischaemic heart disease, and blindness can be attributed to solid fuel usein developing countries. Acute respiratory infections in children under five years of age arethe largest single category of deaths (64%) and DALYs (81%) from indoor air pollution,apparently being responsible globally for about 1.2 million premature deaths annually in theearly 1990s.

Key words: Indoor air pollution developing countries household solid fuel use riskassessment global burden of disease

Introduction

Air pollution has been consistently linked withsubstantial burdens of ill-health in developed anddeveloping countries (Schwartz 1994; WHO 1999;Bruce et al., 2000; Smith et al., 2000). Most recent-ly, as part of the Comparative Risk Assessment(CRA) project of the World Health Organization

(WHO), Geneva, the global and regional burdens ofdiseases (death, illness, injury, lost life years, andyears lost to disability) by age, sex, and 14 worldregions were calculated for a range of risk factors,including indoor and outdoor air pollution (WHO2002). It is the intention here to discuss the history

1438-4639/03/206-04-05-279 $ 15.00/0

Corresponding author: Professor Kirk R. Smith, Environmental Health Sciences, School of Public Health, Universityof California Berkeley, CA 94720-7360, USA. Phone: 5106430793, Fax: 5106425815, E-mail:[email protected]

7/27/2019 SmithKR_2003

2/11

and briefly compare the results of earlier estimates ofthe health effects of indoor air pollution to the latestresults in the WHO CRA.

The bulk of air pollution research has focused onurban outdoor (ambient) air pollution. With therapid increase in vehicular and other pollutionsources in urban areas of developing countries, and

burgeoning numbers of epidemiological studies indeveloped countries showing effects at what used tobe considered low levels (WHO 1999), outdoorsources have remained the center of most airpollution research worldwide. Not surprisingly,then, the first estimate of the global burden ofdisease from air pollution only addressed outdoorair pollution((Hong, 1995), summarized in (Murrayand Lopez 1996)). This endeavor focused on thehealth effects of two ambient air pollutants, totalsuspended particulates and sulfur dioxide, to esti-mate that some 500000 deaths from pneumonia,COPD, cardiovascular diseases, and all causes

combined could be attributable to outdoor airpollution each year. It estimated regional urbanexposures by reference to the WHO/UNEP GEMSurban air pollution database and applied availableexposure-response information to determine im-pacts. Because few exposure-response studies hadbeen done in developing countries, the results ofthose done in China were applied to the rest of thedeveloping world. The counterfactual levels chosenwere the WHO air quality guidelines (WHO 1979).The 2002 WHO CRA used measured and modeledurban particle pollutionas itssole indicatorof health

risk from outdoor air pollution (Cohen et al., 2003).In reality, however, indoor sources of air pollutionalso pose substantial risks and, for some pollutants,probably dominate global human exposure. This isso even though pollutant emissions are dominatedby outdoor sources. Exposures are a function of thedegree of pollution in places were people spend timeand, globally, peoplespend the majority of their timeindoors. As a result, a gram of pollution releasedindoors is likely to cause many hundreds of timesmore exposure than a gram released outdoors.Similarly, even outdoors, a cookfire near the housewill produce much more exposure per unit emissions

than a vehicle or factory some distance away fromplaces where the population spends most time. Thisconcept, originally referred to as, inter alia, ex-posure factor (Smith 1988) or exposure/doseeffectiveness (Smith 1993) is now termed intakefraction(Bennett et al., 2002).

Unfortunately, there are also important indoor

emission sources throughout the world and conse-quent significant exposures. As with outdoor airpollution, however, the bulk of indoor air pollution(IAP) research and control has focused on sources ofconcern in developed countries. Table 1 provides asimple categorization of indoor sources according topollutant category worldwide.

The important non-occupational indoor environ-ments that might be included in a complete IAP CRAwould be households, schools, and passenger com-partments in vehicles. Unfortunately, however, thereare too few exposure and exposure-response studiesto derive reliable global risk estimates for the latter

two microenvironments.Because they contain the largest fraction of time

spent by nearly all populations worldwide, house-hold sources of pollution can dominate exposures.(We are focusing here on indoor sources, not indoorexposures. The latter is influenced by outdoorsources too, of course, since outdoor pollutionpenetrates indoors. Indeed, overall, the major im-pact of outdoor pollution is probably through theindoor exposures it causes, since such a largefraction of the population's time is spent indoors.)Indeed, based on available measurements, it seems

that bulk of global IAP exposures seem to be due tojust two categories: the combustion of solid fuels forcooking/heating and environmental tobacco smoke(ETS). In fact, these sources probably produce moreexposure for several important pollutants than alloutdoor sources. Here we focus solely on solid fueluse. In terms of available epidemiology and otherrisk estimates, however, it is also possible tocalculate the impact of ETS and radon on diseasefor any population where sufficient exposure dataare available. Insufficient exposure data were avail-able at the time of the WHO CRA to do so globally,however.

Table 1. Indoor pollution sources by major pollutant types.

Particles Combustion by-products (CO, NOx)

Volatile organics Biologicals Pesticides Radon

Solid fuel andtobaccocombustion,cleaning

Fuel and tobaccocombustion

Furnishings, householdproducts, solid fuel andtobacco combustion

Furnishings, ventilationducts, moist areas

Household products,dust from outside

Ground underbuilding, ventilationcharacteristics

280 K. R. Smith, S. Mehta

7/27/2019 SmithKR_2003

3/11

About half of the world continues to cook withsolid fuels, such as dung, wood, agricultural resi-dues, and coal. In simple household stoves, thesefuels emit substantial amounts of a number ofimportant pollutants, including respirable particles,carbon monoxide, toxic organic compounds such asbenzene, formaldehyde, and 1,3-butadiene, andpolyaromatic compounds, such as benzo(a)pyrene

(Smith 1987). In households with limited ventila-tion, as is common in many developing countries,exposures to householders, particularly women andyoung children whospend a large proportion of theirtime indoors, have been measured to be many timeshigher than WHO Guidelines and national stan-dards (Smith 1987).

Different approaches used to estimate

environmental burden of diseaseKnown to us are five different methods that havebeen applied to estimate the burden of disease fromsolid fuel use in developing countries, each withadvantages and disadvantages. Given that theirresults are fairly similar, taken together they providesome credibility, although by no means proof, forthe assertion that the problem is severe. Here webriefly summarize the methods and results fromapplication of three of these methods, as done mostlyby others, and then explain the specific approachused in the recent WHO CRA in more detail. A

summary of the different approaches and theirsources of data is given in Table 2.

Pollutant-based approach

This method involves the following steps: 1) Esti-mate total population exposures from indoor sour-ces to the indicator pollutant. Estimates of this type

have relied on particulate matter as the indicatorpollutant and mean exposure concentrations in mg/m3 as the metric. 2) Determine best availableexposure-response factors for this pollutant. 3)Find the current rates of morbidity and mortalityin the population of concern. 4) Estimate theattributable number of deaths and diseases.

Table 3 is a summary of the results from suchefforts done globally and for the two largest nations,India and China. Shown for comparison are esti-mates for outdoor air pollution done using the samemethod. That the exposure-response relationships

have been derived for outdoor air pollution in urbansituations, where the chief source of particulates isfossil fuel burning, however, raises a number ofquestions about their suitability for applicationindoors mainly with rural populations relying onbiomass fuels. In addition, some of the studies relymostly on developed-country studies. These char-acteristics raise several important questions (manyof these same problems plague attempts to calculateimpacts of outdoor air pollution in developingcountries from epidemiologic results in developedcountries (Cohen et al. 2003)): 1) Differences inpollutant mix due to different sources, i.e. although

Table 2. Summary of risk assessment methods applied to solid fuel use in developing countries.

Approach Methodology utilized Type of data utilized Likely bias of method as utilized

Pollutant-based Exposure-response extrapolation Estimated exposure concentrations for indicator pollutants(usually particulates)Exposure-response relationships fromurban outdoor studies, usually in developed countriesCurrent rates of morbidity and mortality

Overestimate

Child survival Survival analysis Survival curves for different risk factors based on householdsurveys Done only for India to date

Overestimate

Cross-national Regression Cross-country comparisons of national-level data on healthand energy conditions

Overestimate

Fuel-based Disease by disease summationfrom binary exposure classifi-cation based on published epi-demiology literature

Estimated distribution of exposure surrogates, usually fueltype Relative risk primarily from developing-country house-hold studies of specific diseases in specific populationgroups experiencing exposure surrogate Current rates ofmorbidity and mortality for each disease

Underestimate

Remainder-based Estimate fraction of burden ofmajor categories of IAP-relatedill-health that can be attributed

to non-IAP factors, the remain-der is applied to IAP

If there are no unaccounted major risk factors, uncertaintiesin the known ones, and no interactions among risk factors,the remainder can be considered a lower bound estimate for

the fraction attributable to IAP Unfortunately, however,these conditions are not likely to be met*

Unknown

* See estimates by Florig (1997) for China in Table 3. This method is not discussed further in this paper.

Indoor air pollution 281

7/27/2019 SmithKR_2003

4/11

particulates can be used as indicator of hazard inboth cases, biomass fuels as commonly used in LDChouseholds produce relatively more organic com-pounds (e.g., benzene, formaldehyde, 1,3-buta-diene, polyaromatic hydrocarbons) and fossil fuelsmore sulfur oxides. Thus risk (exposure-response)estimates derived in the latter situation may notapply to the former. 2) In a similar fashion, thechemical and other characteristics of the particles

produced by biomass combustion are not the sameas those produced by fossil fuel use, although ofcourse woodsmoke is found seasonally in the out-door air of many developed-country cities. 3)Differences in exposure patterns, i.e., indoor con-centrations tend to vary much more during the day,because of household cooking and heating sche-dules, than do outdoor urban levels. 4) Differentexposure levels, i.e., the average exposure levels of

Table 3. Estimates of annual pre-mature mortality from air pollution for world, India, and China using the pollutant-based method.

Location Outdoor Exposure(1000 deaths)

Indoor Exposure(1000 deaths)

Pollutant Comments Reference

World 570 - PM, SOx

First draft done for GBD database using local airmonitoring data and local exposure-responsedata where available

(Hong 1995)

200 2800 PM Using local air pollution monitoring data andMDC exposure-response information at lowexposures and half risk at higher exposures

(WHO 1997),based on(Smith 1994)

510 2200 PM Using local air pollution monitoring data andlocal exposure-response data where available

(WHO 1997),(Schwela 1996)

799 PM Second iteration of GBD, includes only urban airpollution. Relies on econometric model topredict PM levels in those cities without data

(WHO 2002)

India 50 300 850 3300 PM Using urban air quality data and rural exposuresfrom rural microenvironment studies and urbandistribution; MDC exposure-response informa-tion; range comes from spread between dailyand annual studies

(Smith 1994)

40 - PM 36 cities only based on MDC exposure-responsedata

(Brandon andHommann 1995)

86 - PM, SOx Uses Chinese exposure-response data since nonein India

(Hong 1995)

84 590 PM Using local air pollution monitoring data andChinese exposure-response data since none inIndia

(WHO 1997),(Schwela 1996)

200 2000 PM Based on estimates of time and exposures inmajor microenvironments by important popula-tion groups and MDC exposure-response data

(Saksena andDayal 1997)

52 PM Extrapolation of ( Brandon and Hommann 1995)using 1995 air pollution monitoring data

(Kumar, et al. 1997)

107 PM Assumes here that India impacts proportional toits population in the WHR SEAR-D region, i.e.81%

(WHO 2002)

China 68 - PM, SOx

Uses exposure-response data developed in Chi-na.

(Hong 1995)

70 370* PM Uses Chinese exposure-response data. ( WHO 1997)180 110* PM Using exposure-response data from Chinese

cities. Assumes only 13% of rural populationexposed to IAP.

(World Bank 1997)

298 PM Assumes here that China impacts proportional toits population in the WHR WPR-B region, i.e.84%.

(WHO 2002)

17 2 90 720 1 200* PM Based on evaluation of Chinese exposure-re-sponse data for COPD, lung cancer, coronaryheart disease, and childhood ARI. Combinationof exposure-based and pollutant-based data.

(Florig 1997)**

* All these estimates use (Sinton et al., 1996).** Remainder-based approach

COPD

Chronic Obstructive Pulmonary Disease; ARI

Acute Respiratory InfectionsLDC less-developed country; MDCmore-developed country, PMparticulate matter


7/27/2019 SmithKR_2003

5/11

concern in households using unvented biomass fuelsare 10 50 times greater than the levels studied inmost recent urban outdoor studies (Smith 1993). Asis common with toxicants, there may be a diminish-

ing of the effect per unit increase in exposure(shallowing of the exposure-response curve's slope)at these high levels. 5) Different populations, i.e., thepatterns of disease, competing risk factors, and agedistributions differ dramatically between urbandeveloped-country populations, the world's richest,healthiest, and oldest populations, and peopleexposed to indoor air pollution in developingcountries who tend to be the poorest, most stressed,and youngest in the world. 6) The largest number ofdeveloped-country studies are time-series studiesthat determine short-term changes in mortality andother endpoints in association with short-term

changes in air pollution. The implication for long-term health patterns is unclear, however (McMi-chael et al., 1998). 7) The few long-term cohortstudies of outdoor air pollution may be biased byslight misclassification of smokers, because smokingis such a powerful risk factor for the same healthendpoints. Similar concerns exist for other potentialconfounders,the pattern of which is likely to be quitedifferent between developed and developing coun-tries. 8) Research from developed countries has notfocused on some of the most relevant health out-comes for developing countries. In particular, ALRI,the chief single cause of ill-health globally and

probably the major health impact of IAP exposuresworldwide, is not a major cause of mortality indeveloped countries and thus has not been examinedin many studies (Romieu et al., 2002). 9) These morefundamental concerns are in addition to severeconstraints imposed by incomplete information onthe distribution of air pollution levels experiencedindoors worldwide. There have been no studies ofpollution levels in households based on stratifiedrandom sampling designs,forexample(also a problemwith available urban outdoor pollution measure-ments in many parts of the world). 10) Additional

uncertainty is created because those relatively fewparticulate measurements done to date have beenmostly with respect to total particulates, althoughmostof the consistent exposure-responseresults havebeen with regard to smaller size fractions (PM10 orPM2.5, i.e. particles less than 10 mm or 2.5 mm inmean aerodynamic diameter, respectively). 11) Asthere is no possibility of zero exposure, it is necessaryto define an arbitrarycounterfactual level to whichexposures can be reduced.

The likely result of these problems is overestima-tion of impacts. As shown in Table 3, for example,applying this method directly to India results in 2

million annual excess deaths (Saksena and Dayal1997), which is above the available mortality in theappropriate disease categories. To compensate forthis tendency to overestimate, some of the estimates

in the table arbitrarily reduced the exposure-re-sponse slope at higher concentrations, which is not areliable or replicable approach.

Child survival approach

Here we summarize briefly the results of an analysisof National FamilyHealth Survey (NFHS) data doneunder the auspices of the South Asia Office of theWorld Bank. The National Family Health Survey ispart of a routinely collected series of the Demo-graphic and Health Surveys (DHS) funded primarilyby USAID and undertaken in about three dozen

countries that focuses on fertility, family planning,mortality, and child health.Although as yet unpublished, this analysis has

undergone substantial review both inside and out-side the Bank (Hughes and Dunleavy, 2000). Itdetermines the survival curves for Indian children0 5 years under different household conditions,with careful attention to control for potentialconfounders, such as house type, mother's educa-tion, parity, caste, household size, etc. Comparisonsamong the curves thus indicate the impact on childmortality of differences in those conditions. A totalof nearly 60000 children is included in the analysis,

about 3200 of whom died before age 5. (Since thecause of death for newborns is difficult to determineand may be due to quite different risk factors, deathsbefore 7 days are excluded.)

Compared to households with clean fuels, childrenin households using dirty fuels had a substantiallyhigher mortality rate. Indeed, the negative effect ofdirty fuels in the model exceeded that of lack ofprivate water supplies and/or private toilet facilities.The relative risks (RR) for using unclean fuels (here,clean fuels were defined as electricity, kerosene,LPG, biogas, and charcoal) were 2.0 (95% con-

fidence level: 1.4 2.8) and 1.22 (1.004 1.5) forrural and urban children respectively. It is interestingto note the much smaller effect observed in urbanareas. This could be partially because NFHS col-lected information only on the primary fuel used inhouseholds, although a significant proportion ofurban households are known to use a mixture offuels. People living in rural areas would not likelyhave access to as wide a range of fuel types.

Extrapolating to India as a whole using under-fivechild mortalitycalculated from theNational Census,the model indicates potential mortality reductionsfrom a switch to clean fuels as shown in Table 4.


7/27/2019 SmithKR_2003

6/11

Calculated by us in the table arethe rough associatedloss of DALYs, which are equivalent to about 7percent of the national totalor 15 percent of the totallost by children under 5 years. (The DALY (dis-ability-adjusted life year) is one type of QualityAdjusted Life Year (QALY), which is an index

combining mortality and morbidity using lost heal-thy years as the measure. Its derivation and potentialproblems are discussed in (Murray and Lopez1996).) Note the extreme domination of totalDALYs by YLLs (years of life lost) compared toYLDs (years loss to disability), a ratio of 32 : 1. Thisis because most childhood diseases produce rela-tively few days of illness compared to the years oflost life and also because young child illness days areheavily discounted by age weighting in the GBD.

The NFHS questionnaire was not specific enoughto allow this model to determine the mix of causes ofdeath, which is in any case notoriously difficult to

determine by survey. An examination of this mixwould perhaps serve as a test as to biologicalplausibility of attributing these deaths to IAP. Onthe other hand, even if a significant portion was notdue to direct IAP impacts, such as ALRI, it could stillbe causal through two potentially important indirectroutes: 1) pre-natal exposure to the mother leadingto adverse pregnancy outcomes such as low birthweight (LBW). LBW is a risk factor for a range ofchildhood mortality that would not be associateddirectly with air pollution,includingdiarrhea, whichis the chief cause of death in this age group after

ALRI and2

) the unhealthy mother effect, bywhich air pollution impacts on mothers' healthimpair her ability to take good care of her children's'health.

As with all observational studies, there can alwaysbe questions about whether all potential confoun-ders have been sufficiently accounted, but this studyhas made prodigious attempts to do so. In addition,of course, this approach does not address thepotential impacts on other population groups,particularly women.

The India totals from this method are staggering,more than 30% percent of all under five deaths or

equivalent to about 90 percent of all ALRI. Evenconsidering that some of the mortality may beexpressed in indirect pathways that affect diarrheaand other major non-respiratory childhood diseases,it is difficult to accept such large attributablemortality to use of solid fuels alone.

Cross-national comparisons

Another approach is to develop a regression modelof demographic and health statistics cross-nation-ally corrected for confounders as has been done for122 nations by Bloom and Zaidi (2000 rev. 2002).The results of the model are shown in Table 5. Notethat at least in the countries where biomass useaccounts for a significant fraction of energy use thatabout half of the under five childhood mortalitydifference between countries could be attributed todifference in percent of total fuel use from biomass.

Additional analysis indicates that approximately173000 infant deaths can be associated with bio-mass use in India (Bloom and Zaidi 2000 rev.2002).

Until the full method used for this analysis ispublished it is difficult to interpret these results. Ingeneral, of course, such studies suffer from the lackof specificity common to all ecological studies,which examine relationships on a population basiswithout linking exposure and effect at the householdor individual level. In addition, such broad-scaleanalyses must inevitably rely on parameters that are

widely available and thus have a significant chancefor residual confounding. In addition, the exposuremeasure, percent traditional fuel use, is difficulttointerpret with regard to the parameter of particularinterest here because it is percent of total fuel use inthe economy, not just of households.

Fuel-based approach

The following steps summarize the bottom-up(disease by disease) approach taken here to estimatethe burden of disease from indoor pollution in solid-fuel-using households. In parallel to the pollutant-

Table 4. Estimated annual child mortality due to not using clean fuels in India (Hughes and Dunleavy, 2000).

Ages Urban Rural All India YLL* DALY**

7 days to

7/27/2019 SmithKR_2003

7/11

based approach, the fuel-based approach utilizesrelative risk estimates for health outcomes that havebeen associated with exposures to indoor air pollu-tion from solid fuel use. In contrast to the pollutant-based approach, which focuses on specific indicatorpollutants that occur as a result of combustion, thefuel-based approach takes advantage of the largenumber of available LDC epidemiological studies

that have been conducted that treat exposure toindoor air pollution from solid fuel use as a binaryvariable, e.g., using dirty or clear fuel. A descriptionof the methodology used in the fuel-based approachis provided below and in more detail by Smith et al.(2003). Using data from the International EnergyAgency, UN statistical office, World Bank, FAO,national censuses and specific fuel-use surveys indeveloping countries, the sizes of the exposed andnon-exposed populations, which are defined simplyas those using solid fuels and those not, aredetermined by region using a model incorporating

energy, income, and demographic parameters toextrapolate to nations not covered. Using the resultsof epidemiological studies in biomass-burninghouseholds in South Asia, Latin America, sub-Saharan Africa, and elsewhere, and coal-burninghouseholds in China, appropriate risk factors (rel-ative risks) for specific diseases in specific agegroups are determined using formal meta-analysistechniques to combine the results from separatestudies. Such studies are available in sufficientquantity and quality only for three disease endpoints(acute lower respiratory infections, chronic obstruc-tive pulmonary disease, and lung cancer from coal)

in two age/sex groups: adult women and childrenunder 5, who have the highest exposures to stoveemissions. Using the regional population and burdenof disease (death and disability) database from theGBD, the current patterns of these diseases in thesepopulation groups are determined. Using the stan-dard procedure for determining the populationattributable fraction, the total disease burden attrib-

utable to use of household fuels is determined byregion. Using the known mortality-morbidity rela-tionships for specific diseases for each age group ineach region, an estimate of the total lost life yearsand total sick days attributable to indoor airpollution is estimated. By combining the variancein the fuel-use model and the 95% confidenceinterval from the meta-analyses, objective uncer-tainty bounds are calculated.

This approach, although not without weaknesses,substantially reduces all the problems noted abovefor the pollutant-based approach (numbered as

before):1 4

) Being based solely on studies done inbiomass-using households, the differences in pollu-tant mix, particle composition, exposure patterns,and exposure levels should be substantially reducedif not eliminated. 5) The studies were all done inpoor, mostly rural, developing-country populationspresumably much more similar to the exposed LDCpopulations than urban developed-country popula-tions. 6) The studies address directly the specifichealth endpoints over time periods appropriate tothe each and thus do not reflect the possibleharvesting that may be seen in time-series studies.7) Confining the assessment to women, who have

Table 5. Association of demographic indicators and biomass use (Bloom et al., 2002).

Dependent variable Constant* Percent Log Inverse Log R2***Traditional GNP GNPbiomass use per capita per capita

Female life expectancy 231.647** 0.102** 6.234 889.108** 0.86Life expectancy 213.215** 0.088** 5.453 816.024** 0.86Male life expectancy 195.649** 0.076** 4.708 746.366** 0.84Infant mortality rate 795.305** 0.247** 37.945** 4203.384** 0.83Under 5 mortality rate 1377.804** 0.494** 67.613** 7066.313** 0.82Total fertility rate 0.011 0.025** 0.213 37.326 0.78Crude birth rate 66.412 0.176** 5.336 6.581 0.77Crude death rate 227.919** 0.007 13.247** 1044.726** 0.54Population growth rate 3.031 0.021** 0.184 3.546 0.43Life expectancy gap ( F M ) 35.998 0.026** 1.526 142.741 0.35

* point where regression line crosses the axis** indicates statistical significance at the 5% level*** square of the regression coefficient indication fraction of variation in the health parameter explained by the variation in biomass fuel useData from 1993 and surrounding years. Traditional fuel includes fuelwood, bagasse, charcoal, animal wastes, vegetable wastes, and other wastes. Traditional fuel use isexpressed as a percentage of total fuel use.Source: Traditional fuel use data from United Nations Energy Statistics Yearbook 1993.Demographic data from World Development Indicators 1998, World Bank CD ROM.


7/27/2019 SmithKR_2003

8/11

very low smoking rates in most rural LDC areas, andchildren under 5, greatly reduces possible confound-ing by smoking. 8) The diseases studied are the mostimportant ones in developing countries. 9, 10) The

epidemiological studies used rely on binary exposurevariables, i.e., exposed or less-exposed, it is notnecessary to extrapolate quantitative pollution ex-posures from incomplete data or to estimate therelative contribution of different particle size frac-tions. 11) Since the epidemiological studies compareactual exposed versus less-exposed populations(with different stoves or fuels), there is no need todefine an arbitrary counterfactual value.

As there is a spectrum of quality and quantity ofevidence for different disease endpoints, the methodcategorizes the available evidence into strong, mod-erate, and inconclusive. Only diseases with strong

evidence (ALRI in children under 5; COPD and lungcancer from coal smoke) are included in the finalresults of the CRA (Smith et al., 2003). Althoughthere is also evidence for TB, cataracts, asthma, andadversepregnancyoutcomes, because theevidence isnot yet as convincing, these endpoints are notincluded. The method is explained in more detailby Smith et al. (2000, 2003). It has remaining

weaknesses, however. Some of these would tend tobias the estimates upwards, for example residualconfounding, although others would tend to lead tounderestimates, such as exposure misclassification.

In general, of course, since the method only ad-dresses effects in children under 5 years and adultwomen, it tends to be an underestimate of thepopulation total. Since these two groups experiencethe greatest exposures, it does not seem the resultingunderestimate is likely to be large. Perhaps the mostimportant possibility for underestimation stemsfrom the method's current inability to address theeffects of in utero exposure on birth outcomes, suchas low birth weight, that might affect overall child(and adult) disease burdens. This inability is due to alack of available studies on these endpoints. Inaddition, there seem to be no indoor studies in

developing countries examining cardiovascular dis-ease, which is known to be one of the majoroutcomes from outdoor air pollution.

To put the results of the WHO CRA for indoor airpollution using the fuel-based method into perspec-tive, the global burden of disease for the top 10 riskfactors plus selected other environmental risks arecompared in Figure 1. Indoor air pollution ranks

Fig. 1. Global burden of disease from top 10 risk factors plus selected other risk factors according to the World Health Reports 2001, 2002.


7/27/2019 SmithKR_2003

9/11

10th in burden (DALYs) and 11th in prematuredeaths. It is the second most important environ-

mental risk factor examined (after poor water/sanitation/hygiene), with about twice the deathsand five times the DALYs of urban outdoor airpollution. We have added two other major riskfactors not found in (WHO 2002) road trafficaccidents and the child cluster of vaccine-treatablediseases (measles, tetanus, polio, pertussis, anddiphtheria) to provide a more complete list of therisk factors amenable to publicaction. Data fortheselatter two risk factors are taken from (WHO 2001).If these two were left out, indoor air pollution wouldrank 8th globally.

Conclusion

Figure 2 summarizes the approaches used for thethree most viable methods discussed above andTable 6 shows their results for India, which is theonly country in which all three were examined. Alsoshown for comparison are the results for China andthe World from the fuel-based method so as tocompare with previous estimates in Table 3.

As utilized, most methods except for the fuel-based approach arelikelyto result in an overestimate

of disease burden. In the pollutant-based approach,exposure response relationships from ambient airpollution studies in developed countries with differ-ent epidemiological profiles have been linearlyextrapolated to concentrations that are often ordersof magnitude higher. The child-survival approachused a remarkably high baseline estimate of mortal-ity (over 5% mortality in children under five). Inaddition, the child survival approach, as well as thecross- national approach, both estimate diseaseburdens based on all cause mortality. As notedbefore, this results in higher values, because it picks

up indirect effects. In addition, it is extremelydifficult to separate out the effects of other socio-economic risk factors, such as access to water andsanitation,as well as povertyas a risk factor in and ofitself.

The results of the fuel-based approach, i.e. theestimates in the WHO CRAfor the world, India, andChina, are less, sometimes much less, than thosederived in pollutant-based approaches summarizedin Table 3. As noted in the text, the CRA approachwould seem to avoid many problems inherent in thepollutant-based methods applied to date. At thesame time, the inability to address potentially

Fig. 2. Comparison of methods discussed in text.

Table 6. Deaths from household solid fuel use estimated by different methods (uncertainty bounds not shown).

Method Region Infant Deaths Under 5 deaths Total deaths Reference

Child survival India 395 000 570 000 ( Hughes and Dunleavy 2000)Cross national India 173 000 ( Bloom and Zaidi 2000 rev. 2002)Fuel-based India 287 000 424 000 ( WHO 2002)Fuel-based China 52 000 423 000 ( WHO 2002)Fuel-based World 910 000 1 600 000 ( WHO 2002)


7/27/2019 SmithKR_2003

10/11

important health outcomes, including adverse preg-nancy outcomes and cardiovascular disease, is likelyto result in a substantial underestimate of diseaseburden.

Note that China and India have quite differentmixtures of adult deaths, which are mostly fromCOPD, and children, which are from ALRI. This isat least partly due to better access to health care inChina such that ALRI kills many fewer children. Incontrast to India, however, China has a much higherCOPD rate in non-smokers. This may have to dowith the differential impacts in China of coal smoke,which is much less prevalent in India.

The different exposure methods employed, andthe different health outcomes addressed make adirect comparison of the approaches difficult.Agreement between the results would not necessa-

rily guarantee the accuracy of the estimates, but doeslend some credibility that the burden is significant.With the most conservative estimate suggesting thataround 1.6 million deaths can be associated withindoor air pollution each year, solid fuel use clearlyremains one of the major risk factors facing human-ity today.This paper is an updated and shortened version of theconference paper of the same name and authors given atthe WHO/USAID Global Technical Consultation HealthImpacts of Indoor Air Pollution in Developing Countries,May 3 4, 2000.

References

Bennett, D., McKone, T., Evans, J., Nazaroff, W.,Margni, M., Jolliet O. and Smith K. (2002). Definingintake fraction. Environ Sci and Technol 36: 207A 211A.

Bloom, D., Rosenfield A. and Malaney P. (2002). TheQuality of Life in Rural Asia. Hong Kong, OxfordUniversity Press.

Bloom, D. and Zaidi A. (2000 rev. 2002). Thedemographic impact of biomass fuel use. Internation-al Conference on Environmental and Occupational

Respiratory Disease, Lucknow, India, Industrial Tox-icology Reserch Institute.Brandon, C. and Hommann K. (1995). The cost of

inaction: valuing the economy-wide cost of environ-mental degradation in India. Conference on theSustainable Future of the Global System, Tokyo, UNUniversity.

Bruce, N., Perez-Padilla R. and Albalak R. (2000).Indoor air pollution in developing countries: a majorenvironmental and public health challenge for thenew millennium. Bulletin of the World HealthOrganization 78 (9): 1078 1092.

Cohen, A. J., Anderson, H. R., Ostro, B., Pandey, K. D.,Krzyzanowski, M., Kuenzli, N. Gutschmidt, K., Pope,

C. A., Romieu, I., Samet, J. M., Smith, K. R.: Mor-tality Impacts of Urban Air Pollution. In: Compara-tive Quantification of Health Risks: Global andRegional Burden of Disease due to Selected MajorRisk Factors (Ezzati, M., Rodgers A. D., Lopez A. D.,

Murray, C. J. L. eds.). World Health Organization,Geneva, 2 volumes, in press 2003.

Florig, H. K. (1997). China's air pollution risks.Environ Sci Tech 31 (6): 276A 279A.

Hong, C. (1995). Global Burden of Disease from AirPollution. Geneva, World Health Organization.

Hughes, G. and Dunleavy M. (2000). Why do babies andyoung children die in India? The role of the householdenvironment. Washington, DC, South Asia Office,World Bank.

Kumar, P. et al. (1997). Death is in the air. Down ToEarth 6 (12): 29 43.

McMichael, A. J., Anderson, H. R., Brunekreef B. andCohen A. J. (1998). Inappropriate use of daily

mortality analyses to estimate longer-term mortalityeffects of air pollution. International Journal ofEpidemiology 27: 450 453.

Murray, C. and Lopez, A., Eds. (1996). The GlobalBurden of Disease: A Comprehensive Assessment ofMortality and Disability from Diseases, Injuries, andRisk Factors in 1990 and Projected to 2020. GlobalBurden of Disease and Injury Series. Cambridge,Harvard School of Public Health on behalf of theWorld Health Organization and the World Bank.

Romieu, I., Samet, J., Smith K. and Bruce N. (2002).Outdoor air pollution and acute respiratory infec-tions among children in developing countries. J.Occupational and Environmental Medicine 44 (7):640649.

Saksena, S. and Dayal V. (1997). Total exposure as abasis for the economic valuation of air pollution inIndia. Energy Environment Monitor 13 (2): 93 102.

Schwartz, J. (1994). Air pollution and daily mortality:A review and meta analysis. Environ Res 64: 36 52.

Schwela, D. (1996). Exposure to environmental chemi-cals relevant for respiratory hypersensitivity: Globalaspects. Toxicology Letters 86 (2 3): 131 142.

Sinton, J. E., Smith, K. R., Hu H. and Liu J. (1996).Indoor Air Pollution Database for China. Geneva,WHO/EHG/95.8.

Smith, K. (1987). Biofuels, Air Pollution, and Health: A

Global Review. New York, Plenum.Smith, K. (1988). Total exposure assessment: Part 1,Implications for the U. S. Environment 30 (8): 10 15; 22 38.

Smith, K. (2000). National burden of disease in Indiafrom indoor air pollution. Proceedings of the Na-tional Academy of Sciences 97 (24): 13286 13293.

Smith, K. R., Mehta S. and Feuz M. (2003). Indoorsmoke from household solid fuels. In: ComparativeQuantification of Health Risks: Global and RegionalBurden of Disease due to Selected Major Risk Factors(Ezzati M., Rodgers A. D., Lopez A. D. and MurrayC. J. L., eds.). World Health Organization, Geneva, 2volumes, in press 2003.


7/27/2019 SmithKR_2003

11/11

Smith, K. R. (1993). Fuel combustion, air pollutionexposure, and health: the situation in developingcountries. Ann Rev Energy Environ 18: 529 566.

Smith, K. R. (1994). Health risks from air pollution inIndia: Implications of new studies. Workshop on the

Energy-Environment Nexus: Indian Issues and GlobalImpacts, Philadelphia, Center for the Advanced Studyof India, Univ of Penn.

Smith, K. R., Samet, J. M., Romieu I. and Bruce N.(2000). Indoor air pollution in developing countriesand acute lower respiratory infections in children.Thorax 55: 518 532.

WHO (1979). Sulfur Oxides and Suspended ParticulateMatter. Environmental Health Criteria #8. Geneva,World Health Organization.

WHO (1997). Health and Environment in SustainableDevelopment. Geneva, World Health Organization.

WHO (1999). Global Air Quality Guidelines. Geneva.WHO.

WHO (2001). World Health Report. Geneva, World

Health Organization.WHO (2002). World Health Report: Reducing Risks,

Promoting Healthy Life. Geneva, World HealthOrganization.

World Bank (1997). Clear Water, Blue Skies: China'sEnvironment in the New Century. Washington, DC,World Bank: 114.


smithkr_2003

Documents