We used population-attributable fractions to quantify the impact of air pollution on asthma-related outcomes in LAC for year 2007 for children < 18 years of age. We followed an existing methodological framework (Künzli et al. 2008
; Perez et al. 2009
) that we adapted for this new study as summarized below.
To estimate the prevalence of asthma attributable to near-roadway pollution exposure, we used a concentration–response function (CRF) from the Children’s Health Study (CHS), a large population-based cohort in Southern California, in which living near major roadways, a proxy for traffic-related pollution exposure, was associated with increased prevalence of asthma (McConnell et al. 2006
). Details on CHS study design and recruitment methods have been published previously (McConnell et al. 2006
; Peters et al. 1999
). To be consistent with the exposure assignment of the CRF study, we used the TeleAtlas MultiNet roads network (http://www.tomtom.com/en_gb/licensing/products/maps/multinet/
) to map major LAC roads, defined as freeways, highways, or major arterial roads. These were then linked to census population data to derive the percentage of persons living within 75 m of these roads. For the present study we linked exposure to census population data given at the parcel level, which increased accuracy relative to linkage at the census block level used in a previous analysis (Perez et al. 2009
). To be consistent with the prior CRF outcome definition, we used as background risk the asthma prevalence reported in the CHS (defined by use of controller medications in the previous year or lifetime asthma with any wheeze in the previous year or severe wheeze in the previous 12 months).
Regional pollutants including particulate matter, nitrogen dioxide (NO2
) and ozone (O3
) are among the many causes of acute exacerbation among children with asthma, regardless of the cause of asthma onset (Jackson et al. 2011
). However, an important consideration is that among those children with asthma attributable to living near a major road, all subsequent exacerbation should be attributed to air pollution, regardless of the trigger, assuming that these children would not otherwise have had the disease (Künzli et al. 2008
). Conceptually, the total burden of asthma due to near-source and regional pollution includes the number of yearly asthma exacerbations triggered by causes other than regional air pollution among children whose asthma was caused (at least in part) by near-roadway pollution (). These exacerbations are in addition to those directly triggered by regional air pollution exposure among all children with asthma, including children whose asthma was caused by near-roadway exposure and children whose asthma was caused by something other than traffic proximity. Air pollution risk assessments typically calculate only the asthma exacerbation burden triggered directly by regional pollution exposures, regardless of the underlying cause of asthma, whereas we included the additional burden of disease among children with asthma caused by near-roadway exposure but with exacerbations triggered by factors other than regional pollution.
Figure 1 Conceptual model used to calculate asthma-related exacerbation attributable to air pollution for Los Angeles County based on Künzli et al. (2008). The thick dashed line indicates children with asthma attributable to near-roadway exposure. (more ...)
To avoid double counting the burden associated with correlated regional pollutants, we estimated exacerbation attributable to NO2
was selected to represent urban-scale combustion-related pollution because it is correlated with particulate mass and other regional pollutants associated with respiratory health effects in southern California (Gauderman et al. 2004
is produced as a result of photo-oxidation that is uncorrelated with other regional pollutants in the Los Angeles air basin (Gauderman et al. 2004
The CRFs for bronchitis episodes among those with asthma, and for prevalent asthma attributable to near-roadway exposure, were derived from the CHS (McConnell et al. 2003
) (). CRFs from appropriate studies of Southern California populations were not available for doctor visits, emergency department visits, or hospital admissions. Therefore, we applied CRFs used in a previous Southern California health impact assessment (Perez et al. 2009
) or averaged the coefficient used in the previous analysis with the coefficient from a more recent study conducted in a similar population, as indicated in .
Concentration–response functions (CRF) with 95% confidence intervals (CI) considered in the evaluation of air pollution burden.
The number of children < 18 years of age (> 2.5 million) was obtained from the American Community Survey (U.S. Census Bureau 2011). Background rates of the outcomes were obtained from the CHS or from local surveys (). Annual average daily concentrations of NO2 and O3 obtained from the 2007 U.S. Environmental Protection Agency Air Quality System (AQS) (U.S. Environmental Protection Agency 2009) and CHS monitoring stations were interpolated based on inverse distance-squared weighting to each census block group in the county to estimate population exposures. Because of the seasonality of school attendance and both the seasonal and day-of-week variability of O3, the O3 population exposure for school absences was based on 2007 daily maps, rather than annual maps, obtained from interpolated hourly ambient school-week concentrations projected to 2000 census block group centroids.
Population size and baseline health outcome and exposure estimates used to evaluate the burden of asthma due to air pollution in LAC in 2007.
We estimated that 17.8% of LAC children lived within 75 m of major roads, and that the annual population-weighted exposure to NO2
was 23.3 ppb (24-hr average) and to O3
was 39.3 ppb (8-hr maximum) in LAC (). We assumed background concentrations of 4 ppb for NO2
annually and 38 ppb for 8-hr maximum O3
on all days, based on long-term measurements (1994–2003) from CHS monitoring stations in clean coastal locations (i.e., Lompoc, CA) (McConnell et al. 2003
). [The average population-weighted annual O3
for LAC was near background because population exposures in the areas with high O3
are offset by population exposures in areas with high oxides of nitrogen (NOx
) emissions and very low O3
concentrations, due to nitric oxide (NO) in fresh vehicular exhaust scavenging O3
in those areas.] We considered three near-roadway proximity exposure reduction scenarios ():
Exposure reduction scenarios for near-roadway exposure, regional NO2 and O3, and corresponding reduction in childhood asthma cases attributable to near-roadway pollution exposure (based on total of 320,500 children with asthma in LAC).
- A reduction in annual concentrations of regional pollutants for each census block group to levels found in clean CHS communities (from 23.3 ppb to 4 ppb for NO2 and 39.3 ppb to 36.3 ppb for O3) in combination with a reduction in the proportion of children in the county living within 75 m of a major road from 17.8% to 0%
- A 20% reduction in the annual concentrations of regional pollutants for each census block group (from 23.3 ppb to 19.4 ppb for NO2 and 39.3 ppb to 38.7 ppb for O3) in combination with a 3.6% reduction in the proportion of all children in the county living within 75 m of a major road (from 17.8% to 14.2%, corresponding to a 20% decrease in the proportion of children currently living within 75 m)
- A 20% reduction in regional pollutant concentrations in combination with a 3.6% increase in the proportion of children living within 75 m of a major road (from 17.8% to 21.2%).
Scenario 1 reflects the total burden of preventable illness from both exposures. At this time there is considerable uncertainty regarding the potential impact of compact urban growth strategies on near-roadway exposures, so scenarios 2 and 3 were selected assuming moderate reductions in regional pollutants from continued regulatory efforts and a moderate 20% increase or decrease in near-roadway exposure—a value that was chosen for illustration and could be refined using data from regional planners as they become available. Regional pollutant concentrations aggregated to the census block group level that exceeded background levels were reduced linearly, whereas we assumed that concentrations at or below the background level would be unaffected by changes in emissions.
There are intrinsic limitations and uncertainties in risk analysis. We estimated a 95% confidence interval (CI) derived from the propagation of the CIs for the CRFs to address uncertainty in these estimates. In addition, proximity to major roadways has uncertainty as a proxy for near-roadway pollution exposure that depends on traffic volume, the emissions of the vehicular fleet, and local meteorological factors. Therefore, we also estimated the total burden of asthma-related exacerbations associated with a 100% and a 20% reduction in population-weighted exposure to the near-roadway dispersion-modeled pollution mixture (instead of a change in roadway proximity in exposure scenarios 1 and 2 in ) using the CHS CRF from an estimate of the association of asthma prevalence with dispersion-modeled near-roadway pollution exposure accounting for traffic volume and emission factors (McConnell et al. 2006
). Specifically, we used modeled NOx
to represent the incremental contribution of local traffic to a more homogeneous community background concentration of NOx
that included both primary and secondary pollution resulting from long-range transport and regional atmospheric photochemistry. It is a marker for correlated pollutants in the near-roadway mixture (rather than the etiologic agent for near-roadway health effects). We developed new estimates of population-weighted yearly average of local traffic-related NOx
concentrations for 2007 in LAC using the CALINE4 dispersion model with the 2007 TeleAtlas MultiNet Roadway network, and 2007 vehicle emission factors for Los Angeles from the EMFAC model (California Air Resources Board 2007
). Vehicle emission factors were developed for winter (55o
F/50% relative humidity) and summer (75o
F/50% relative humidity) conditions using average speeds of 65 mph on freeways and highways [FCC (functional class code) 1 and FCC2 class roads], 50 mph on major arterials (FCC3 class roads), and 30 mph on minor arterials and collectors (FCC4 roads). The model used year 2000 traffic volumes adjusted to 2007 VMT provided by the California Department of Transportation (17.5% increase in VMT for LAC). Modeled NOx
concentrations were estimated for the block group centroids. The CHS CRF was developed for the contribution of local traffic on non-freeways using an older road functional roadway classification (FRC) scheme (McConnell et al. 2006
) that is no longer available in a form that matches the most current FCC classification that we used. To minimize overestimation of population exposure to near-roadway exposure in LAC, we used estimates of exposure from FCC3 (major arterials) as representative of non-freeway roads used in developing the CHS CRF. We considered the impact of all near-FCC3 roadway NOx
(corresponding to a scenario of 100% reduction in modeled near-roadway pollution at the block group centroid) and of a 20% decrease in population exposure. This corresponds to the 3.6% reduction in the total population of children within 75 m of a major roadway (a 20% reduction the proportion of the total population living within 75 m) ().