Secondary Logo

Share this article on:

Indoor Exposure to “Outdoor PM10”: Assessing Its Influence on the Relationship Between PM10 and Short-term Mortality in U.S. Cities

Chen, Chuna; Zhao, Bina; Weschler, Charles J.a,b,c

doi: 10.1097/EDE.0b013e31826b800e
Air Pollution

Background: Seasonal and regional differences have been reported for the increase in short-term mortality associated with a given increase in the concentration of outdoor particulate matter with an aerodynamic diameter smaller than 10 μm (PM10 mortality coefficient). Some of this difference may be because of seasonal and regional differences in indoor exposure to PM10 of outdoor origin.

Methods: From a previous study, we obtained PM10 mortality coefficients for each season in seven U.S. regions. We then estimated the change in the sum of indoor and outdoor PM10 exposure per unit change in outdoor PM10 exposure (PM10 exposure coefficient) for each season in each region. This was originally accomplished by estimating PM10 exposure coefficients for 19 cities within the regions for which we had modeled building infiltration rates. We subsequently expanded the analysis to include 64 additional cities with less well-characterized building infiltration rates.

Results: The correlation (r = 0.71 [95% confidence interval = 0.46 to 0.86]) between PM10 mortality coefficients and PM10 exposure coefficients (28 data pairs; four seasons in each of seven regions) was strong using exposure coefficients derived from the originally targeted 19 National Morbidity, Mortality, and Air Pollutions Study cities within the regions. The correlation remained strong (r = 0.67 [0.40 to 0.84]) when PM10 exposure coefficients were derived using 83 cities within the regions (the original 19 plus the additional 64).

Conclusions: Seasonal and regional differences in PM10 mortality coefficients appear to partially reflect seasonal and regional differences in total PM10 exposure per unit change in outdoor exposure.

Supplemental Digital Content is available in the text.

From the aDepartment of Building Science, School of Architecture, Tsinghua University, Beijing, China; bEnvironmental and Occupational Health Sciences Institute, University of Medicine and Dentistry of New Jersey (UMDNJ)–Robert Wood Johnson Medical School and Rutgers University, Piscataway, NJ; and cInternational Centre for Indoor Environment and Energy, Technical University of Denmark, Lyngby, Denmark.

Submitted 29 January 2012; accepted 8 May 2012.

Supported by the National Natural Science Foundation of China (Grant No. 51078216) and the Tsinghua University Initiative Scientific Research Program.

Supplemental digital content is available through direct URL citations in the HTML and PDF versions of this article (www.epidem.com). This content is not peer-reviewed or copy-edited; it is the sole responsibility of the author.

The authors report no conflict of interest.

Correspondence: Bin Zhao, Department of Building Science, School of Architecture, Tsinghua University, Beijing, China. E-mail: binzhao@tsinghua.edu.cn.

Numerous epidemiologic studies have shown an association between outdoor PM10 concentration and short-term mortality.1–7 However, a large fraction of exposure to outdoor PM10 occurs indoors.8 A given increase in the concentration of outdoor PM10 in different cities can result in varying increases in total PM10 exposure (sum of outdoor and indoor exposures), owing to various parameters that influence indoor exposure to outdoor pollution. Important factors include the air change rate (the rate at which indoor air is replaced with outdoor air), the size distribution of outdoor PM10 at the time it is transported indoors, the fraction of residences with central air conditioning (AC), the fraction of time that cooling and heating occurs, and the amount of time that residents spend indoors.

Peng et al7 estimated the seasonal percent increase in short-term mortality per 10 μg/m3 increase in PM10 (PM10 mortality coefficient) for seven U.S. regions using data from 100 U.S. cities included in the National Morbidity, Mortality, and Air Pollutions Study (NMMAPS). Their estimates for the percent change in daily mortality per 10 μg/m3 increase in daily PM10 ranged from −0.19% for the Upper Midwest in the fall to 0.92% for the Northeast in the summer. Several investigators have reported a modifying effect of AC on the association between PM10 and short-term mortality.9–15 However, these studies did not systematically examine the possibility that the modifying effect of AC may be owing to decreased outdoor-to-indoor transport of PM10 coupled with filtration of recirculated air, reducing the indoor exposure to PM10.

Sheppard et al16,17 have noted that variation in the concentration-response effect estimates in air pollution time-series studies is partially owing to differences in population exposures. Recently, we assessed the influence of indoor exposure to ozone of outdoor origin on the relation between ozone and short-term mortality in U.S. communities.18 Estimates of outdoor-to-indoor transport of ozone were based, in part, on distributions of residential air infiltration rates in 19 cities chosen to represent different U.S. climatic conditions.19 The results suggest that city-to-city differences in total ozone exposure per unit change in outdoor exposure (ozone exposure coefficient) partially explain differences in ozone mortality coefficients among these cities.

The present analysis was prompted by the analysis of Peng et al7 reporting seasonal and regional differences in PM10 mortality among the NMMAPS cities, the detailed estimates of infiltration rate distributions for the U.S. housing stock in representative U.S. cities,19 and the strong association18 that we previously observed between ozone exposure coefficients and mortality coefficients. Our aim is to examine whether differences reported for PM10 mortality coefficients in the various seasons in seven U.S. regions7 might be partially explained by seasonal and regional differences in total PM10 exposure per unit change in outdoor exposure (PM10 exposure coefficient). We have addressed this question by (1) estimating seasonal PM10 exposure coefficients for 19 NMMAPS cities whose infiltration rates have been modeled by Persily et al,19 (2) extending these estimates to 83 NMMAPS cities (the original 19 plus 64 with less well-characterized infiltration rates), (3) using these seasonal estimates for the various cities to calculate seasonal PM10 exposure coefficients for each region, and (4) examining potential correlations between PM10 mortality coefficients for each season in each region and corresponding PM10 exposure coefficients.

Back to Top | Article Outline

METHODS

Cities and Regions

We initially focused on 19 U.S. cities included in the NMMAPS study and also selected by Persily et al19 to represent different climatic regions of the United States. We then extended the analysis to 64 additional NMMAPS cities with climatic conditions and housing stock similar to one of the 19 cities used in the original analysis. Figure 1 shows the locations of these 83 cities as well as the seven geographical regions used in this analysis.

FIGURE 1

FIGURE 1

FIGURE 2

FIGURE 2

Back to Top | Article Outline

Changes in Indoor PM10

In all residences, when windows are closed and central AC (if present) is off, the monthly change in indoor PM10 concentration per change in outdoor PM10 concentration, {Δ[PM10]in/Δ[PM10]out}windows_closed, can be approximated using:

where P is the particle penetration factor, λinfilt is the average annual infiltration rate, and k sr,infilt is the surface removal rate constant when windows are closed.

In all residences, the monthly change in indoor PM10 concentration per change in outdoor PM10 concentration when the windows are open, {Δ[PM10]in/Δ[PM10]out}windows_open, can be approximated using:

where λwin_open is the air change rate when windows are open, and k sr,open is the surface removal rate constant when windows are open.

In residences with central AC, the monthly change in indoor PM10 concentration per change in outdoor PM10 concentration when the central AC is operating, {Δ[PM10]in/Δ[PM10]out}AC_on, can be approximated using:

where η is the filtration efficiency of the filter installed in the central AC system, and λrecirc is the air change rate due solely to air recirculated by the central AC system. The average change in indoor PM10 concentration per change in outdoor PM10 for a given city in a given month can be approximated using a combination of Equations (1)–(3):

where x cool, x heat, and x mild are, respectively, the fraction of time that cooling, heating, and mild weather occurs in a given month; z is the average fraction of time that windows are open in mild weather, and y is the fraction of residences with central AC. Details regarding values used for the input parameters are provided in the subsections “City-specific parameters” and “Fixed parameters.”

Back to Top | Article Outline

PM10 Exposure Coefficients

We define exposure in a specific microenvironment as the product of time spent in the microenvironment and pollutant concentration in the microenvironment during that time. The change in total PM10 exposure is the sum of the change in PM10 exposure in each of the microenvironments that a person spends a fraction of their time; it is approximated in this study as the sum of the changes in outdoor exposure and indoor exposure at home. The change in total PM10 exposure per unit change in outdoor PM10 exposure, PM10 exposure coefficient, is given by

where t in is the fraction of time indoors and t out the fraction outdoors. This equation allows us to estimate how, in various cities, the total PM10 exposure changes when the outdoor PM10 exposure changes by a given amount. The derivation of Equation 5 is outlined in page 4 of the eAppendix (http://links.lww.com/EDE/A607).

For each U.S. region, the monthly PM10 exposure coefficient was obtained by population-weight averaging monthly PM10 exposure coefficients of the representative U.S. cities in that region. The 19 cities originally targeted and 83 cities used in the extended analysis were distributed among the regions as shown in Figure 1. Seasonal PM10 exposure coefficients were obtained by averaging the relevant monthly PM10 exposure coefficients.

Back to Top | Article Outline

City-specific Parameters

Average annual infiltration rate (λinfilt) was estimated starting with the percent of hours below five key infiltration rates as reported in Table 8 of the article by Persily and colleagues.19 Assuming that these infiltration rates were log-normally distributed, we calculated an average annual infiltration rate for each of the 19 NMMAPS cities. Further details are provided in the report of our previous study.18

Fraction of time that cooling occurred in a given month (x cool) was estimated using an approach based on the monthly maximum and minimum temperature throughout a statistical year. We assumed that cooling occurred if the temperature was higher than 24°C. Further details are provided in our previous report.18

Fraction of time that heating occurred in a given month (x heat) was estimated using a method analogous to that used for estimating x cool, but with an assumed critical temperature of 15°C.

Fraction of time that mild weather occurred in a given month (x mild) was simply the fraction of time with neither cooling nor heating (ie, 1 − x coolx heat).

Fraction of residences with central AC (y) was taken from the data set assembled by R. L. Smith (personal communication) and described in the study by Smith and colleagues.20

Penetration factor (P) and surface removal rate constant (k sr) were determined taking into account the size distribution of outdoor PM10. Detailed information on the size distributions of PM10 in the cities of interest was not available. However, city-specific mean concentrations of PM10 and PM2.5 (particulate matter smaller than 2.5 μm aerodynamic diameter) are available at the website of the Internet-based Health & Air Pollution Surveillance System (http://www.ihapss.jhsph.edu). This information was used to estimate city-specific PM10 penetration factors as well as surface removal rate constants when windows were closed (k sr,infilt) and open (k sr,open). Details are provided in page 2 of the eAppendix (http://links.lww.com/EDE/A607).

Fraction of time indoors (t in) and outdoors (t out) was set as the region-specific values from the National Human Activity Pattern Survey.21 The ratio of time indoors to time outdoors (t in/t out) is provided in eTable 2 (http://links.lww.com/EDE/A607).

Back to Top | Article Outline

Fixed Parameters

The air change rate when windows are open (λwin_open) was assumed to be 2.3 h 1.22,23 The filtration efficiency of the filter installed in the central AC system (η) was estimated to be 0.1.24,25 The air change rate due solely to air recirculated by the central AC system (λrecirc) was assumed to be 4.0 h 1.24 The average fraction of time windows are open in mild weather (z) was assumed to be 0.23.26 Further details regarding the parameter, z, are provided in page 3 of the eAppendix (http://links.lww.com/EDE/A607).

Back to Top | Article Outline

PM10 Mortality Coefficients

Seasonal PM10 mortality coefficients for the seven U.S. regions were based on plots of these coefficients versus day-of-year presented in Figure 3 of the report by Peng et al7 and are listed in eTable 1. They correspond to the percent increase in short-term mortality per 10 μg/m3 increase in outdoor PM10 at a lag of 1 day based on data from 100 NMMAPS cities, which have been adjusted for outdoor temperature and other pollutants (sulfur dioxide, ozone, and nitrogen dioxide). Further details regarding the calculation of these PM10 mortality coefficients are provided in the article by Peng and colleagues.7 The analysis was based on daily cause-specific mortality counts for 1987–2000. The population was divided into three age groups (<65, 65–74, and >75 years). The daily average mortality counts ranged from two deaths per day in Arlington, VA, to 190 deaths per day in New York City, NY. The daily mean of PM10 ranged from 13 μg/m3 in Coventry, RI, to 49 μg/m3 in Fresno, CA. The average daily temperature ranged from 3°C to 25°C.

FIGURE 3

FIGURE 3

Back to Top | Article Outline

Exploring Correlations for Other NMMAPS Cities

In Table A1 of the article by Vandemusser Design,27 U.S. cities with a population greater than 250,000 were paired with one of the original 19 cities for the purpose of modeling ventilation rates. From this listing we have selected an additional 64 NMMAPS cities for which the match in terms of climate and housing stock is reasonably close. Each was assigned a value for λinfilt equal to that of its paired city, whereas x cool, x heat, P, and k sr were calculated using the same procedure as that for the original 19 cities. The fraction of residences with central AC (y) was taken from the data set assembled by R. L. Smith (personal communication) and described in Smith.20

Back to Top | Article Outline

Sensitivity Analysis

We have examined the sensitivity of our results to five parameters that are difficult to estimate—the fraction of time in cooling mode (x cool), the air change rate when windows are open (λwindow_open), the air recirculation rate during central AC operation (λrecirc), the efficiency of the filter installed in the central AC system (η), and the fraction of time that windows were open in mild weather (z). This was accomplished by determining correlations between PM10 exposure coefficients and PM10 mortality coefficients, when the former are calculated with what we judge to be reasonable lower and upper limits for these five parameters. Justifications for the lower and upper bounds are presented in page 6 of the eAppendix (http://links.lww.com/EDE/A607).

An additional sensitivity evaluation involved mean and median values of outdoor PM10 and PM2.5. Our initial estimates of PM10 exposure coefficients were obtained using mean values; these were recalculated using median values to examine the impact on the reported correlations.

Back to Top | Article Outline

Correlations

Pearson correlation coefficients (r) were used to examine the associations between PM10 mortality coefficients and other parameters examined in this study. We used the Fisher transformation to estimate the 95% confidence interval (CI) for r.

Back to Top | Article Outline

RESULTS

Changes in Indoor PM10 Concentration and PM10 Exposure

For the originally targeted 19 NMMAPS cities and the seven U.S. regions containing these cities, Table 1 lists population, seasonal values for Δ[PM10]in/Δ[PM10]out, and seasonal values for the PM10 exposure coefficients. Other input parameters are provided in eTable 2 (http://links.lww.com/EDE/A607). Monthly average values of Δ[PM10]in/Δ[PM10]out and PM10 exposure coefficients were estimated using Equations 4 and 5, respectively. Values for individual months are listed in eTables 3–5 (http://links.lww.com/EDE/A607). For a given season, large city-to-city differences exist in Δ[PM10]in/Δ[PM10]out (eg, in summer: 0.09 in Phoenix compared with 0.39 in New York City). Similarly, for a given season, PM10 exposure coefficients display large city-to-city differences, resulting in substantial regional differences in a given season (eg, in summer: 3.5 in Northeast, 1.7 in Southwest). There are also seasonal differences in a given region (eg, in the Northeast: 3.5 in summer, 2.6 in winter).

TABLE 1

TABLE 1

Back to Top | Article Outline

Correlations Based on 19 Targeted Cities

We have systematically explored potential correlations between PM10 mortality coefficients and PM10 exposure coefficients. In our initial analysis the season-by-season regional values for PM10 exposure coefficients were based on data from the 19 cities for which we had detailed estimates of building infiltration rates.19 Figure 2A shows a scatter plot of PM10 mortality coefficients versus PM10 exposure coefficients for each season in each region. The correlation coefficient for these 28 data pairs is 0.71 (95% CI = 0.46–0.86).

Figure 2B presents comparisons between PM10 mortality coefficients and PM10 exposure coefficients throughout an average year; the scales have been aligned based on maximum and minimum values for each metric. The plots for PM10 mortality coefficients versus time of year (solid line) are adapted from those presented in Figure 3 of the report by Peng et al.7 The plots of PM10 exposure coefficients versus time of year (dotted line) are based on data in eTable 5 (http://links.lww.com/EDE/A607). It is apparent from Figure 2B that: (1) in the Northeast, Industrial Midwest, and Northwest there are seasonal variations in PM10 exposure coefficients that match the seasonal variations in PM10 mortality coefficients; (2) the summer peak in both PM10 exposure coefficients and PM10 mortality coefficients is largest in the Northeast, followed by the Industrial Midwest and the Northwest; (3) in the Southeast and Southwest, the seasonal variations in PM10 exposure coefficients are less pronounced and roughly match the small seasonal variations in PM10 mortality coefficients for these regions; (4) the match between PM10 exposure coefficients and PM10 mortality coefficients is poor in the Upper Midwest and Sothern California—clearly, other factors influence PM10 mortality coefficients.

Back to Top | Article Outline

Correlations Based on 83 Cities

Figure 3A, 3B is analogous to Figure 2A, 2B, but with seasonal PM10 exposure coefficients for the seven regions obtained by population-weight averaging seasonal PM10 exposure coefficients for the 83 NMMAPS cities—the original 19 targeted cities and the 64 additional NMMAPS cities for which there were roughly equivalent cities among the original 19. For all 83 cities, key input parameters (monthly values for Δ[PM10]in/Δ[PM10]out and monthly PM10 exposure coefficients) are listed in eTable 6 (http://links.lww.com/EDE/A607); this table also lists regional values for monthly PM10 exposure coefficients derived from the 83 NMMAPS cities. As was the case in Figure 2A, the association between PM10 mortality coefficients and PM10 exposure coefficients for each season in each region (Fig. 3A) is reasonably described by a linear relationship, with a correlation coefficient of 0.67 (95% CI = 0.40–0.84).In addition, the comparisons among the plots in Figure 3B are similar to those presented in Figure 2B.

Back to Top | Article Outline

Sensitivity Analysis

Table 2 summarizes the results of the sensitivity analyses, listing correlation coefficient and 95% CI for the associations when the default, lower estimate or upper estimate values of x cool, λwindow_open, λrecirc, η, and z are the basis for PM10 exposure coefficients for the original 19 cities. Although lower estimates for λwindow_open and z result in somewhat smaller values for the correlation coefficients (0.61 and 0.62, respectively), the association remains strong in all cases. Interestingly, the correlation improves when a larger estimate for x cool is used. The last two rows in Table 2 indicate that the correlation also remains strong when median rather than mean values of outdoor PM10 and PM2.5 are used to define the size distributions in the various cities.

TABLE 2

TABLE 2

Back to Top | Article Outline

DISCUSSION

In addition to the correlations between PM10 exposure coefficients and PM10 mortality coefficients, we also assessed the correlations between other parameters and PM10 mortality coefficients—those parameters include fraction of residences with central AC or window AC, average indoor temperature, and ratio of PM2.5 to PM10 outdoors (an indicator of the size distribution of PM10). Table 3 summarizes correlation coefficients (r) and 95% CIs for these correlations when the selected parameters are based on data from the original 19 cities or the extended group of 83 cities. The next three sections discuss these findings.

TABLE 3

TABLE 3

Back to Top | Article Outline

Effect Modification by Air Conditioning

Several studies have noted that PM10 mortality coefficients tends to be lower in cities that have a higher fraction of residences with central AC.9–15 Residential central AC systems typically are not designed to provide outdoor air; when such systems are operating, air change occurs primarily as a consequence of infiltration.19,20,28 In addition, such systems contain particle filters. Although these filters have relatively low efficiency, multiple passes of recirculated room air through these filters improve their overall efficiency.29,30 Furthermore, buildings with central AC tend to be newer buildings with less infiltration. Consequently, residents in homes with central AC tend to be less exposed to PM10 of outdoor origin than residents in homes without central AC. PM10 mortality coefficients are negatively correlated with the fraction of residences that have central AC (Table 3: 19 cities, r = −0.63; 83 cities, r = −0.64). This parameter is one of the major factors affecting PM10 exposure coefficients.

In addition to estimating the fraction of residences in a given city with central AC, Smith et al estimated the fraction with window AC.20 PM10 mortality coefficients are positively correlated with the fraction of residences having window AC. However, this association is much stronger when based on data from the original 19 cities rather than the extended collection of 83 cities (Table 3: 19 cities, r = 0.66; 83 cities, r = 0.38). There are several potential reasons for the positive correlation: (1) outdoor-to-indoor transport may be higher in residences with window AC owing to window opening in rooms without window units, coupled with the option on many window AC units of opening a vent to introduce some outdoor air20; (2) cities containing a larger fraction of residences with window AC have a larger fraction of older buildings, which have higher average infiltration rates than newer buildings; and (3) the filters used in window AC units tend to have lower particle-removal efficiencies than those used in central AC systems.

The height above ground level may influence the extent to which AC is an effect modifier. For example, residences on higher floors in a multistory building tend to receive more fresh air than those on lower floors owing to the nature of the wind profile.

Back to Top | Article Outline

Effect Modification by Indoor Temperature

An association has been observed between outdoor temperature and mortality.31–33 Peng et al have adjusted the short-term PM10 mortality coefficients that we are using for outdoor temperatures7 but not indoor temperatures. There is a possibility that the correlation we observe between PM10 exposure coefficients and PM10 mortality coefficients reflects indoor temperature differences driven by regional differences in the fraction of residences with AC, although this is not supported by the positive correlation between window AC and PM mortality coefficients (see above). We further explored this possibility by examining the relationship between seasonal average indoor temperature and PM10 mortality coefficients for the seven U.S. regions. The method for estimating seasonal average indoor temperature is presented in page 29 of the eAppendix (http://links.lww.com/EDE/A607). The correlation between average indoor temperatures and PM10 mortality coefficients (Table 3: 19 cities, r = 0.15; 83 cities, r = 0.23) is much weaker than that between PM10 exposure coefficients and PM10 mortality coefficients. Several parameters used in calculating average indoor temperature are difficult to estimate, but in sensitivity analyses, the uncertainty in those parameters does not alter the weak correlation between average indoor temperatures and PM10 mortality coefficients (eTable 7, http://links.lww.com/EDE/A607). It seems unlikely that the correlation between PM10 exposure coefficients and PM10 mortality coefficients is driven by indoor temperature.

Back to Top | Article Outline

Effect Modification of Outdoor PM2.5/PM10

The ratio of PM2.5 to PM10 outdoors is a crude indicator of the size distribution of outdoor PM10. This ratio differs among the 83 NMMAPS cities in this study. As the size distribution of PM10 shifts toward smaller particles, the penetration factor increases and both the filtration efficiency and the surface removal rate decrease. If all else were equal, this would result in larger PM10 exposure coefficients for cities with larger ratios of outdoor PM2.5/PM10. There is some correlation between outdoor PM2.5/PM10 and corresponding PM10 mortality coefficients (Table 3: 19 cities, r = 0.53; 83 cities, r = 0.47), but the relationship is weaker than that between PM10 exposure and mortality coefficients.

Back to Top | Article Outline

Limitations

There are likely multiple factors responsible for the heterogeneity observed in PM10 mortality coefficients. In addition, the correlations presented in Figures 2 and 3 are based on several approximations and assumptions. We assume that there is no threshold concentration for the impact of particles on short-term mortality. In pooled data for the whole country, Dominici et al3 found a linear concentration-response relationship, although at the regional level, there were modest departures from a linear model in cities from certain locales (Northwest, Southwest, upper Midwest, and Southeast).

We have not adjusted exposure for socioeconomic status. In earlier studies, the effects of PM10 on either mortality or hospitalization were not highly sensitive to socioeconomic status factors such as poverty or race.34,35 However, a recent study suggests that the fraction of outdoor PM2.5 that penetrates and persists indoors is higher with an increased prevalence of poverty.36 We also acknowledge that our analysis does not consider potential confounding owing to regional variation in factors such as age and body mass index.

We used a fixed value for the fraction of time that windows are open in mild weather because city-specific information is unavailable. However, this parameter may vary by city, even for the same temperature distribution, owing to variations in noise, sunshine, and neighborhood security. With that said, we judge that temperature is the dominant factor influencing window opening, and have used the outlined approach as a first approximation.

Increased physical activity and higher breathing rates typically occur outdoors. A larger breathing rate means a larger intake of PM10 per unit time. Breathing rate also affects how deeply particles penetrate the respiratory system. The difference between average breathing rate outdoors and indoors has not been considered in the present analysis.

Some residences with central AC use a forced-air distribution system for heating and cooling; hence, filtration of recirculated air occurs during heating periods as well as cooling periods. Such an approach is more common in milder climates where homes can use heat pumps for both cooling and heating. We have not included filtration during heating periods in Equation 5 because we were unable to obtain information regarding the fraction of residences in a given community with forced-air heating. We do know that this is more common in 2012 than it was in the period 1987–2000.

PM10 may be only a proxy for the actual causative pollutants. Certain constituents of particles may be especially toxic. Bell et al37 have discussed spatial and temporal variation in the chemical composition of PM2.5 in the United States. In addition, the chemical composition of outdoor particles may be modified by the indoor environment. For example, PM10 transported indoors may acquire semivolatile organic compounds present at higher levels indoors than outdoors and lose semivolatile organic compounds present at lower levels indoors than outdoors. Finally, our analysis ignores indoor sources of PM10 such as smoking, cooking, cleaning, and ozone-initiated chemistry. With the exception of the last source, we do not anticipate that indoor sources vary substantially from region to region and season to season. Furthermore, in time-series studies when indoor and outdoor sources are independent, exposure variation owing to indoor-source exposures does not bias the effect estimates.16

Other limitations, including the lack of representative cities for NMMAPS cities located in the middle of the United States and time spent in nonresidential indoor locations, are discussed in our previous report,18 as well as in page 31 of the eAppendix.

In sum, this and our previous study18 indicate that mitigation measures designed to reduce adverse health effects of outdoor air pollutants should consider indoor as well as outdoor exposure. However, these findings should not be interpreted as reason to reduce building ventilation rates, which would likely exacerbate health problems caused by pollutants that originate indoors.38 A better approach is to minimize indoor sources of pollutants while maintaining adequate ventilation with outdoor air (filtered to remove outdoor pollutants when feasible).

Back to Top | Article Outline

ACKNOWLEDGMENTS

We thank R. L. Smith for generously providing data on the fraction of homes in various U.S. cities with central AC and window AC, A. Persily and A. Musser for answering numerous questions regarding their modeling of infiltration rate distributions for U.S. housing, and J. Sundell and J. Baumgartner for valuable comments. C.J.W. thanks Yinping Zhang for sponsoring his Visiting Professorship at Tsinghua University.

Back to Top | Article Outline

REFERENCES

1. Daniels MJ, Dominici F, Samet JM, Zeger SL. Estimating particulate matter-mortality dose-response curves and threshold levels: an analysis of daily time-series for the 20 largest US cities. Am J Epidemiol. 2000;23:397–406
2. Samet JM, Dominici F, Curriero FC, Coursac I, Zeger SL. Fine particulate air pollution and mortality in 20 U.S. cities, 1987-1994. N Engl J Med. 2000;343:1742–1749
3. Dominici F, Daniels M, Zeger SL, Samet JM. Air pollution and mortality: estimating regional and national dose-response relationships. J Am Stat Assoc. 2002;97:100–111
4. Dominici F, McDermott A, Zeger SL, Samet JM. National maps of the effects of particulate matter on mortality: exploring geographical variation. Environ Health Perspect. 2003;111:39–44
5. Dominici F, McDermott A, Daniels M, Zeger SL, Samet JM. Revised analyses of the National Morbidity, Mortality, and Air Pollution Study: mortality among residents of 90 cities. J Toxicol Environ Health Part A. 2005;68:1071–1092
6. Anderson HR, Atkinson RW, Peacock JL, Marston L, Konstantinou K. Meta-analysis of time-series studies and panel studies of Particulate Matter (PM) and Ozone (O3). Report of a WHO task group. 2004 Available at: http://www.euro.who.int/__data/assets/pdf_file/0004/74731/e82792.pdf. Accessed September 20, 2011.
7. Peng RD, Dominici F, Pastor-Barriuso R, Zeger SL, Samet JM. Seasonal analyses of air pollution and mortality in 100 US cities. Am J Epidemiol. 2005;161:585–594
8. Chen C, Zhao B. Review of relationship between indoor and outdoor particles: I/O ratio, infiltration factor and penetration factor. Atmos Environ. 2011;45:275–288
9. Sunyer J, Schwartz J, Tobías A, Macfarlane D, Garcia J, Antó JM. Patients with chronic obstructive pulmonary disease are at increased risk of death associated with urban particle air pollution: a case-crossover analysis. Am J Epidemiol. 2000;151:50–56
10. Janssen NA, Schwartz J, Zanobetti A, Suh HH. Air conditioning and source-specific particles as modifiers of the effect of PM10 on hospital admissions for heart and lung disease. Environ Health Perspect. 2002;110:43–49
11. Medina-Ramón M, Zanobetti A, Schwartz J. The effect of ozone and PM10 on hospital admissions for pneumonia and chronic obstructive pulmonary disease: a national multicity study. Am J Epidemiol. 2006;163:579–588
12. Franklin M, Zeka A, Schwartz J. Association between PM2.5 and all-cause and specific-cause mortality in 27 US communities. J Expo Sci Environ Epidemiol. 2007;17:279–287
13. Bell ML, Ebisu K, Peng RD, Dominici F. Adverse health effects of particulate air pollution: modification by air conditioning. Epidemiology. 2009;20:682–686
14. Meng QY, Spector D, Colome S, Turpin B. Determinants of indoor and personal exposure to PM2.5 of indoor and outdoor origin during the RIOPA Study. Atmos Environ. 2009;43:5750–5758
15. Vedal S. Does air conditioning modify the health effects of exposure to outdoor air pollution? Epidemiology. 2009;20:687–688
16. Sheppard L, Slaughter JC, Schildcrout J, Liu LJ, Lumley T. Exposure and measurement contributions to estimates of acute air pollution effects. J Expo Anal Environ Epidemiol. 2005;15:366–376
17. Sheppard L, Burnett RT, Szpiro AA, et al. Confounding and exposure measurement error in air pollution epidemiology, air quality, atmosphere & health. 2011
18. Chen C, Zhao B, Weschler CJ. Assessing the influence of indoor exposure to “outdoor ozone” on the relationship between ozone and short-term mortality in U.S. communities. Environ Health Perspect. 2012;120:235–240
19. Persily A, Musser A, Emmerich SJ. Modeled infiltration rate distributions for U.S. housing. Indoor Air. 2010;20:473–485
20. Smith RL, Xu B, Switzer P. Reassessing the relationship between ozone and short-term mortality in U.S. urban communities. Inhal Toxicol. 2009;21(suppl 2):37–61
21. Klepeis NE, Nelson WC, Ott WR, et al. The National Human Activity Pattern Survey (NHAPS): a resource for assessing exposure to environmental pollutants. J Expo Anal Environ Epidemiol. 2001;11:231–252
22. Pellizzari ED, Michael LC, Perritt R, Smith DJ, Hartwell TD, Sebestik J. Comparison of indoor and outdoor toxic air pollutant levels in several Southern California communities. Final Report, EPA Contract #68-02-4544. 1989, January.
23. Alevantis LE, Girman JR The Effects of Window-opening on Residential Ventilation. IAQ’89. 1989 Atlanta, GA American Society of Heating, Refrigerating and Air Conditioning Engineers
24. Riley WJ, McKone TE, Lai AC, Nazaroff WW. Indoor particulate matter of outdoor origin: importance of size-dependent removal mechanisms. Environ Sci Technol. 2002;36:200–207
25. Lennox Inc.. 2011 Product Manual for IAQ Systems and Home Comfort Controls. 2011 Available at: www.DaveNet.com. Accessed September 30, 2011.
26. Wallace LA, Emmerich SJ, Howard-Reed C. Continuous measurements of air change rates in an occupied house for 1 year: the effect of temperature, wind, fans, and windows. J Expo Anal Environ Epidemiol. 2002;12:296–306
27. Vandemusser Design LLC. . Ventilation rate frequency distributions for U.S. housing stock. Report PWS NB863060-7–03758 to the National Institute of Standards and Technology. 2007 Gaithersburg, MD
28. Weschler CJ. Ozone’s impact on public health: contributions from indoor exposures to ozone and products of ozone-initiated chemistry. Environ Health Perspect. 2006;114:1489–1496
29. Pui DY, Qi C, Stanley N, Oberdörster G, Maynard A. Recirculating air filtration significantly reduces exposure to airborne nanoparticles. Environ Health Perspect. 2008;116:863–866
30. Fadeyi MO, Weschler CJ, Tham KW. The impact of recirculation, ventilation and filters on secondary organic aerosols generated by indoor chemistry. Atmos Environ. 2009;43:3538–3547
31. Semenza JC, Rubin CH, Falter KH, et al. Heat-related deaths during the July 1995 heat wave in Chicago. N Engl J Med. 1996;335:84–90
32. McGeehin MA, Mirabelli M. The potential impacts of climate variability and change on temperature-related morbidity and mortality in the United States. Environ Health Perspect. 2001;109(suppl 2):185–189
33. Curriero FC, Heiner KS, Samet JM, Zeger SL, Strug L, Patz JA. Temperature and mortality in 11 cities of the eastern United States. Am J Epidemiol. 2002;155:80–87
34. Health Effects Institute (HEI).. Revised analyses of time-series studies of air pollution and health. Special Report. 2003
35. Pope CA 3rd, Ezzati M, Dockery DW. Fine-particulate air pollution and life expectancy in the United States. N Engl J Med. 2009;360:376–386
36. Hodas N, Meng Q, Lunden MM, et al. Variability in the fraction of ambient fine particulate matter found indoors and observed heterogeneity in health effect estimates. J Expo Sci Environ Epidemiol. 2012;22:448–454
37. Bell ML, Dominici F, Ebisu K, Zeger SL, Samet JM. Spatial and temporal variation in PM2.5 chemical composition in the United States for health effects studies. Environ Health Perspect. 2007;115:989–995
38. Sundell J, Levin H, Nazaroff WW, et al. Ventilation rates and health: multidisciplinary review of the scientific literature. Indoor Air. 2011;21:191–204

Supplemental Digital Content

Back to Top | Article Outline
© 2012 Lippincott Williams & Wilkins, Inc.