Age-specific seasonal associations between acute exposure to PM2.5 sources and cardiorespiratory hospital admissions in California

Atmospheric Environment 218 (2019) 117029

Contents lists available at ScienceDirect

Atmospheric Environment journal homepage: http://www.elsevier.com/locate/atmosenv

Age-specific seasonal associations between acute exposure to PM2.5 sources and cardiorespiratory hospital admissions in California Keita Ebisu a, *, Brian Malig a, Sina Hasheminassab b, Constantinos Sioutas b a b

Office of Environmental Health Hazard Assessment, California Environmental Protection Agency, 1515 Clay Street, 16th floor, Oakland, CA, 94612, USA Department of Civil and Environmental Engineering, University of Southern California, 3620 South Vermont Avenue, Los Angeles, CA, 90089, USA

H I G H L I G H T S

G R A P H I C A L A B S T R A C T

� Explored the age-specific associations between PM2.5 sources and hospitalizations. � Vehicular emissions was associated with cardiovascular hospitalizations for elderly. � Vehicular emissions was associated with respiratory hospitalizations for children. � Estimates in the warm season were higher than in the cool season among children.

A R T I C L E I N F O

A B S T R A C T

Keywords: Fine particulate matter PM2.5 sources Hospital admissions Season Children’s health

Numerous studies have explored the relationships between short-term exposure to fine particulate matter (PM2.5) and morbidity. However, few studies have investigated which PM2.5 sources and constituents contribute to the health associations, and even fewer studies are available which explored age or seasonal effect modification for the associations between PM2.5 sources and health. We explored age-specific associations between short-term exposure to PM2.5 chemical constituents and its sources, and hospital admissions in California. We linked hos­ pital admission data (n ¼ 1,679,094) with PM2.5 chemical constituents and source apportionment data for eight sites in California for the period of 2002–2009. Site-specific source apportionment was conducted using Positive Matrix Factorization, and five PM2.5 sources were commonly identified in most sites (biomass burning, soil, secondary ammonium nitrate, secondary ammonium sulfate, and vehicular emissions). Age-stratified Poisson time-series regression was conducted for each site, and the health risk estimates were combined to generate overall age-specific associations with cardiovascular- and respiratory-related hospital admissions. We further conducted seasonal interaction models to assess seasonal effect modification. An interquartile range increase in PM2.5 vehicular emissions was associated with increased risk of cardiovascular-related hospital admission at lag 0 (1.32% [95% confidence interval (CI): 0.16, 2.49]) for elderly people (�65 years old). Exposure to PM2.5 vehicular emissions increased the risk of respiratory-related hospitalizations at lag 2 (3.58% [95% CI: 0.90, 6.33]) for children (0–18 years old). Risk estimates of PM2.5 total mass, vehicular emissions, and its related constituents (e.g., iron) for respiratory admissions were higher in the warm season among children. Heteroge­ neous seasonal estimates were not observed for other age groups. Our results suggest that short-term exposures to several PM2.5 sources and their related constituents are more harmful than exposures to other pollutants,

* Corresponding author. 1515 Clay Street, 16th floor, Oakland, CA, 94612, USA. E-mail address: [email protected] (K. Ebisu). https://doi.org/10.1016/j.atmosenv.2019.117029 Received 13 April 2019; Received in revised form 16 August 2019; Accepted 1 October 2019 Available online 4 October 2019 1352-2310/© 2019 Elsevier Ltd. All rights reserved.

K. Ebisu et al.

Atmospheric Environment 218 (2019) 117029

particularly for children in summer. Identifying toxic sources is important for developing effective interventions and protecting susceptible populations.

1. Introduction

found no seasonal difference (O’Donnell et al., 2011). A study in London found no seasonal difference between PM2.5 black carbon, an indicator of vehicle emissions, and cardiovascular hospital admissions for people aged �65 years old (Samoli et al., 2016b). In contrast, a study in New York City found a significant effect of PM2.5 traffic source only in winter (Lall et al., 2011). Heterogeneous seasonal effects could be due to different contribution of PM2.5 constituents/sources, the pre­ sence/absence of an inversion layer, photochemical activity in the at­ mosphere, differing seasonal indoor penetration rates, or human behavior patterns, all of which further warrant the need for additional research on this subject (Zanobetti and Schwartz, 2009). In this study, we explored the age-specific associations between short-term exposure to PM2.5 chemical constituents and its sources, and hospital admissions in California. We applied two-stage hierarchical time-series model using source apportionment data. To examine effect modifications by season, we conducted an interaction model. We were particularly interested in the health effects of traffic-related pollutants, since they showed inconsistent associations with adverse health out­ comes in the previous studies conducted in California (Berger et al., 2018; Ostro et al., 2016). We hypothesized these inconclusive findings are due to aggregation of season and/or age groups. Isolating toxic PM2.5 constituents/sources is an important research topic as well as revealing disproportional health risks across age group and season from these exposures, which will help achieving effective regulation.

The U.S. Environmental Protection Agency (EPA) concluded that short-term exposure to particulate matter with an aerodynamic diam­ eter � 2.5 μm (PM2.5) has a causal relationship with cardiovascular outcomes and is likely to have a causal relationship with respiratory outcomes (U.S. EPA, 2009). However, health effect estimates of PM2.5 vary among studies (Luben et al., 2017), which may be attributed to differences in the levels of PM2.5 chemical constituents and/or of their sources. PM2.5 consists of many chemical constituents, and certain chemical constituents are tracers of specific PM2.5 sources (Hashemi­ nassab et al., 2014b). Recently, numerous studies have shifted their research focus from PM2.5 total mass to PM2.5 chemical constituents or sources (Samoli et al., 2016b; Sarnat et al., 2008), so that they can unveil toxic constituents to understand pathological mechanisms and identify specific sources to implement emission-specific regulations. Several literature reviews have summarized the associations be­ tween short-term exposure to PM2.5 chemical constituents and adverse health outcomes. Levy et al. showed that the effects of PM2.5 organic carbon (OC) matter, sulfate, and nitrate on mortality are stronger than that of PM2.5 total mass (Levy et al., 2012). Similarly, Lippmann et al. concluded that among the constituents of PM2.5, sulfate, OC, and copper (Cu) are associated with cardiovascular diseases (Lippmann, 2014). In addition to PM2.5 chemical constituents, reviews of health effects of PM2.5 sources are also available. One review concluded that PM2.5 from crustal material, combustion, traffic, or road dust sources are more likely associated with cardiovascular-related mortality (Stanek et al., 2011). The same review also concluded that evidence for the association be­ tween PM2.5 sources and respiratory health effects was limited. Another review found that PM2.5 secondary sulfate is linked with adverse health outcomes, such as mortality (Adams et al., 2015). These studies called for further investigations to demonstrate more consistent relationships between adverse health outcomes, and PM2.5 sources and chemical constituents. Based on these literature reviews, as well as individual studies, several organizations reported that there are not enough studies available to provide collective evidence identifying specific PM2.5 con­ stituents and/or sources as toxic pollutants (U.S. EPA, 2009; Vedal et al., 2013). The World Health Organization encourages the collection of more data on PM2.5 chemical constituents to be used in conjunction with large-scale health studies to develop better understanding of their relationship (World Health Organization, 2007). It is well documented that the burden from air pollution is unequal; certain subpopulations are not only exposed to high pollution levels but also have high health risks from exposures (Bell and Ebisu, 2012; Mar­ tenies et al., 2017). For instance, young and old populations are more susceptible to cardiovascular and respiratory morbidity from PM2.5 exposure compared to adults (Atkinson et al., 2014; Samoli et al., 2016a; Strosnider et al., 2018; Yu and Chien, 2016). Limited findings are available, however, examining whether these demographic factors modify associations between PM2.5 sources/constituents and morbidity. From an environmental justice standpoint, it is an important topic to investigate whether specific PM2.5 sources or constituents dispropor­ tionally affect certain age groups. In addition to age, effect modification by season was reported in several studies, although their results were inconsistent. The effect es­ timate of PM2.5 on mortality was high in the heating season, between November 15 and March 15, in Xi’an, China (Huang et al., 2012), but a national study in the U.S. did not find any seasonal differences (Krall et al., 2013). Canadian studies explored associations between PM2.5 exposure and ischemic stroke hospitalization; one study found strong association in warm season (Villeneuve et al., 2012), but the other study

2. Materials and methods 2.1. Health data Daily unscheduled hospital admission counts from 2002 to 2009 were obtained from the Office of Statewide Health Planning and Development (Office of Statewide Health Planning and Development, 2017), which covers hospital admissions in California excluding federal hospitalizations. The sum of daily hospital admission numbers for each site stratified by health outcomes, age, and race/ethnicity was calcu­ lated. To match with exposure data, 8 sites were selected: Bakersfield, El Cajon, Fresno, Los Angeles, Rubidoux, Sacramento, San Jose, and Simi Valley (Supplemental Fig. 1). We included subjects whose population-weighted centroids of residential ZIP code tabulation areas were within 20 km from the 8 ambient monitors. Outcomes were based on primary diagnosis, and two health out­ comes were considered: cardiovascular-related hospital admissions (International Classification of Disease, Ninth Revision [ICD-9] codes 390–459) and respiratory-related hospital admissions (ICD-9 codes 460–519). Age was divided into three groups (children: � 18 years old, adults: 19–64 years old, and elderly: � 65 years old), and race/ethnicity was divided into four groups (non-Hispanic White, non-Hispanic Black, Hispanic, and non-Hispanic Asian). If age was missing or race/ethnicity did not belong to these four groups, they were removed from the rele­ vant analyses. 2.2. Exposure data PM2.5 total mass and chemical constituents data from 2002 to 2009 were obtained from the U.S. EPA Speciation Trends Network (STN). Eight STN monitors were considered (listed above), and their sampling frequency was every 3rd or 6th day. Among the measured/reported chemical species by STN, we selected 15 PM2.5 chemical constituents based on our previous studies and literature reviews (Bell et al., 2014; Liu et al., 2016; Ostro et al., 2016): aluminum (Al), bromine (Br), cal­ cium (Ca), Cu, elemental carbon (EC), iron (Fe), potassium ion (Kþ), 2

K. Ebisu et al.

Atmospheric Environment 218 (2019) 117029

ammonium ion (NHþ 4 ), nitrate ion (NO3 ), OC, silicon (Si), sulfate ion (SO¼ 4 ), titanium (Ti), vanadium (V), and zinc (Zn). To estimate PM2.5 source contributions, a source apportionment model was conducted at each site separately using the Positive Matrix Factorization (PMF) receptor method on the PM2.5 chemical constituent data. Briefly, the PMF model identifies major sources of PM2.5 and quantifies their relative contributions to PM2.5 total mass by solving chemical mass balance equations using a weighted least-squares algo­ rithm and by imposing non-negativity constrains on the factors (Reff et al., 2007). Five to nine major PM2.5 sources were identified at each site. In our study, we used 5 sources for analyses, which were commonly found in at least 7 study sites: biomass burning, secondary ammonium nitrate, secondary ammonium sulfate, soil, and vehicular emissions mixed with road dust. Secondary ammonium nitrate and secondary ammonium sulfate are secondary pollutants formed through photo­ chemical reactions of constituents in the atmosphere, and their pre­ cursors are primarily emitted by mobile sources (e.g., vehicles, ships, etc.) and stationary sources (e.g., power plants) (Hasheminassab et al., 2014b). Further descriptions regarding monitoring locations, PMF modeling, and site-specific source profiles/contributions used in this study are provided by previous studies (Hasheminassab et al. 2014a, 2014b). Apparent temperature (AT) was calculated and used as an adjust­ ment variable in the model. AT is an index reflecting overall temperature discomfort using temperature and humidity (Michelozzi et al., 2006). Hourly temperature and relative humidity were obtained from the U.S. EPA, the California Irrigation Management Information System, and the National Oceanic Atmospheric Administration, and they were converted to hourly AT. Daily AT was estimated by averaging hourly AT. The closest weather monitors to the air monitors were assigned to each site.

first-stage model. This provided season-specific estimates and covari­ ance matrix for a given site, and these were taken into account in the second stage (Bell et al., 2008; Krall et al., 2013). Seasons were defined as the warm season (May to October) and the cool season (November to April), following previous California studies (Basu et al., 2017; Delfino et al., 2014).

2.3. Statistical model

3. Results

We applied a two-stage hierarchical time-series model to estimate associations between PM2.5-related sources/constituents and hospital admissions (Bell et al., 2008). In the first stage, a site-specific log-linear Poisson model was applied by regressing the daily count of hospital admissions (cardiovascular or respiratory) on the pollutant level. In previous studies, different exposure windows (lag) were associated with different health outcomes (Bell et al., 2008; Ostro et al., 2016), and we considered 3 single-day lags: same day exposure (lag 0), previous day exposure (lag 1), and 2 days’ previous exposure (lag 2). It should be noted that population at different lags are mutually exclusive pop­ ulations since PM2.5 total mass and chemical constituents were not measured every day. The non-daily observation also hinders us from conducting distributed lag models. The model was adjusted by race/­ ethnicity, indicators for day, AT, and long-term temporal trend. The indicator for day was weekday, weekend/holiday, or day after week­ end/holiday, which allowed to adjsut trends of hospital admission numbers by day. Natural spline of AT (degrees of freedom (df) ¼ 3) was used to control for weather, and natural spline of calendar date (df ¼ 4 per year) was used to control for long-term trend. Covariates and df were determined based on previous studies (Basu et al., 2017; Ito et al., 2011; Krall et al., 2013). Once effect estimates of each site were obtained, they were combined to estimate overall association between pollutants and hospital admis­ sions. In the second stage, an overall association was estimated while taking into account within-site statistical error and between-site vari­ ability of the estimates (Dominici et al., 2006). Models were fitted separately for each age group (children, adults, and elderly) because admission numbers and seasonal trands of hospitalization differed by age group. We did not conduct an analysis for cardiovascular-related hospital admissions among the children, because it was a rare event among this group. To explore seasonal differences, we conducted a seasonal model in which we added an interaction term between pollutant and season in the

Our data consist of 1,679,094 total hospital admissions. Median number of cardiovascular-related hospital admissions per day over all sites was 2 (25th and 75th percentile: 1, 3), 132 (113, 146), and 213 (186, 237) for children, adults, and elderly people, respectively. Simi­ larly, median number of respiratory-related hospital admissions per day was 41 (29, 61), 67 (57, 80), and 106 (90, 129). Table 1 shows a sum­ mary of the daily hospital admissions by age group and site. Table 1 also shows site-specific summary statistics of PM2.5 total mass, sources, and AT. With the exception of PM2.5 biomass burning, PM2.5 total mass and its source levels were higher overall in the Los Angeles region (Los Angeles and Rubidoux) and Central Valley (Bakersfield and Fresno) compared to northern California sites (Sacramento and San Jose). Summary statistics of PM2.5 chemical constituents are provided in Supplemental Table 1. PM2.5 vehicular emissions were associated with respiratory-related hospital admissions for the children: an IQR increase in PM2.5 vehic­ ular emissions (2.86 μg/m3) increased the risk of hospitalization by 2.45% (95% confidence interval (CI): 0.02, 4.95) and 3.58% (0.90, 6.33) at lag 0 and lag 2, respectively (Table 2). Strong associations were also observed for PM2.5 vehicular emissions and cardiovascular admissions for the elderly people: the excess risks per IQR were 1.32% (0.16, 2.49) at lag 0 and 1.54% (0.28, 2.81) at lag 1. An IQR increase of exposure to PM2.5 biomass burning (2.71 μg/m3) increased the risk of respiratory hospitalizations 2.59% (0.20, 5.03) at lag 0 and 2.56% (0.01, 5.18) at lag 2 among the elderly group. PM2.5 total mass and other PM2.5 sources (i.e., secondary ammonium nitrate, secondary ammonium sulfate, and soil) did not indicate associations with any type of hospitalizations for either children or the elderly people. Associations between PM2.5 chemical constituents and hospital ad­ missions for the children and elderly people are shown in Fig. 1. Asso­ ciations were observed between cardiovascular admissions and PM2.5 Br (1.65% (0.16, 3.15)), Fe (1.03% (0.11, 1.97)), and OC (1.10% (0.00, 2.22)) for elderly people at lag 1. Elderly people were also sensitive to

2.4. Sensitivity analyses Sensitivity analyses were performed to check robustness of esti­ mates. First, the model was additionally adjusted by the remaining PM2.5 (i.e., the remaining fraction after subtracting the source of interest from the PM2.5 total mass) (Mostofsky et al., 2012; Ng et al., 2017). Site-specific correlations between the PM source of interest and remaining PM2.5 were low to moderate, and the highest correlation was between biomass burning and the remaining PM2.5 in Sacramento (r ¼ 0.61). Second, we conducted the model adjusted for ozone expo­ sure. Ozone was chosen as an adjusted pollutant because it has been associated with morbidity in California (Malig et al., 2016), and its level is spatially homogeneous compared to other gaseous pollutants such as nitrogen dioxide (NO2) or carbon monoxide (CO) (Berman et al., 2015). Ozone data were assigned from the nearest EPA ozone monitors within 20 km of the 8 PM2.5 monitors described above. It should be noted that ozone monitors may not be co-located with PM2.5 monitors. Results were presented as percent change of risk per interquartile range (IQR) increase of the corresponding pollutants. The IQRs were estimated by aggregating all available sites. All statistical analyses were conducted in R version 3.4.1 (R Core Team, 2017). We obtained Insti­ tutional Review Board approval for our protocol from the State of Cal­ ifornia Committee for the Protection of Human Subjects.

3

60.8 (53.8, 70.6)

(10.1, 16.6) (0.10, 2.64) (0.72, 3.58) (0.64, 3.79) (0.22, 0.72) (0.94, 3.57) 60.1 (48.7, 74.9)

12.7 (8.4, 23.7) 2.88 (0.90, 5.95) 3.76 (1.60, 10.65) N.A. N.A. 1.92 (1.10, 3.07)

3 (2, 5) 4 (3, 6) 6 (4, 8)

0 (0, 0) 7 (5, 9) 12 (9, 14)

Fresno

61.7 (56.3, 68.5)

17.2 (12.5, 23.6) 1.01 (0.08, 2.01) 2.96 (1.13, 7.06) 2.73 (0.76, 6.16) 0.76 (0.45, 1.31) 2.87 (1.53, 4.98)

18 (12, 27) 27 (22, 33) 45 (37, 55)

1 (0, 1) 55 (46, 62) 91 (78, 104)

Los Angeles

60.1 (52.8, 70.4)

18.7 (12.2, 29.4) 0.72 (0.04, 1.38) 6.97 (2.14, 14.70) 2.32 (0.81, 4.60) 0.63 (0.32, 1.15) 3.33 (1.81, 5.48)

6 (4, 9) 8 (6, 10) 9 (6, 11)

0 (0, 1) 15 (12, 18) 17 (14, 20)

Rubidoux

4

0.71 ( 0.55, 1.99)

0.32 ( 0.53, 1.18)

0.04 ( 1.77, 1.73)

0.39 ( 0.48, 1.27) 1.54 (0.28, 2.81)

0.19 ( 1.20, 1.59)

0.47 ( 0.35, 1.30)

0.23 ( 1.63, 1.19)

0.69 ( 2.11, 0.75) 1.32 (0.16, 2.49)

Lag 1

0.77 ( 0.16, 1.70)

0.57 ( 0.30, 1.44)

Lag 0

Age �65

Cardiovascular Lag 2

0.36 ( 1.20, 0.49) 0.30 ( 2.03, 1.45)

0.52 ( 1.12, 2.19)

0.06 ( 0.88, 0.77)

0.59 ( 0.44, 1.64)

0.35 ( 0.63, 1.35)

Lag 0

0.05 ( 2.51, 2.46) 2.45 (0.02, 4.95)

0.97 ( 2.29, 4.34)

0.41 ( 1.31, 2.15)

0.54 ( 3.36, 2.36)

0.71 ( 1.20, 2.66)

0.89 ( 1.73, 3.58) 2.29 ( 0.50, 5.16)

0.53 ( 3.09, 4.30)

0.28 ( 1.38, 1.98)

0.19 ( 2.17, 2.61)

0.73 ( 1.04, 2.54)

Lag 1

Age �18 Lag 2

1.10 ( 1.26, 3.52) 3.58 (0.90, 6.33)

1.11 ( 4.77, 2.68)

0.06 ( 2.43, 2.37)

1.31 ( 2.38, 5.14)

0.38 ( 1.87, 2.68)

Lag 0

0.74 ( 2.59, 1.16) 0.64 ( 1.43, 2.76)

0.19 ( 2.08, 1.73)

0.14 ( 1.17, 0.90)

2.59 (0.20, 5.03)

0.89 ( 0.40, 2.19)

Respiratory

57.6 (48.0, 69.5)

9.6 (6.6, 15.5) 2.02 (0.39, 4.69) 0.50 (0.01, 2.39) 1.13 (0.62, 1.99) 0.31 (0.12, 0.64) 2.67 (1.55, 4.49)

3 (2, 5) 8 (6, 10) 12 (9, 16)

0 (0, 0) 13 (11, 16) 22 (19, 26)

Sacramento

0.13 ( 1.14, 1.43) 0.37 ( 1.96, 2.76)

1.52 ( 4.52, 1.57)

0.32 ( 1.70, 1.07)

1.08 ( 1.05, 3.26)

0.26 ( 1.39, 1.94)

Lag 1

Age �65

56.5 (49.6, 63.9)

10.3 (7.2, 15.0) 3.02 (1.77, 5.40) 0.86 (0.25, 2.35) 0.94 (0.43, 1.72) 0.11 (0.05, 0.20) 1.04 (0.48, 1.81)

3 (2, 5) 5 (4, 7) 11 (8, 15)

0 (0, 0) 12 (9, 14) 22 (18, 26)

San Jose

2 (1, 4) 4 (3, 6) 10 (7, 13)

0 (0, 0) 9 (6, 11) 18 (15, 22)

Simi Valley

0.22 ( 2.34, 1.94) 1.27 ( 0.88, 3.47)

0.41 ( 2.26, 1.48)

0.15 ( 1.30, 1.02)

2.56 (0.01, 5.18)

0.84 ( 0.74, 2.44)

Lag 2

60.5 (53.0, 69.1)

11.6 (7.2, 16.2) 0.30 (0.09, 1.13) 1.19 (0.32, 3.44) 1.72 (0.60, 4.33) 0.67 (0.31, 1.20) 2.30 (1.29, 3.54)

Note: IQR values (μg/m3) were 11.6 (PM2.5 total mass), 2.71 (biomass burning), 5.97 (secondary ammonium nitrate), 2.87 (secondary ammonium sulfate), 0.80 (soil), and 2.86 (vehicular emissions).

PM2.5 Total Mass Biomass Burning Secondary Ammonium Nitrate Secondary Ammonium Sulfate Soil Vehicular Emissions

Pollutant

Table 2 Excess percent risk and 95% confidence intervals per IQR increase of PM2.5 total mass and sources for selected age groups and hospitalizations.

Note: Numbers were median (25th and 75th percentile). Secondary ammonium sulfate and soil were not identified in Fresno.

61.0 (50.1, 76.0)

Weather (F� ) Apparent Temperature

13.2 1.22 1.71 1.94 0.44 1.83

3 (2, 5) 7 (5, 9) 10 (8, 13)

2 (1, 3) 4 (2, 5) 4 (3, 6)

14.8 (10.2, 24.7) 0.91 (0.47, 1.83) 2.25 (0.77, 10.93) 2.97 (1.79, 4.35) 1.09 (0.44, 2.12) 2.89 (1.56, 4.47)

0 (0, 0) 12 (9, 15) 20 (17, 24)

El Cajon

0 (0, 0) 6 (5, 8) 7 (5, 9)

Bakersfield

Pollution (μg/m3) PM2.5 Total Mass Biomass Burning Secondary Ammonium Nitrate Secondary Ammonium Sulfate Soil Vehicular Emissions

Hospital Admission Number Cardiovascular Age �18 Age 19-64 Age �65 Respiratory Age �18 Age 19-64 Age �65

Variable

Table 1 Summary statistics of daily hospitalized number, pollution levels, and apparent temperature by site.

K. Ebisu et al.

Atmospheric Environment 218 (2019) 117029

K. Ebisu et al.

Atmospheric Environment 218 (2019) 117029

Fig. 1. Excess percent risk and 95% confidence intervals per IQR increase of PM2.5 chemical constituents for selected age groups and hospitalizations. IQR values (μg/m3) were 0.044 (aluminum), 0.0041 (bromine), 0.051 (calcium), 0.0070 (copper), 0.79 (elemental carbon), 0.096 (iron), 0.066 (potassium ion), 1.88 (ammonium ion), 4.55 (nitrate ion), 3.29 (organic carbon), 0.13 (silicon), 1.71 (sulfate ion), 0.0065 (titanium), 0.0024 (vanadium), and 0.0091 (zinc).

Fig. 2. Excess percent risk and 95% confidence intervals per IQR increase of PM2.5 total mass and sources stratified by seasons. IQR values were shown in footnote of Table 2.

5

K. Ebisu et al.

Atmospheric Environment 218 (2019) 117029

Fig. 3. Excess percent risk and 95% confidence intervals per IQR increase of PM2.5 constituents stratified by seasons for �18 years old. IQR values were shown in footnote of Fig. 1.

Fig. 4. Excess percent risk and 95% confidence intervals per IQR increase of PM2.5 total mass and sources for single pollutant models and models adjust by remaining PM2.5. IQR values were shown in footnote of Table 2.

exposure to PM2.5 OC and V; excess risks of respiratory admissions were 1.95% (0.26, 3.66) and 1.19% (0.02, 2.38) with exposure to PM2.5 OC at lag 2 and V at lag 2, respectively. Associations were observed between PM2.5 Fe and respiratory admissions among children at lag 2. The adult group did not show any associations in either cardiovascular or respi­ ratory hospitalizations with PM2.5 total mass, sources, and chemical constituents (Supplemental Table 2). Heterogeneous seasonal estimates were observed for respiratory hospital admissions among children (Fig. 2). Excess risk estimates were mostly higher in the warm season than in the cool season, and

differences were particularly profound for PM2.5 total mass and vehic­ ular emissions. For instance, an IQR increase of PM2.5 vehicular emis­ sions increased the risk of hospitalization 9.40% (2.81, 16.40) in the warm season and 1.16% ( 1.80, 4.22) in the cool season at lag 0; esti­ mated seasonal differences were 8.14% (1.44, 15.29). The same patterns also appeared at lag 1 and 2. Similarly, associations between PM2.5 total mass and respiratory admissions differed by season at all lags. At lag 0, excess risk was 7.53% (1.42, 14.02) in the warm season, whereas it was 0.18% ( 2.58, 2.27) in the cool season; seasonal difference was 7.73% (1.57, 14.26). Seasonal effect modification was also observed for traffic6

K. Ebisu et al.

Atmospheric Environment 218 (2019) 117029

and soil-related PM2.5 constituents: Br, Ca, EC, Fe, OC, Si, Ti, and V showed higher estimates in the warm season than in the cool season (Fig. 3). We did not see strong seasonal differences among the adult and elderly people for either cardiovascular or respiratory admissions. Estimates were robust after controlling for the remaining PM2.5; traffic emissions and biomass burning retained the associations (Fig. 4). Similarly, estimates were mostly unchanged after being adjusted for ozone, except for estimates of vehicular emissions at lag 1 for children, which had a wider CI (Supplemental Fig. 2).

metals, which included Fe, showed association with asthma emer­ gency department visits for age groups between 5 and 17 years old (Strickland et al., 2010). In addition to these hospitalization studies, a study in South Korea found that metal components of PM2.5, including Fe, was associated with decreasing the children’s peak expiratory flow rate (Hong et al., 2010). Whereas several PM2.5 sources and constituents showed associations with hospital admissions, PM2.5 total mass did not show any association in any age groups. This indicates that specific sources and/or constitu­ ents are more toxic than others, and looking at associations with PM2.5 total mass solely might mask critical associations. Previous studies also found that some sources/constituents are more toxic than PM2.5 total mass (Berger et al., 2018; Ebisu et al., 2018). Reviews on PM2.5 sources and constituents indicated that carbon-containing constituents (i.e., OC) and traffic emissions are more likely to show associations with adverse health outcomes (Rohr and Wyzga, 2012; Stanek et al., 2011). The findings in reviews are in agreement with our findings, but it should be noted that not enough scientific evidence has been accumulated to identify specific sources or constituents that are unequivocally linked to adverse health outcomes. This is partly due to inconsistent labeling of factors among source apportionment studies. For example, one study might name a source factor as “crustal material,” whereas another might call it “soil” or “road dust.” Conducting large-scale health studies com­ bined with consistent PM source apportionment might help identify toxic sources and constituents, which will enable more efficient public health intervention. Seasonal effect modification was observed for children, and its esti­ mates were higher in the warm season than in the cool season, partic­ ularly for PM2.5 total mass, vehicular emission, and its related constituents. It should be noted that the levels of the aforementioned pollutants are typically higher in the cool season than in the warm season in our study areas. Our findings are consistent with a previous study in Atlanta, which found associations between pediatric asthma emergency visits and PM2.5 EC, OC, and water-soluble metals during the warm season (Strickland et al., 2010). Winquist et al. also showed that warm season risk estimates of PM2.5 OC were higher than cold season estimates for an association with asthma-related emergency department visits in St. Louis (Winquist et al., 2015). Our results are also in agree­ ment with the findings of an earlier California study, which did not find any associations between respiratory hospital admissions and same day exposure to PM2.5 total mass and constituents in the cool season (Ostro et al., 2009). We did not observe notable seasonal differences for either cardio­ vascular or respiratory admissions among elderly people. For cardio­ vascular admissions, estimates in the cool season and warm season were similar. Studies in St. Louis and New York City found higher estimates during cold season for the associations between PM2.5 EC and OC, and cardiovascular morbidity (Ito et al., 2011; Winquist et al., 2015), though the CIs are substantially overlapping. For respiratory admissions, our season-specific estimates were mostly similar between seasons, except for biomass burning, which showed slightly higher cool season esti­ mates. Biomass burning in California mainly comes from residential wood burning and wildfire emissions in cool and warm seasons, respectively (Hasheminassab et al., 2014b). Different emission sources may contribute to the various seasonal health estimates. In contrast to our findings, previous studies found that estimates of traffic sources, including EC and OC, were higher in the warm season (Sarnat et al., 2008; Winquist et al., 2015), although the analyses were not separated by age group. By aggregating for all ages, it is possible that the children contributed to higher warm season estimates in those studies. There were some inconsistencies in the observed associations across outcomes and ages. Vehicular emissions were associated with respira­ tory morbidity in children, but not in the elderly. In previous California studies, traffic emission was associated with respiratory emergency room visits (Ostro et al., 2016), but not respiratory mortality (Berger et al., 2018). This may in part due to the aforementioned greater

4. Discussion Vehicular emissions and biomass burning of PM2.5 were associated with hospital admissions in California. It should be noted that associa­ tions were only observed for children and elderly people, confirming that these groups are more susceptible to the associations with PM2.5 total mass, constituents, and sources compared to young and middleaged adults (Bell et al., 2009; Strickland et al., 2010). Associations differed by health outcomes and pollutants. For cardiovascular disease, vehicular emissions were associated with hospitalizations at lag 0 and 1 among elderly people. A study in New York City and Boston also found a short-term association between traffic-related PM2.5 and cardiovascular admissions in the elderly pop­ ulation (Kioumourtzoglou et al., 2014; Lall et al., 2011). A New England study among people �65 years old did not observe a relationship be­ tween motor vehicle and cardiovascular admissions, but associations were observed between cardiovascular admissions and road dust, which is partly attributed to vehicular emissions in our source apportionment analysis (Bell et al., 2014). Vehicular emissions were also linked to cardiovascular emergency room visits in California and Atlanta (Ostro et al., 2016; Sarnat et al., 2008). Further, we found that exposures to Br, Fe, and OC were related to cardiovascular admissions, which were among the marker species of traffic and road dust (Hasheminassab et al., 2014b). Br and OC were reported as toxic pollutants in previous morbidity or mortality studies (Cakmak et al., 2009; Ito et al., 2011; Sarnat et al., 2008). Fe was associated with oxidized low density lipo­ protein, which is a biomarker for oxidative stress and atherosclerosis (Wu et al., 2015). For respiratory-related admissions, biomass burning showed associ­ ations at lag 0 and 2 among elderly people. Biomass burning was linked to respiratory emergency room visits or hospitalizations in California, Atlanta, and Copenhagen (Andersen et al., 2007; Krall et al., 2017; Ostro et al., 2016). Exposure to smoke from wildfire, one of the main con­ tributors to biomass burning and an emerging ambient risk due to climate change, increased the risk of respiratory hospitalizations in the western U.S. (Liu et al., 2017). Among PM2.5 chemical constituents, OC, which in part originates from biomass burning (Hasheminassab et al., 2014b), showed associations with respiratory hospitalizations (Ostro et al., 2016). Association was also observed for PM2.5 V. Major sources of V in California include oil refineries, oil-fired power plants, and ship emissions (Spada et al., 2018), which were not resolved as separate sources in our PMF model, mainly due to lack of additional unique tracers for these sources. PM2.5 OC and V were previously reported to be associated with respiratory hospital admissions in the U.S. and Europe (Basagana et al., 2015; Bell et al. 2009, 2014; Peng et al., 2009). Among children, our study observed associations with PM2.5 vehic­ ular emissions. We also found an association with Fe, which is a key ingredient of fuel additives used for reducing soot emissions (Marsh et al., 2007). Fe is also a major constituent of traffic-induced road dust (Pant and Harrison, 2013). Compared to studies targeting the elderly population, few studies are available exploring the relationships be­ tween PM2.5 sources and constituents, and respiratory-related admis­ sions among children. This is because many studies use Medicare data whose eligibility is 65 or older. Among the limited studies, Fe was associated with respiratory admissions among children under 19 years old in California (Ostro et al., 2009). In Atlanta, PM2.5 water-soluble 7

K. Ebisu et al.

Atmospheric Environment 218 (2019) 117029

vulnerability in children which are represented strongly in emergency room visits and lesser vulnerability among the elderly, which dominated respiratory mortality. Additionally, biomass burning was associated with cardiovascular admissions but not respiratory admissions among elderly people. This could imply differing pathological mechanisms for the two outcomes, with biomass burning more strongly impacting the cardiovascular pathway. There are several plausible explanations for heterogeneous seasonal effects across age groups. First, ambient pollutant levels in the warm season may represent personal exposure level more accurately than in the cool season because of high ventilation rates caused by opening windows (Xu et al., 2014). Furthermore, children play outside more often in the warm season, making their exposure levels more accurate than other age groups, who likely stay indoors. Besides housing and behavioral factors, seasonality of respiratory-related admissions may contribute to seasonal heterogeneous estimates. In general, the number of respiratory-related admissions is high in the cool season, particularly among children because of the high risk of viral infection (Finianos et al., 2016). Given that there is a strong competing factor in the cool season, estimates in the cool season are difficult to ascertain and esti­ mates in the warm season get higher because of low attribution by infection (Sarnat et al., 2008; Strickland et al., 2010). Seasonal estimates of PM2.5 sources or constituents on hospitalizations are not well inves­ tigated, particularly for age-specific estimates. Further studies are war­ ranted to confirm our findings and suggest effective interventions to reduce the risk for susceptible age groups in specific seasons. Results of several biomarker studies help to understand pathological mechanisms. A study in Finland showed that exposure to traffic emis­ sions was associated with cytokines interleukin-12, a risk factor for cardiovascular disease (Siponen et al., 2015). Exposure to PM2.5 OC can change the von Willebrand Factor, a biomarker of acute vascular response linked to the risk of cardiovascular disease (Strak et al., 2013). Furthermore, exposure to V was associated with altering DNA methyl­ ation of proinflammatory asthma genes and lowering the fractional concentration of expired nitric oxide, both indicating adverse respira­ tory responses (Jung et al., 2017; Kim et al., 2003). This study has several limitations. We conducted PMF for each site to capture local emission sources, but we did not use any other source apportionment methods to confirm the consistency of the results. Several apportionment techniques are available. For example, chemical transport models, such as the Community Multiscale Air Quality Modeling System (CMAQ), utilize emissions inventories and dispersion function to estimate PM2.5 source profiles unlike the PMF model. While there are some pros and cons for each method, none of the source apportionment methods are considered the gold standard (Thurston et al., 2005). Second, results of lags should be interpreted with caution. Our results sometimes showed mixed results across lags (e.g. associa­ tions were only observed among lag 0 and 2). While there might be a pathological explanation for lag associations, this finding could be due to heterogeneous population. Because the PM2.5 sampling frequency was either every 3 or 6 days, subjects included in the analyses differ by lags. Therefore, mixed results could be due to random population differences. Furthermore, we conducted co-pollutant models for the remaining PM2.5 and ozone, but we did not adjust for other gaseous pollutants such as NO2. Unlike ozone, spatial heterogeneity is high and only a few monitors were available for NO2, resulting in serious exposure misclassification. Additionally, given that NO2 is a strong indicator of traffic, collinearity with vehicular emissions might introduce statistical challenges. More­ over, monitor locations for NO2 do not necessarily co-locate with PM2.5 constituents’ monitors, leading to a potential exposure misclassification; one person might reside close to the NO2 monitor, whereas another person might live far from the monitor even though they both live within 20 km circle from the PM2.5 monitor. Lacking individual level infor­ mation, such as time activity patterns, is another potential source of exposure misclassification. People commuting or working in heavy traffic environment can be exposed to higher vehicular emissions than

others. Lastly, only 8 PM2.5 chemical constituents’ monitoring sites were available in California during our study period. Although PM2.5 total mass can be estimated via satellite or CMAQ, limited techniques are available to estimate constituents or sources using these methods. Reliance on ground monitors is still high, and increasing the monitor number measuring PM2.5 constituents is critical to obtain further insights. 5. Conclusions We explored associations between short-term exposure to PM2.5 sources and constituents and hospital admissions. Our findings corrob­ orate the toxicity of certain PM2.5 sources, such as traffic. We stratified by age group to obtain age-specific estimates. Aggregating all age groups likely masks or underestimates the risks to susceptible populations. Another novelty of this study is investigating seasonal estimates. Few studies are available on the seasonal estimates of PM2.5 constituents for the short-term relationships with hospitalizations, and even fewer studies have examined seasonal estimates of PM2.5 sources. Our findings might help develop effective interventions for reducing targeted sources and protecting susceptible populations by drawing attention to chil­ dren’s PM2.5 exposures, especially from traffic emissions, in the warm season. Further research is critical to shed light on environmental justice issues and to clarify the toxicity of PM2.5 sources and constituents. Disclaimer The contents in this article are solely those of the authors, and do not necessarily represent those of the California Environmental Protection Agency, the Office of Environmental Health Hazard Assessment, or the State of California. Declaration of competing interest The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper. Acknowledgement The source-apportionment analysis used in this paper was funded by the California Environmental Protection Agency, Office of Environ­ mental Health Hazard Assessment (12-E0021). Appendix A. Supplementary data Supplementary data to this article can be found online at https://doi. org/10.1016/j.atmosenv.2019.117029. References Adams, K., Greenbaum, D.S., Shaikh, R., van Erp, A.M., Russell, A.G., 2015. Particulate matter components, sources, and health: systematic approaches to testing effects. J. Air Waste Manag. Assoc. 65, 544–558. Andersen, Z.J., Wahlin, P., Raaschou-Nielsen, O., Scheike, T., Loft, S., 2007. Ambient particle source apportionment and daily hospital admissions among children and elderly in copenhagen. J. Expo. Sci. Environ. Epidemiol. 17, 625–636. Atkinson, R.W., Kang, S., Anderson, H.R., Mills, I.C., Walton, H.A., 2014. Epidemiological time series studies of pm2.5 and daily mortality and hospital admissions: a systematic review and meta-analysis. Thorax 69, 660–665. Basagana, X., Jacquemin, B., Karanasiou, A., Ostro, B., Querol, X., Agis, D., et al., 2015. Short-term effects of particulate matter constituents on daily hospitalizations and mortality in five south-european cities: results from the med-particles project. Environ. Int. 75, 151–158. Basu, R., Gavin, L., Pearson, D., Ebisu, K., Malig, B., 2017. Examining the association between temperature and emergency room visits from mental health-related outcomes in California. Am. J. Epidemiol. 187 (4), 726–735. Bell, M.L., Ebisu, K., Peng, R.D., Walker, J., Samet, J.M., Zeger, S.L., et al., 2008. Seasonal and regional short-term effects of fine particles on hospital admissions in 202 us counties, 1999-2005. Am. J. Epidemiol. 168, 1301–1310.

8

K. Ebisu et al.

Atmospheric Environment 218 (2019) 117029

Bell, M.L., Ebisu, K., Peng, R.D., Samet, J.M., Dominici, F., 2009. Hospital admissions and chemical composition of fine particle air pollution. Am. J. Respir. Crit. Care Med. 179, 1115–1120. Bell, M.L., Ebisu, K., 2012. Environmental inequality in exposures to airborne particulate matter components in the United States. Environ. Health Perspect. 120, 1699–1704. Bell, M.L., Ebisu, K., Leaderer, B.P., Gent, J.F., Lee, H.J., Koutrakis, P., et al., 2014. Associations of pm(2).(5) constituents and sources with hospital admissions: analysis of four counties in Connecticut and Massachusetts (USA) for persons >/¼ 65 years of age. Environ. Health Perspect. 122, 138–144. Berger, K., Malig, B.J., Hasheminassab, S., Pearson, D.L., Sioutas, C., Ostro, B., et al., 2018. Associations of source-apportioned fine particles with cause-specific mortality in California. Epidemiology 29, 639–648. Berman, J.D., Breysse, P.N., White, R.H., Waugh, D.W., Curriero, F.C.J.E.T., Innovation, 2015. Evaluating Methods for Spatial Mapping: Applications for Estimating Ozone Concentrations across the Contiguous united states, vol. 3, pp. 1–10. Cakmak, S., Dales, R.E., Gultekin, T., Vidal, C.B., Farnendaz, M., Rubio, M.A., et al., 2009. Components of particulate air pollution and emergency department visits in Chile. Arch. Environ. Occup. Health 64, 148–155. Delfino, R.J., Wu, J., Tjoa, T., Gullesserian, S.K., Nickerson, B., Gillen, D.L., 2014. Asthma morbidity and ambient air pollution: effect modification by residential traffic-related air pollution. Epidemiology 25, 48–57. Dominici, F., Peng, R.D., Bell, M.L., Pham, L., McDermott, A., Zeger, S.L., et al., 2006. Fine particulate air pollution and hospital admission for cardiovascular and respiratory diseases. J. Am. Med. Assoc. 295, 1127–1134. Ebisu, K., Malig, B., Hasheminassab, S., Sioutas, C., Basu, R., 2018. Cause-specific stillbirth and exposure to chemical constituents and sources of fine particulate matter. Environ. Res. 160, 358–364. Finianos, M., Issa, R., Curran, M.D., Afif, C., Rajab, M., Irani, J., et al., 2016. Etiology, seasonality, and clinical characterization of viral respiratory infections among hospitalized children in beirut, Lebanon. J. Med. Virol. 88, 1874–1881. Hasheminassab, S., Daher, N., Ostro, B.D., Sioutas, C., 2014. Long-term source apportionment of ambient fine particulate matter (pm2.5) in the los angeles basin: a focus on emissions reduction from vehicular sources. Environ. Pollut. 193, 54–64. Hasheminassab, S., Daher, N., Saffari, A., Wang, D., Ostro, B., Sioutas, C., 2014. Spatial and temporal variability of sources of ambient fine particulate matter (pm 2.5) in California. Atmos. Chem. Phys. 14, 12085–12097. Hong, Y.C., Pan, X.C., Kim, S.Y., Park, K., Park, E.J., Jin, X., et al., 2010. Asian dust storm and pulmonary function of school children in seoul. Sci. Total Environ. 408, 754–759. Huang, W., Cao, J., Tao, Y., Dai, L., Lu, S.E., Hou, B., et al., 2012. Seasonal variation of chemical species associated with short-term mortality effects of pm(2.5) in xi’an, a central city in China. Am. J. Epidemiol. 175, 556–566. Ito, K., Mathes, R., Ross, Z., Nadas, A., Thurston, G., Matte, T., 2011. Fine particulate matter constituents associated with cardiovascular hospitalizations and mortality in New York city. Environ. Health Perspect. 119, 467–473. Jung, K.H., Torrone, D., Lovinsky-Desir, S., Perzanowski, M., Bautista, J., Jezioro, J.R., et al., 2017. Short-term exposure to pm2.5 and vanadium and changes in asthma gene DNA methylation and lung function decrements among urban children. Respir. Res. 18, 63. Kim, J.Y., Hauser, R., Wand, M.P., Herrick, R.F., Amarasiriwardena, C.J., Christiani, D. C., 2003. The association of expired nitric oxide with occupational particulate metal exposure. Environ. Res. 93, 158–166. Kioumourtzoglou, M.A., Coull, B.A., Dominici, F., Koutrakis, P., Schwartz, J., Suh, H., 2014. The impact of source contribution uncertainty on the effects of source-specific pm2.5 on hospital admissions: a case study in boston, ma. J. Expo. Sci. Environ. Epidemiol. 24, 365–371. Krall, J.R., Anderson, G.B., Dominici, F., Bell, M.L., Peng, R.D., 2013. Short-term exposure to particulate matter constituents and mortality in a national study of u.S. Urban communities. Environ. Health Perspect. 121, 1148–1153. Krall, J.R., Mulholland, J.A., Russell, A.G., Balachandran, S., Winquist, A., Tolbert, P.E., et al., 2017. Associations between source-specific fine particulate matter and emergency department visits for respiratory disease in four u.S. Cities. Environ. Health Perspect. 125, 97–103. Lall, R., Ito, K., Thurston, G.D., 2011. Distributed lag analyses of daily hospital admissions and source-apportioned fine particle air pollution. Environ. Health Perspect. 119, 455–460. Levy, J.I., Diez, D., Dou, Y., Barr, C.D., Dominici, F., 2012. A meta-analysis and multisite time-series analysis of the differential toxicity of major fine particulate matter constituents. Am. J. Epidemiol. 175, 1091–1099. Lippmann, M., 2014. Toxicological and epidemiological studies of cardiovascular effects of ambient air fine particulate matter (pm2.5) and its chemical components: coherence and public health implications. Crit. Rev. Toxicol. 44, 299–347. Liu, J.C., Wilson, A., Mickley, L.J., Dominici, F., Ebisu, K., Wang, Y., et al., 2017. Wildfire-specific fine particulate matter and risk of hospital admissions in urban and rural counties. Epidemiology 28, 77–85. Liu, S., Ganduglia, C.M., Li, X., Delclos, G.L., Franzini, L., Zhang, K., 2016. Fine particulate matter components and emergency department visits among a privately insured population in greater houston. Sci. Total Environ. 566–567, 521–527. Luben, T.J., Nichols, J.L., Dutton, S.J., Kirrane, E., Owens, E.O., Datko-Williams, L., et al., 2017. A systematic review of cardiovascular emergency department visits, hospital admissions and mortality associated with ambient black carbon. Environ. Int. 107, 154–162. Malig, B.J., Pearson, D.L., Chang, Y.B., Broadwin, R., Basu, R., Green, R.S., et al., 2016. A time-stratified case-crossover study of ambient ozone exposure and emergency department visits for specific respiratory diagnoses in California (2005-2008). Environ. Health Perspect. 124, 745–753.

Marsh, N.D., Preciado, I., Eddings, E.G., Sarofim, A.F., Palotas, A.B., Robertson, D.J., 2007. Evaluation of organometallic fuel additives for soot suppression. Combust. Sci. Technol. 179, 987–1001. Martenies, S.E., Milando, C.W., Williams, G.O., Batterman, S.A., 2017. Disease and health inequalities attributable to air pollutant exposure in detroit, Michigan. Int. J. Environ. Res. Public Health 14. Michelozzi, P., De Sario, M., Accetta, G., de’Donato, F., Kirchmayer, U., D’Ovidio, M., et al., 2006. Temperature and summer mortality: geographical and temporal variations in four Italian cities. J. Epidemiol. Community Health 60, 417–423. Mostofsky, E., Schwartz, J., Coull, B.A., Koutrakis, P., Wellenius, G.A., Suh, H.H., et al., 2012. Modeling the association between particle constituents of air pollution and health outcomes. Am. J. Epidemiol. 176, 317–326. Ng, C., Malig, B., Hasheminassab, S., Sioutas, C., Basu, R., Ebisu, K., 2017. Source apportionment of fine particulate matter and risk of term low birth weight in California: exploring modification by region and maternal characteristics. Sci. Total Environ. 605–606, 647–654. O’Donnell, M.J., Fang, J., Mittleman, M.A., Kapral, M.K., Wellenius, G.A., Investigators of the Registry of Canadian Stroke N, 2011. Fine particulate air pollution (pm2.5) and the risk of acute ischemic stroke. Epidemiology 22, 422–431. Office of Statewide Health Planning and Development, 2017. California patient discharged data file documentation. Available: https://www.oshpd.ca.gov/. Ostro, B., Roth, L., Malig, B., Marty, M., 2009. The effects of fine particle components on respiratory hospital admissions in children. Environ. Health Perspect. 117, 475–480. Ostro, B., Malig, B., Hasheminassab, S., Berger, K., Chang, E., Sioutas, C., 2016. Associations of source-specific fine particulate matter with emergency department visits in California. Am. J. Epidemiol. 184, 450–459. Pant, P., Harrison, R.M., 2013. Estimation of the contribution of road traffic emissions to particulate matter concentrations from field measurements: a review. Atmos. Environ. 77, 78–97. Peng, R.D., Bell, M.L., Geyh, A.S., McDermott, A., Zeger, S.L., Samet, J.M., et al., 2009. Emergency admissions for cardiovascular and respiratory diseases and the chemical composition of fine particle air pollution. Environ. Health Perspect. 117, 957–963. R Core Team, 2017. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria. Reff, A., Eberly, S.I., Bhave, P.V., 2007. Receptor modeling of ambient particulate matter data using positive matrix factorization: review of existing methods. J. Air Waste Manag. Assoc. 57, 146–154. Rohr, A.C., Wyzga, R.E., 2012. Attributing health effects to individual particulate matter constituents. Atmos. Environ. 62, 130–152. Samoli, E., Atkinson, R.W., Analitis, A., Fuller, G.W., Beddows, D., Green, D.C., et al., 2016. Differential health effects of short-term exposure to source-specific particles in london, u.K. Environ. Int. 97, 246–253. Samoli, E., Atkinson, R.W., Analitis, A., Fuller, G.W., Green, D.C., Mudway, I., et al., 2016. Associations of short-term exposure to traffic-related air pollution with cardiovascular and respiratory hospital admissions in london, UK. Occup. Environ. Med. 73, 300–307. Sarnat, J.A., Marmur, A., Klein, M., Kim, E., Russell, A.G., Sarnat, S.E., et al., 2008. Fine particle sources and cardiorespiratory morbidity: an application of chemical mass balance and factor analytical source-apportionment methods. Environ. Health Perspect. 116, 459–466. Siponen, T., Yli-Tuomi, T., Aurela, M., Dufva, H., Hillamo, R., Hirvonen, M.R., et al., 2015. Source-specific fine particulate air pollution and systemic inflammation in ischaemic heart disease patients. Occup. Environ. Med. 72, 277–283. Spada, N.J., Cheng, X., White, W.H., Hyslop, N.P., 2018. Decreasing vanadium footprint of bunker fuel emissions. Environ. Sci. Technol. 52 (20), 11528–11534. Stanek, L.W., Sacks, J.D., Dutton, S.J., Dubois, J.-J.B., 2011. Attributing health effects to apportioned components and sources of particulate matter: an evaluation of collective results. Atmos. Environ. 45, 5655–5663. Strak, M., Hoek, G., Godri, K.J., Gosens, I., Mudway, I.S., van Oerle, R., et al., 2013. Composition of pm affects acute vascular inflammatory and coagulative markers the raptes project. PLoS One 8, e58944. Strickland, M.J., Darrow, L.A., Klein, M., Flanders, W.D., Sarnat, J.A., Waller, L.A., et al., 2010. Short-term associations between ambient air pollutants and pediatric asthma emergency department visits. Am. J. Respir. Crit. Care Med. 182, 307–316. Strosnider, H.M., Chang, H.H., Darrow, L.A., Liu, Y., Vaidyanathan, A., Strickland, M.J., 2018. Age-specific associations of ozone and pm2.5 with respiratory emergency department visits in the us. Am. J. Respir. Crit. Care Med. 199 (7), 882–890. Thurston, G.D., Ito, K., Mar, T., Christensen, W.F., Eatough, D.J., Henry, R.C., et al., 2005. Workgroup report: workshop on source apportionment of particulate matter health effects–intercomparison of results and implications. Environ. Health Perspect. 113, 1768–1774. U.S. EPA, 2009. Integrated Science Assessment for Particulate Matter. EPA/600/R-08/ 139F. Vedal, S., Campen, M.J., McDonald, J.D., Larson, T.V., Sampson, P.D., Sheppard, L., et al., 2013. National Particle Component Toxicity (Npact) Initiative Report on Cardiovascular Effects. Research Report. Health Effects Institute, pp. 5–8. Villeneuve, P.J., Johnson, J.Y., Pasichnyk, D., Lowes, J., Kirkland, S., Rowe, B.H., 2012. Short-term effects of ambient air pollution on stroke: who is most vulnerable? Sci. Total Environ. 430, 193–201. Winquist, A., Schauer, J.J., Turner, J.R., Klein, M., Sarnat, S.E., 2015. Impact of ambient fine particulate matter carbon measurement methods on observed associations with acute cardiorespiratory morbidity. J. Expo. Sci. Environ. Epidemiol. 25, 215–221. World Health Organization, 2007. Health Relevance of Particulate Matter from Various Sources: Report on a Who Workshop, Bonn, germany 26-27 March 2007. Wu, S., Yang, D., Wei, H., Wang, B., Huang, J., Li, H., et al., 2015. Association of chemical constituents and pollution sources of ambient fine particulate air pollution

9

K. Ebisu et al.

Atmospheric Environment 218 (2019) 117029

and biomarkers of oxidative stress associated with atherosclerosis: a panel study among young adults in beijing, China. Chemosphere 135, 347–353. Xu, J., Bai, Z., You, Y., Zhou, J., Zhang, J., Niu, C., et al., 2014. Residential indoor and personal pm10 exposures of ambient origin based on chemical components. J. Expo. Sci. Environ. Epidemiol. 24, 428–436.

Yu, H.L., Chien, L.C., 2016. Short-term population-based non-linear concentrationresponse associations between fine particulate matter and respiratory diseases in taipei (taiwan): a spatiotemporal analysis. J. Expo. Sci. Environ. Epidemiol. 26, 197–206. Zanobetti, A., Schwartz, J., 2009. The effect of fine and coarse particulate air pollution on mortality: a national analysis. Environ. Health Perspect. 117, 898–903.

10