Contribution of the in-vehicle microenvironment to individual ambient-source nitrogen dioxide exposure: the Multi-Ethnic Study of Atherosclerosis and Air Pollution.

Exposure estimates that do not account for time in-transit may underestimate exposure to traffic-related air pollution, but exact contributions have not been studied directly. We conducted a 2-week monitoring, including novel in-vehicle sampling, in a subset of the Multi-Ethnic Study of Atherosclerosis and Air Pollution cohort in two cities. Participants spent the majority of their time indoors and only 4.4% of their time (63 min/day) in-vehicle, on average. The mean ambient-source NO2 concentration was 5.1 ppb indoors and 32.3 ppb in-vehicle during drives. On average, indoor exposure contributed 69% and in-vehicle exposure contributed 24% of participants' ambient-source NO2 exposure. For participants in the highest quartile of time in-vehicle (≥1.3 h/day), indoor and in-vehicle contributions were 60 and 31%, respectively. Incorporating infiltrated indoor and measured in-vehicle NO2 produced exposure estimates 5.6 ppb lower, on average, than using only outdoor concentrations. The indoor microenvironment accounted for the largest proportion of ambient-source exposure in this older population, despite higher concentrations of NO2 outdoors and in vehicles than indoors. In-vehicle exposure was more influential among participants who drove the most and for participants residing in areas with lower outdoor air pollution. Failure to characterize exposures in these microenvironments may contribute to exposure misclassification in epidemiologic studies.

INTRODUCTION

Accurate exposure assessment remains a central challenge in large epidemiologic studies investigating the health effects of ambient air pollution exposure. Long-term monitoring at the individual level is not feasible on such a scale and thus studies rely on surrogate measures to estimate individual exposure to air pollution. Typical measures include city-wide averages based on centrally located monitors, distance from residences to the nearest major roadways, or modeled outdoor concentrations at residences.[1-8] However, environmental exposures such as air pollution can vary greatly among different microenvironments.[9-12] People spend the majority of their time indoors and thus the indoor microenvironment is particularly important.[13-16] To better account for this microenvironment, the Multi-Ethnic Study of Atherosclerosis and Air Pollution (MESA Air) improved estimates of individual exposure by developing and incorporating modeled particulate infiltration into residences.[17-20] Yet even these improved estimates including indoor exposures do not account for another potentially important microenvironment: inside motor vehicles.

Concentrations of traffic-related air pollution (TRAP) can be much higher on the roadway and inside motor vehicles than at residential locations.[21-23] Despite the limited time spent in-vehicle relative to indoors, exposures in this microenvironment may comprise a large proportion of individual exposure to air pollution,[24,25] and if so, exposure estimates not accounting for time spent in-transit may underestimate exposure to TRAP. Though TRAP includes mixtures of many pollutants, nitrogen dioxide (NO2) is often used as a surrogate marker of TRAP. NO2 is also of interest because it is designated by the US EPA as a criteria air pollutant and is associated with a number of health effects including cardiovascular disease.[26,27]

Given the challenge of accurate exposure assessment and the potential impact of in-vehicle exposures, the primary aim of this study was to quantify the contribution of time spent in the in-vehicle microenvironment to individual exposure to ambient source NO2. We hypothesized that, although most participants spend only a small portion of their time in-transit, the in-vehicle microenvironment may still substantially contribute to overall exposure to NO2, and that the magnitude of this contribution may vary by factors such as city and season. We conducted monitoring campaigns with air pollution sampling and intensive time-location assessment to test this hypothesis.

METHODS

A subset of the MESA Air cohort participated in two-week monitoring campaigns. These campaigns were conducted during summer and winter in Winston-Salem, NC in 2013 (n=64) and in Los Angeles, CA in 2014 (n=64). During each monitoring campaign, samplers were deployed over the course of six consecutive days and then collected after two weeks of monitoring.

Study population

Selection and recruitment of the MESA Air cohort has been described in detail elsewhere.[19,28] Briefly, MESA Air is an ancillary study to the Multi-Ethnic Study of Atherosclerosis (MESA), which enrolled participants aged 45–84 who were free of recognized cardiovascular disease at baseline (2000–2002). MESA targeted four racial/ethnic groups for inclusion: white, black, Hispanic, and Chinese. The 128 participants included in this in-vehicle study met the following criteria: previously consented to be contacted about participating in monitoring studies, English-speaking, reported owning and driving a car as their primary mode of transportation, reported driving on average at least 30 minutes per day, non-smoking, not living with any smokers, and not planning to spend multiple nights away from home during the sampling period. All study participants gave informed written consent before data collection and all of the participating centers’ institutional review boards approved the study.

Outdoor concentrations of NO2

Residential indoor and outdoor, in-vehicle, and personal air monitoring was conducted to measure concentrations of NO2 in each microenvironment. Two-week integrated samples were collected using Ogawa passive samplers (Ogawa & Co., Pompano Beach, FL) and analyzed using ion chromatography.

Ambient-source indoor concentrations of NO2

As we were interested solely in the ambient-source portion of indoor air pollution, we used a literature-derived estimate of infiltration to estimate the proportion of the outdoor NO2 concentration that infiltrates inside the residence. This indoor to outdoor (I/O) ratio includes a city- and season-specific air exchange rate (AER), a building fabric filtration factor (f)[9] of 1.00 which reflects the ability of gases to penetrate the building envelope due to natural ventilation, and a decay rate of NO2 (K)[10] of 0.99 h−1, and is calculated using: Based on the literature, AERs of 0.81 and 0.36 were used in Winston-Salem for winter and summer, respectively.[29] AERs of 0.61 and 1.13 were used in Los Angeles for winter and summer, respectively.[30] The literature-derived ratios were compared to the observed indoor to outdoor ratio in homes with no reported indoor sources of NO2.

Indoor sources of NO2 such as gas appliances and relevant housing characteristics such as the presence of an attached garage were collected via questionnaire at the start of each monitoring period. Measured residential indoor concentrations of NO2 in homes with no reported indoor sources were only used to calculate an observed indoor-outdoor ratio (n=8 and n=5 in winter and summer in Winston-Salem, respectively, and n=2 in winter in Los Angeles).

In-vehicle concentrations of NO2

A novel in-vehicle sampler was also placed in each participant’s primary vehicle. Participants were instructed to open the in-vehicle sampler before driving and close the sampler at the conclusion of their drive, so the sampler was only open while the participant was in the vehicle. Additionally, participants were asked to take the sampler with them if they traveled in a different vehicle.

These novel samplers sealed tightly when closed, and were designed to minimize empty space while still providing room for the passive air samplers to be exposed when the housing of the sampler was open. Given these design considerations, 350 mL of dead space remained inside the canister upon closure. The Ogawa sampler inside of the canister continued to sample the NO2 from the trapped air in this dead space until all NO2 was consumed. This was confirmed using laboratory tests with known levels of NO2, which indicated that the Ogawa badge was absorbing NO2 from the air inside the canister after the canister was closed (Table S3). To account for this additional sampling occurring upon each closure of the canister, we calculated an effective sampling rate assuming a high affinity of the sorbent for NO2 (see equation in supplement). This simplified sampling rate assumed concentration independence and was calculated using the molecular weight of NO2 and molar volume of air at standard temperature and pressure and assuming a relative humidity of 43%. Given the 350 mL volume of air trapped in the canister, we estimated that it would take 40.2 minutes for all of the NO2 to be absorbed. When calculating NO2 vehicle concentrations, the aggregate sampling time included this additional sampling time for each opening and closure of the canister.

Two-week time-location measurements

During each of the sampling periods, participants were asked to complete two methods of intensive location tracking: carrying a GPS unit and filling out Time-Location Diaries (TLDs). Participants were instructed to carry a customized 747ProS GPS Data Logger unit (G&V Global Tech Co., LTD, Taipei City, Taiwan) whenever they left the house. Participants were also instructed to fill out a TLD providing specific location information categorizing locations as home indoors, home outdoors, other indoors, other outdoors, and in-vehicle, throughout each day of the monitoring period. Time-location patterns measured by these two methods were integrated to capitalize on their respective strengths.[31] GPS data were used to classify time as in-vehicle, at home, or in other locations, and time at home and in other locations was further divided into indoor and outdoor using proportions reported in the TLDs. Ultimately, the analysis of GPS and TLD data yielded a single estimate of time spent in each of three types of microenvironments: indoors, outdoors, and in-vehicle.

Additionally, participants provided information on the TLDs about transit-related activities such as traffic conditions, road type, and open vehicle windows. Specific information about the make, model, and year of the primary and secondary vehicles was collected via questionnaire at the start of each monitoring period.

Data Analysis

Overall individual exposure to ambient-source NO2 was calculated by summing exposures from three microenvironments—indoors, outdoors, and in-vehicle—using a simple time-weighted model.

Total Individual Exposure = Exposureoutdoor + Exposureindoor + Exposurevehicle The proportional contributions of each of the three microenvironments to overall individual exposure to ambient-source NO2 were compared with a focus on the relative importance of the in-vehicle environment.

RESULTS

Characteristics of the subset of MESA Air participants who participated in each two-week monitoring campaign, as well as characteristics of the entire MESA Air cohort, are shown in Table 1. Each subset included a similar proportion of female participants as the entire cohort, except for the Los Angeles summer campaign, which included a higher proportion of male participants. Chinese-Americans were under-represented in the monitoring subset, in part because Chinese-Americans were not recruited in Winston-Salem in the overall cohort. The median age was lower in the Los Angeles subset than in the Winston-Salem subset.

Table 1

Characteristics of the MESA Air cohort and the subset of participants who completed two-week monitoring.

Characteristic	Winston-Salem, NC1		Los Angeles, CA2		Entire MESAAir Cohort3,4(n=4 920)
	Winter(n=46)	Summer(n=47)	Winter(n=47)	Summer(n=46)
Sex, n (%)
Female	25 (54)	23 (51)	24 (51)	19 (41)	2 625 (53)
Male	21 (46)	24 (49)	23 (49)	27 (59)	2 295 (47)
Race/Ethnicity, n (%)
White	20 (43)	21 (45)	14 (30)	15 (33)	2 055 (42)
Black	26 (57)	26 (55)	9 (19)	5 (15)	1 285 (26)
Chinese	0 (0)	0 (0)	3 (6)	7 (11)	541 (11)
Hispanic	0 (0)	0 (0)	21 (45)	19 (41)	1 039 (21)
Age,4 n (%)
45–54	1 (2)	1 (2)	2 (4)	2 (4)	77 (2)
55–64	9 (20)	14 (30)	21 (45)	16 (34)	1 616 (33)
65–74	18 (39)	19 (40)	16 (34)	21 (45)	1 569 (32)
75–84	15 (33)	12 (26)	8 (17)	6 (13)	1 300 (26)
85+	3 (7)	1 (2)	0 (0)	1 (2)	358 (7)
Median Age4	72	69	65	67	69
Age Range4	54 – 89	54 – 89	54 – 83	54 – 90	53 – 94
Employed,4,5 n (%)	18 (40)	25 (56)	27 (57)	27 (59)	2 158 (44)

Monitoring completed in 2013.

Monitoring completed in 2014.

Includes participants in Winston-Salem, NC; Los Angeles, CA; Baltimore, MD; New York City, NY; Chicago, IL; and St. Paul, MN.

At MESA Exam 5 in 2010–2012.

Employed outside of the home.

Only a small number of homes did not have NO2 sources identified during each monitoring campaign (n=8 and n=5 in winter and summer in Winston-Salem, respectively, and n=2 in winter in Los Angeles); no homes during the Los Angeles summer campaign had zero indoor NO2 sources identified. However, even with these small sample sizes, the observed I/O ratios were similar to the literature-based ratio in both cities (Table 2). The literature-based I/O ratios used in this analysis were 0.45 in winter and 0.27 in summer in Winston-Salem, and 0.38 in winter and 0.53 in summer in Los Angeles. The mean observed I/O ratios for without indoor sources were 0.43, 0.37, and 0.49 during monitoring in winter and summer in Winston-Salem and winter in Los Angeles, respectively. In contrast, the mean observed I/O ratios for homes with indoor sources ranged from 1.2 during the winter monitoring in Los Angeles to 2.6 during the summer monitoring in Winston-Salem.

Table 2

Observed (mean ± standard deviation) and literature-based indoor to outdoor (I/O) ratios of NO2 for each monitoring campaign.

City	Season	N1	ObservedI/O ratio	Literature-based2I/O ratio
Winston-Salem	Winter	8	0.43 ± 0.07	0.45
	Summer	5	0.37 ± 0.24	0.27
Los Angeles	Winter	2	0.49 ± 0.02	0.38
	Summer	0	-	0.53

Number of participants with no identified indoor sources of NO2 whose measured NO2 concentrations were used to calculate the observed I/O ratio.

Using AERs from Breen et al (2010) for Winston-Salem and from Isaacs et al (2013) for Los Angeles.[29,30]

Concentrations of NO2 were highest in-vehicle and lowest indoors during each monitoring campaign, while participants spent the majority of their time indoors and on average only 4 to 5% (approximately 1 hr/day) of their time in a motorized vehicle (Table 3). As expected, NO2 concentrations were on average higher in Los Angeles compared to Winston-Salem and tended to be higher outdoors during the winter campaigns than the summer campaigns. Average in-vehicle concentrations of NO2 ranged from 23.0 ppb in summer in Winston-Salem to 38.2 ppb in winter in Los Angeles. The average time per trip was similar across cities (28 minutes). Participants drove on average 27% percent of their trips on major roads (defined as freeways, expressways, highways, or toll roads), and 21% percent of their trips in heavy traffic, with participants in Los Angeles reporting driving more on major roads and in heavy traffic than participants in Winston-Salem.

Table 3

Mean ± standard deviation of NO2 concentration and percent of time spent in each of three microenvironments—indoors, outdoors, and in-vehicle—during four monitoring campaigns.

City	Season	Microenvironment
		Indoor1	Outhoor	In-vehicle2
NO2 Concentration (ppb)
Winston-Salem	Winter	4.1 ± 1.1	9.3 ± 2.5	30.8 ± 12.8
	Summer	1.4 ± 0.7	5.0 ± 2.8	23.0 ± 14.0
Los Angeles	Winter	7.9 ± 1.7	20.7 ± 4.5	36.6 ± 18.6
	Summer	6.7 ± 1.5	12.6 ± 2.8	38.2 ± 15.0
% Time
Winston-Salem	Winter	94 ± 4	2 ± 3	4 ± 2
	Summer	92 ± 6	4 ± 4	4 ± 3
Los Angeles	Winter	91 ± 6	4 ± 5	5 ± 2
	Summer	89 ± 8	7 ± 7	4 ± 2

Indoor ambient-source NO2 was calculated using the measured outdoor concentration and literature-based I/O ratio.

In-vehicle concentrations account for additional sampling time due to air trapped in sampler upon closure.

The indoor microenvironment contributed the highest proportion of individual exposure to ambient-source NO2 (Figure 1) compared to the other two microenvironments monitored in this study, ranging from 53% in Winston-Salem in the summer to 75% in Los Angeles in the winter. However, the in-vehicle microenvironment contributed a larger proportion of exposure than the outdoor microenvironment, overall: 21%, 40%, 17%, and 18% from the in-vehicle microenvironment on average in Winston-Salem in the winter and summer, and Los Angeles in the winter and summer, respectively. Among participants who drove at least 1.3 hours per day (i.e. the highest quartile of time in-vehicle in this sample) during the sampling period, the in-vehicle microenvironment contributed a larger proportion than among those participants who drove less (Figure 2). For these participants, in Winston-Salem in the winter and summer, and in Los Angeles in the winter and summer, the average contribution of the in-vehicle microenvironment to the overall exposure was 34%, 56%, 23%, and 21%, respectively.

Figure 1

Percent contribution of each microenvironment—indoor, outdoor, and in-vehicle—to overall individual exposure to ambient-source NO2 during each monitoring campaign in Winston-Salem (A) and Los Angeles (B). Gray boxes show the 25th percentile to 75th percentile, the horizontal black bar shows the median, the whiskers show either the minimum or 1.5 times the interquartile range below 25th percentile and either the maximum or 1.5 times the interquartile range above the 75th percentile, and points are outliers.

Figure 2

Percent contribution of each microenvironment—indoor, outdoor, and in-vehicle—to overall individual exposure to ambient-source NO2 during each monitoring campaign in Winston-Salem (A) and Los Angeles (B), only for those participants who drove on average at least 1.3 hours per day (i.e. in the highest quartile of time spent in-vehicle, n=38) during the study period. Gray boxes show the 25th percentile to 75th percentile, the horizontal black bar shows the median, the whiskers show either the minimum or 1.5 times the interquartile range below 25th percentile and either the maximum or 1.5 times the interquartile range above the 75th percentile, and points are outliers.

Estimates of total ambient-source NO2 exposure using only outdoor NO2 concentrations were highest compared to estimates incorporating ambient-source indoor or in-vehicle concentrations (Table 4). Estimates using infiltrated indoor concentrations in addition to outdoor concentrations were on average 6.4 ppb lower (standard deviation 3.8 ppb) than estimates using only outdoor concentrations. Incorporating both infiltrated indoor and measured in-vehicle concentrations resulted in exposure estimates on average 5.6 ppb lower (standard deviation 4.0 ppb) than estimates using only outdoor concentrations. Larger absolute differences in the exposures calculated using information from indoor and in-vehicle microenvironments compared to those only using outdoor concentrations, were observed in winter compared to summer and in Los Angeles compared to Winston-Salem. Among the participants in the highest quartile of time spent in-vehicle (at least 1.3 hrs/day in this sample), incorporating in-vehicle exposures resulted in higher estimates than among those who drove less.

Table 4

Total individual ambient-source NO2 exposure calculated using three methods: concentrations measured outdoors only; concentrations outdoors and infiltrated indoors; and concentrations outdoors, infiltrated indoors, and measured in-vehicle.

City andSeason	Method	Individual Ambient-Source NO2 Exposure (ppb) 1
		Overall(N=151)	<1.3 hrs/day driving2(N=113)	≥1.3 hrs/day driving2(N=38)
Winston-Salem
Winter	O	9.2 ± 2.5	9.4 ± 2.5	8.1 ± 2.7
	O+I	4.5 ± 1.3	4.6 ± 1.2	4.2 ± 1.3
	O+I+V	5.2 ± 1.2	5.2 ± 1.3	5.3 ± 0.8
Summer	O	5.0 ± 2.8	5.3 ± 3.0	4.1 ± 1.5
	O+I	1.6 ± 0.9	1.7 ± 0.9	1.5 ± 0.5
	O+I+V	2.4 ± 1.0	2.2 ± 0.9	2.9 ± 1.3
Los Angeles
Winter	O	20.7 ± 4.5	20.8 ± 4.6	20.3 ± 4.5
	O+I	9.0 ± 2.2	8.8 ± 2.3	9.3 ± 2.1
	O+I+V	9.6 ± 2.1	9.4 ± 2.2	10.1 ± 1.8
Summer	O	12.6 ± 2.8	12.6 ± 3.0	12.5 ± 2.4
	O+I	7.3 ± 1.7	7.3 ± 1.9	7.5 ± 1.4
	O+I+V	8.3 ± 1.8	8.3 ± 2.1	8.3 ± 1.1

Abbreviations: O = outdoor concentrations only; O+I = outdoor and infiltrated indoor concentrations; O+I+V = outdoor, infiltrated indoor, and measured in-vehicle concentrations.

Summary statistics are mean ± standard deviation.

Participants in the highest quartile of time spent in-vehicle (75th percentile=1.3 hours/day) were compared to those who drove less.

We also conducted a sensitivity analysis in which we used lower and higher AER values from the literature.[29,30,32] The contribution of indoor infiltration of ambient NO2 to overall exposure ranged from an average of 53% during summer in Winston-Salem using the lowest AER to an average of 84% during winter in Los Angeles using the highest AER (see Table S2). Even with the lowest AER, the indoor contribution to total exposure remained higher on average than the in-vehicle contribution.

DISCUSSION

Using novel in-vehicle samplers deployed concurrently with indoor and outdoor monitors, this study demonstrated the importance of the in-vehicle and indoor microenvironments for accurate exposure characterization. The in-vehicle microenvironment, on average, contributed more to overall exposure to NO2 than the outdoor microenvironment. However, exposure indoors to NO2 originating outdoors accounted for the majority of ambient-source NO2 exposure within this older population.

A limited number of studies have investigated the role of the in-vehicle microenvironment in overall exposure,[33] with some relying on modeled commute exposures rather than measured in-vehicle concentrations.[34] A study in Barcelona found that in-vehicle exposures may contribute up to 24% of overall NO2 exposure, even for participants spending on average only 6% of time in-vehicle,[35] which is a similar finding to the current study. The in-vehicle microenvironment may contribute to measurement error in exposure assessment, particularly among those who travel most and in areas where the outdoor levels of air pollution are generally low.[33,36]

Indoor concentrations were the largest contributor to overall individual exposure, particularly in the city with the highest outdoor concentrations, and thus focus on improved estimation of ambient-source indoor exposures is warranted. Studies in a variety of U.S. cities have examined the effect of various housing characteristics on AERs[29,30,37,38] and I/O ratios of NO2,[39] and report AERs ranging from 0.36 h−1 in North Carolina during the summer[29] to 1.92 h−1 in Michigan during the summer.[30] In a sensitivity analysis, variation in AER altered the proportion of overall exposure due to the indoor microenvironment, though even with the lowest AER the mean contribution of the indoor microenvironment remained greater than 50%. Differences in AER are attributed to meteorological conditions including wind speed, indoor versus outdoor temperature differences, and humidity, as well as housing ventilation choices such as open windows versus air conditioning.[29,30,37] Such characteristics may be important to consider in further characterizing air pollution exposure variability to reduce measurement error.

While some evidence suggests that ignoring in-vehicle exposure may contribute to bias in epidemiologic studies,[40-42] others have found only a small contribution from the in-vehicle microenvironment and suggest that the indoor microenvironment may be the most important source of measurement error in air pollution epidemiology.[43,44] Epidemiologic studies have most often used modeled outdoor concentrations at the residential location, which may overestimate overall exposure as shown in the current study. Future studies may reduce bias by accounting for exposure magnitude and variability related to the indoor and in-vehicle microenvironments.

The measurement error contributed to epidemiological studies by ignoring high exposures from in-vehicle microenvironments could bias results. If exposure assessment systematically underestimates variation in exposure across study participants, the main effect would be over-estimation of the concentration-response slope. If exposure assessment misclassifies exposure in a random way, the effect will typically (although not always) be biased toward the null and underestimation of uncertainty in the concentration-response relationship. If the misclassification is correlated with the outcome or other predictors of outcome, the effect may be confounded and could be either over- or under-estimated. It is difficult to predict which of these is most likely.[45]

The in-vehicle concentrations measured in this study (overall median two-week NO2 concentration: 30.4 ppb) are consistent with near-roadway regulatory monitoring in California; no near-roadway sites have been established in Winston-Salem. Specifically, at the designated near-roadway regulatory site in Los Angeles County, which was intentionally located near a major roadway with heavy traffic, the annual average NO2 concentration in 2015 was 23.9 ppb.[46]

The concentrations measured in this study are also consistent with prior literature examining NO2 concentrations inside of vehicles. However, most studies of adults utilized occupational settings such as patrol cars[47] and taxis,[33,48-50] or scripted drive protocols.[23,51] Furthermore, only a small number of these studies were conducted in North America.[21,23,47,51] The earliest of these studies, conducted in Raleigh, NC in 1988, reported a median in-vehicle NO2 concentration of 43.1 ppb during scripted drives in both the morning and evening rush hour.[21] In patrol cars during 3pm to 12am shifts in and around Raleigh, NC, the median concentration was 22.2 ppb.[47] The median NO2 concentration measured inside the vehicle during scripted drives primarily on the New Jersey Turnpike during morning commute hours was 20.6 ppb[51] and in three Canadian cities during three hour commutes was 39.0 ppb.[23] Most prior studies monitored NO2 for a limited amount of time (e.g. only during rush hour or only during some shifts) or on small number of days. Our study builds upon this literature by implementing in-vehicle monitoring in a population—selected for reasons other than occupational exposures—who were instructed to go about their usual activities (un-scripted) over a two-week period in two distinct cities.

While this study expanded on prior work to understand the influence of in-vehicle exposures by implementing a novel sampling campaign, it was also subject to several limitations. First, this population is a sample of older adults, only about half of whom are employed, and thus their driving habits, time-location patterns, and exposures while driving may limit generalizability to younger populations. Second, limitations in the passive sampling equipment due to the unscripted nature of this study and requirement for ease of use by participants, required additional post-hoc adjustments to calculate in-vehicle concentrations of air pollutants. We assumed that the in-vehicle NO2 concentration at the end of each trip was equal to the trip average, though it is possible that this assumption was not correct in each case. The adjustments also assumed a concentration-independent sampling rate within the closed in-vehicle sampler and thus are likely an oversimplification. Third, we used NO2 as a marker for the suite of potential air pollutants associated with traffic proximity. It is possible that NO2 is not a perfect indicator. For example, non-reactive pollutants such as carbon monoxide or volatile organic compounds such as 1,3 butadiene, may have higher contributions from in-vehicle exposures.

Even given the limitations mentioned above, this study deployed a novel sampling design to measure NO2 as a marker of TRAP in an older population with well-characterized time-activity patterns. In-vehicle concentrations were relatively similar across cities, but outdoor and infiltrated indoor concentrations varied substantially. Therefore, resulting in-vehicle exposures made up a larger proportion of overall NO2 exposure in the city with lower outdoor air pollution levels and lower air exchange rates (i.e., in Winston-Salem, compared to Los Angeles). This study suggests that while in-vehicle exposure is an important component of overall individual-level NO2 exposure, overall exposure in this older population is largely determined by ambient-source exposure inside the residence.