Abstract

Objectives: An exposure assessment method was developed for use in assigning an exposure score to New York State personnel who responded to the World Trade Center disaster site after the 11 September 2001 terrorist attacks.

Methods: The method consists of an algorithm with two instantiations. Each represents a major component of the overall exposures at the site: dust and smoke. The algorithm uses US Environmental Protection Agency air monitoring data collected between 23 September 2001 and 28 February 2002, as well as information on duration, location and time period of work assignment and type and frequency of personal protective respiratory equipment (PPE) use, collected by a self-administered mailed questionnaire. These data were used to calculate an overall exposure score for each participant. For each time period/location combination, individuals provided average number of hours and number of days worked. This was multiplied by a weighting factor derived from the median of the air monitoring data for the time period/location. Calcium was chosen as a surrogate for the dust exposure, so the weighting factors for the dust instantiation were calculated from calcium air monitoring data. Total hepta-chlorinated dibenzo-p-dioxin was chosen as a surrogate for the smoke exposure and was similarly used in the smoke instantiation.

Results: More individuals in the highest exposure score category performed tasks such as search/rescue and hand digging than those in the lowest exposure category. Also, those in the highest exposure category had a higher mean number of hours at the site than other exposure groups.

Conclusions: The exposure assessment method presented accounts for PPE use, amount of time at the site, proximity to the site and ambient air monitoring results taken in the immediate vicinity. The algorithm can be used to rank individuals in the same study with very different patterns of exposure, such as high-level, short-term exposures and low-level, long-term exposures. The concepts could be modified for use in other epidemiological studies where long-term chronic exposure is a concern.

INTRODUCTION

Individuals who responded to the 11 September 2001, World Trade Center (WTC) disaster were exposed to various levels of smoke, fumes, dust and debris generated by the collapse of the buildings and the ongoing fires. Dust contained a variety of materials including cement, silica and asbestos. Three settled dust/smoke samples collected 5–6 days after the disaster in areas that ranged from two blocks to 0.7 km from the site contained sulfate and calcium in relatively large amounts (in nanograms/grams) (Lioy et al., 2002). Also detected in settled dust/smoke were heavy metals, polychlorinated biphenyls, polycyclic aromatic hydrocarbons (PAHs), polychlorinated dibenzo-p-dioxins (PCDDs) and polychlorinated dibenzofurans (PCDFs) (Lioy et al., 2002). Other samples collected 1–2 days after the disaster and within one-half mile of the site were found to have calcium sulfate and calcium carbonate as major components of the particulate matter (PM 2.5) fraction of the settled dust (McGee et al., 2003). Analyzed aerosol plumes contained finely powdered concrete, gypsum and glass with soot-like coatings and anthropogenic metals. The aerosols also contained sulfuric acid and organic matter, including PAHs and their derivatives, in very fine particles (Cahill et al., 2004). The intensity and magnitude of the combustion process at the WTC suggest that unusual products of combustion may also have been formed and released in quantity during this incident. These include trace metals, as well as unusual persistent organic chemicals formed by pyrolysis and combustion of the numerous carbon-based products in the building. During the first days after the incident, many workers did not use appropriate respiratory protection, thus increasing the risk of exposures.

The intent of this project was to develop an exposure assessment method that could be used to assign a semiquantitative overall exposure score to each study participant in New York State Department of Health (NYSDOH) epidemiological studies of WTC responders. Given the complex mixture of compounds present, our goal was not to estimate actual exposures to any specific compounds, but rather to create a scale of likely relative overall exposures that responders may have experienced. An algorithm was created that has two instantiations, one for dust exposures and one for smoke exposures, because these two categories of exposure could potentially be quite different.

The algorithm was developed in response to the needs of WTC studies conducted by the NYSDOH. These included a study of medical monitoring-related data, a pilot biomonitoring study and studies evaluating asthma and respiratory impacts of WTC exposure. In each of these studies, the exposure assessment method described here is being used. Results from these studies will be described in subsequent papers.

METHODS

Algorithm development

Data for the algorithm were collected from two main sources: a mailed questionnaire and US Environmental Protection Agency (USEPA) air monitoring data. Methods and descriptive statistics from the mailed questionnaire are described elsewhere (Mauer et al., 2007). Briefly, the questionnaire asked participants about health outcomes including symptoms and diagnoses and information regarding time spent working at the WTC disaster site. The questionnaire was mailed to all individuals identified by New York State (NYS) as having been assigned to work at the WTC site between 11 September 2001 and 23 December 2001 and who participated in an offered medical monitoring program. Individuals who agreed to participate in a medical monitoring-related study and who had information on work history at the site were included in the data set used to develop the algorithm (n = 1341) (response rate 14%). Individuals were asked to complete the questionnaire and bring it with them to the medical evaluation that was part of the medical monitoring program.

The work assignment portion of the self-administered questionnaire was divided into seven time periods: (i) 11 September, (ii) 12 September, (iii) 13–16 September, (iv) 17–30 September, (v) 1–31 October, (vi) 1 November to 31 December and (vii) 1 January and after. The time periods were chosen to facilitate recall of work patterns, while attempting to collect exposure data in sufficient detail. Other researchers have defined early time periods in a similar, though not identical, manner (Prezant et al., 2002; Edelman et al., 2003; Lioy et al., 2006). The study described here encompassed a longer time span and made more specific distinctions from the end of September through the end of December than other long-term studies (Lioy et al., 2006). Although some of the dates chosen do not directly correspond to some significant events in the aftermath of the disaster (rain events in September and the switch from rescue to recovery efforts in mid-October), they correspond to more easily recalled calendar dates. Earlier time periods are shorter than later time periods because we anticipated exposures might change more dramatically soon after the event compared to later. By doing this, we created time periods small enough to capture some variation in exposure, while minimizing recall bias. The questionnaire also identified six exposure locations (Fig. 1): (i) on the pile; (ii) adjacent to the pile; (iii) in the secure area; (iv) outside the secure area, but within five blocks; (v) transfer or wash station and (vi) Fresh Kills Landfill. A matrix was provided, and for each time period/location, individuals were asked to record the number of days worked and the average number of hours per day.

Fig. 1.

Map of study area. Pile = WTC footprint. Adjacent = streets bordering the WTC site. Secure area = secure zone from January 2002, established by the Office of Emergency Management, City of New York. This zone changed over time and this map represents the smallest secure zone of the study period. Outside area = anywhere outside the secure border but within five blocks of the WTC site. • = site of a USEPA air monitoring station.

Fig. 1.

Map of study area. Pile = WTC footprint. Adjacent = streets bordering the WTC site. Secure area = secure zone from January 2002, established by the Office of Emergency Management, City of New York. This zone changed over time and this map represents the smallest secure zone of the study period. Outside area = anywhere outside the secure border but within five blocks of the WTC site. • = site of a USEPA air monitoring station.

For each time period, individuals were asked what type of personal protective equipment (PPE) they used, with specific focus on respiratory protection. Six respiratory protection items were listed: one-strap dust mask, two-strap dust mask, respirator (cartridge), respirator [powered air-purifying respirator (PAPR)], respirator [self-contained breathing apparatus (SCBA)] and respirator (unknown). Furthermore, individuals were asked how often they used respiratory protection, if at all (choices were always, more than half the time, less than half the time or never).

Attenuation factors, representing the attenuated amount of exposure reaching the individual, were assigned to each type of respiratory protection listed in the questionnaire. Attenuation factors were developed in consultation with NYSDOH Industrial Hygienists and were based on established Occupational Safety and Health Administration guidelines (Department of Health and Human Services (NIOSH), 1987). When cartridge respirators were used, we assumed that the proper cartridge for potential exposures was used. Information on fit testing was not available. We assume that if respirators were used, fit testing had been conducted, although we acknowledge that this may not have been the case. The attenuation factors are listed in Table 1. SCBA respirators provide virtually complete protection with proper fit testing and use, while one-strap dust masks provide virtually no protection according to the National Agricultural Safety Database (Centers for Disease Control and Prevention, 2002).

Table 1.

Attenuation factors assigned to personal respiratory protection equipment (factors assume proper fit testing and use)

Respiratory protection Attenuation factor 
SCBA 0.0001 
PAPR 0.02 
Cartridge respirator 0.10 
Two-strap dust mask 0.10 
One-strap dust mask 1.00 
Unknown 1.00 
Respiratory protection Attenuation factor 
SCBA 0.0001 
PAPR 0.02 
Cartridge respirator 0.10 
Two-strap dust mask 0.10 
One-strap dust mask 1.00 
Unknown 1.00 

Air monitoring data received from the USEPA, Region 2 [data from fixed ambient air sampling locations for calcium (23 September 2001 to May 2002) and PCDDs and PCDFs (16 September 2001 to May 2002)] were also used (United States Environmental Protection Agency, 2003). This data set provided the best temporal and spatial coverage of the study area because there were many stationary locations throughout a five-block radius around Ground Zero. Data from stationary monitors were considered a good choice for this project because the participants in the study had a wide array of job duties at various locations throughout lower Manhattan. For example, tasks ranged from traffic control at the perimeter of the secure area to hand digging on the pile. All monitors used in this algorithm were set up at breathing zone height (∼2 m).

Beginning on 23 September 2001, at least one location selected by the USEPA for air monitoring fell within each of the four lower Manhattan exposure locations listed in the questionnaire (Fig. 1). Data from 16 September 2001 were excluded from consideration here because there were no measurements in some study areas. Data after 28 February 2002 were also excluded because very few participants were at the site beyond that date.

The transfer/wash stations and Fresh Kills Landfill were assumed to be equal to the secure area. Relatively few individuals were ever at these sites, and those who were did not spend long periods of time in these locations. Furthermore, air sampling at Fresh Kills Landfill was set up to detect any compounds moving off-site and not to evaluate potential exposure to people working on the site. Thus, the timing and positioning of air sampling at Fresh Kills Landfill did not fit the needs of this study. Among areas covered by the data set, the secure area was thought to be most similar to these sites because of the percent of individuals in the study involved in major task groups (data not shown).

Although PM was a major component of the overall exposure at the WTC in the period following the collapse of the two towers, a PM data set illustrating spatial variation at the site was lacking. However, many individual metal compounds were monitored near the disaster site and several provided the basis for a data set with few non-detects. Thus, calcium was chosen as a surrogate for dust exposure. Calcium was a major component of the building materials present in the towers, which were pulverized when the collapse occurred. Analysis of dust samples confirm a relatively high concentration of calcium (Lioy et al., 2002; McGee et al., 2003). Furthermore, the calcium data set had few non-detects.

In addition to dust, people were exposed to smoke and combustion products from the fires. Dioxins and furans are a group of compounds released when many types of materials burn. Total hepta-chlorinated dibenzo-p-dioxins (total hepta-CDDs) were chosen as a surrogate for the smoke exposure. This grouping encompasses two isomers—1,2,3,4,6,7,8-HPCDD and 1,2,3,4,7,8,9-HPCDD. Although many different dioxin and furan compounds are released during combustion, total hepta-CDDs were chosen because the data set provided adequate sampling for this group compared to other groups of congeners. Total hepta-CDDs were among the CDD congeners most frequently detected in the USEPA air monitoring data. In addition, 1,2,3,4,6,7,8-HPCDD had the second highest concentration among the eight PCDD isomers analyzed from three settled dust samples (Lioy et al., 2002).

Each of the two surrogates selected is intended to represent a variety of compounds that have a common source. Two separate instantiations of the algorithm were thus developed for these two types of exposure, using calcium data as a surrogate for dust exposures and total hepta-CDDs as a surrogate for smoke exposures.

The median of all air monitoring data points for the chosen surrogate for each time period/location was assigned to that cell as a time period/location weight. If several monitoring stations were located within the boundaries of that location, all data were grouped together for calculation of the median. Non-detect values were assigned a zero. The values of the time period/location weights are listed in Table 2. Note that the first three time periods are equal to time period 4 because the ambient air monitoring data used began on 23 September 2001. Therefore, there are no measurements for time periods encompassing the first week after the disaster. Hepta-CDD data were converted from nanograms meter−3 to femtograms meter−3 for calculations of the smoke exposure scores. Final smoke scores were then divided by 1 000 000 to restore the scale.

Table 2.

Time period/location weights

Time period Location 
 Pile Adjacent Secure/wash/landfill Outside 
 Dusta 
T1, 11 September 2001 44 43.5 41 7.4 
T2, 12 September 2001 44 43.5 41 7.4 
T3, 13–16 September 2001 44 43.5 41 7.4 
T4, 17–30 September 2001 44 43.5 41 7.4 
T5, 1–31 October 2001 47 18.5 21 3.4 
T6, 1 November to 31 December 2001 29 21 12 3.2 
T7, 1 January 2002+b 5.85 8.4 6.5 2.48 
 Smokec 
T1, 11 September 2001 920 000 590 000 80 000 2000 
T2, 12 September 2001 920 000 590 000 80 000 2000 
T3, 13–16 September 2001 920 000 590 000 80 000 2000 
T4, 17–30 September 2001 920 000 590 000 80 000 2000 
T5, 1–31 October 2001 312 000 111 000 34 000 900 
T6, 1 November to 31 December 2001 29 300 2600 1500 900 
T7, 1 January 2002 and afterb 450 
Time period Location 
 Pile Adjacent Secure/wash/landfill Outside 
 Dusta 
T1, 11 September 2001 44 43.5 41 7.4 
T2, 12 September 2001 44 43.5 41 7.4 
T3, 13–16 September 2001 44 43.5 41 7.4 
T4, 17–30 September 2001 44 43.5 41 7.4 
T5, 1–31 October 2001 47 18.5 21 3.4 
T6, 1 November to 31 December 2001 29 21 12 3.2 
T7, 1 January 2002+b 5.85 8.4 6.5 2.48 
 Smokec 
T1, 11 September 2001 920 000 590 000 80 000 2000 
T2, 12 September 2001 920 000 590 000 80 000 2000 
T3, 13–16 September 2001 920 000 590 000 80 000 2000 
T4, 17–30 September 2001 920 000 590 000 80 000 2000 
T5, 1–31 October 2001 312 000 111 000 34 000 900 
T6, 1 November to 31 December 2001 29 300 2600 1500 900 
T7, 1 January 2002 and afterb 450 
a

Time period/location values used for the dust algorithm are the median values of air monitoring data for calcium for the specific time period/location (USEPA).

b

For calculating the time period/location weight, air sampling data through February 2002 were used.

c

Time period/location values used for the smoke algorithm are the median values of air monitoring data for total hepta-chlorinated dibenzo-p-dioxins for the specific time period/location (USEPA). Values were converted from nanograms meter−3 to femtograms meter−3 to allow whole numbers for calculations.

The final algorithm can be summarized as aformula, as follows, using information from the questionnaire, PPE attenuation factors from Table 1 and the time period/location weights from Table 2: 

graphic
Table 3 shows sample data from the questionnaire for time and location worked. The table also contains sample calculations for several calculated variables arising from pieces of the formula above. Finally, the table contains an overall score for the example.

Table 3.

Example of a typical work history for WTC work assignments as reported in a mailed survey and calculation of scores using the formula: forumla

Time period  Location Scores 
  Pile Adjacent Secure Transfer/wash station Fresh Kills Landfill Outside secure area within five blocks PPE use PPE frequency of use Duration Time period exposurec Unprotected exposure Protected exposured Total 
T1 Hours         
Days       
T2 Hours  12     None Not applicable 12 522 522 522 
Days      
T3 Hours   12    Cartridge More than halfa 48 1968 492 147.6 639.6 
Days      
T4 Hours         
Days       
T5 Hours     Cartridge More than halfa 84 1271.2 317.8 95.3 413.1 
Days     
T6 Hours    PAPR Less than halfb 84 887.6 665.7 4.4 670.1 
Days    
T7 Hours         
Days       
             Overall 2244.8 
Time period  Location Scores 
  Pile Adjacent Secure Transfer/wash station Fresh Kills Landfill Outside secure area within five blocks PPE use PPE frequency of use Duration Time period exposurec Unprotected exposure Protected exposured Total 
T1 Hours         
Days       
T2 Hours  12     None Not applicable 12 522 522 522 
Days      
T3 Hours   12    Cartridge More than halfa 48 1968 492 147.6 639.6 
Days      
T4 Hours         
Days       
T5 Hours     Cartridge More than halfa 84 1271.2 317.8 95.3 413.1 
Days     
T6 Hours    PAPR Less than halfb 84 887.6 665.7 4.4 670.1 
Days    
T7 Hours         
Days       
             Overall 2244.8 
a

Assumes 75% use.

b

Assumes 25% use.

c

Incorporates appropriate time period/location weights from Table 2 (dust instantiation).

d

Incorporates appropriate attenuation factors from Table 1.

Sensitivity analysis

Sensitivity analysis was performed on the algorithm by making various substitutions for terms in each instantiation. Substitutions changed the definition of frequency of PPE use including using <100% when a participant responded that they used PPE ‘always’ and percent use assumed when frequency was missing. (The algorithm assumes 50% use when a type of respirator was selected, but no frequency of use was selected.) Other substitutions changed the way the air monitoring data were used. One sub-analysis fitted a single, three-parameter exponential decay regression line to the plotted median values and back extrapolated the data for the first three time periods (SigmaPlot v. 8.02, SPSS, Inc.). Fit statistics (R2) for each of the regression lines were ≥0.90 with the exception of the plot showing median calcium values for the pile (R2 = 0.82). A final sub-analysis plotted each individual data point by date (rather than time period median value) and fit a linear regression line through the points. Concentration of the compound was then read at the date of the midpoint of each study time period and used as the time period/location weight.

Algorithm use

Exposure scores were computed for each individual in the medical monitoring cohort who had acompleted WTC work assignment section (n = 1341) by running each instantiation of the algorithm with the self-reported data from the questionnaires. Based on these scores, individuals were categorized into four groups (quartiles), with Quartile 1 including the lowest exposure scores and Quartile 4 including the highest exposure scores.

RESULTS

Overall, the study population engaged in varied activities over several months in a variety of locations around Ground Zero. NYS assigned its employees to work at the disaster site beginning on 11 September 2001, continuing into mid-2002. The amount of time spent at the site ranged from every day for months to less than one full day. Individuals often worked in multiple locations, as well as during multiple study time periods. Table 4, which uses data for dust exposure, shows the number of participants assigned to the various study locations/time periods by exposure score quartile. Table 4 also shows major categories of participants’ job tasks by exposure score quartile. Finally, Table 4 shows the number of participants who used respiratory protection and the type used, by exposure score quartile. A larger percentage of participants in the highest quartile of exposure scores spent at least some time on the pile and in the adjacent and secure areas, where exposures are expected to have been higher, compared to all other quartiles. In contrast, a smaller percentage of participants in the highest quartile spent any time in the outside area where exposures are expected to have been lowest. Overall, when considering duration at the site and location/time period, individuals in the highest quartile spent more time at the site on average (mean duration in hours) and, specifically, more time on the pile, and in the adjacent and secure areas, where the exposures were greater (data not shown). Also, participants in the highest exposure score group spent more time on average in every time period (data not shown).

Table 4.

Descriptive statistics of study populationa

 Overall
 
Quartile 1 (lower exposure)
 
Quartile 2
 
Quartile 3
 
Quartile 4 (higher exposure)
 
 n = 1341
 
n = 336
 
n = 335
 
n = 335
 
n = 335
 
 n(%) n(%) n(%) n(%) n(%) 
Location 
    Pile 299 (22.3) 38 (11.3) 61 (18.2) 77 (23.0) 123 (36.7) 
    Adjacent 816 (60.9) 178 (53.0) 195 (58.2) 213 (63.6) 230 (68.7) 
    Secure 709 (52.9) 134 (39.9) 155 (46.3) 185 (55.2) 235 (70.2) 
    Wash station 113 (8.4) 22 (6.6) 28 (8.4) 33 (9.9) 30 (9.0) 
    Landfill 87 (6.5) 20 (6.0) 26 (7.8) 30 (9.0) 11 (3.3) 
    Outside 650 (48.5) 173 (51.5) 203 (60.6) 162 (48.4) 112 (33.4) 
Time period 
    T1 242 (18.0) 39 (11.6) 40 (11.9) 66 (19.7) 97 (29.0) 
    T2 458 (34.2) 72 (21.4) 83 (24.8) 131 (39.1) 172 (51.3) 
    T3 762 (56.8) 128 (38.1) 175 (52.2) 224 (66.9) 235 (70.2) 
    T4 921 (68.7) 157 (46.7) 203 (60.6) 252 (75.2) 309 (92.2) 
    T5 699 (52.1) 112 (33.3) 186 (55.5) 175 (52.2) 226 (67.5) 
    T6 437 (32.6) 48 (14.3) 113 (33.7) 132 (39.4) 144 (43.0) 
    T7 221 (16.5) 19 (5.7) 56 (16.7) 68 (20.3) 78 (23.3) 
Job task 
    Security 747 (55.7) 127 (37.8) 179 (53.4) 206 (61.5) 235 (70.2) 
    Traffic control 448 (33.4) 76 (22.6) 112 (33.4) 121 (36.1) 139 (41.5) 
    Emergency responseb 181 (13.5) 24 (7.1) 22 (6.6) 59 (17.6) 76 (22.7) 
    Demolitionc 75 (5.6) 25 (7.4) 21 (6.3) 22 (6.6) 7 (2.1) 
    Engineering 89 (6.6) 24 (7.1) 18 (5.4) 26 (7.8) 21 (6.3) 
    Otherd 419 (31.3) 114 (33.9) 99 (29.6) 102 (30.5) 104 (31.0) 
Respiratory protection 
    SCBAe 10 (0.8) 3 (0.9) 1 (0.3) 2 (0.6) 4 (1.2) 
    PAPRf 9 (0.7) 2 (0.6) 2 (0.6) 4 (1.2) 1 (0.3) 
    Cartridge 297 (22.2) 68 (20.2) 66 (19.7) 85 (25.4) 78 (23.3) 
    Two-strap dust mask 316 (23.6) 91 (27.1) 62 (18.5) 85 (25.4) 78 (23.3) 
    One-strap dust mask 385 (28.7) 48 (14.3) 79 (23.6) 112 (33.4) 146 (43.6) 
    Unknown type 115 (8.6) 12 (3.6) 26 (7.8) 26 (7.8) 51 (15.2) 
 Overall
 
Quartile 1 (lower exposure)
 
Quartile 2
 
Quartile 3
 
Quartile 4 (higher exposure)
 
 n = 1341
 
n = 336
 
n = 335
 
n = 335
 
n = 335
 
 n(%) n(%) n(%) n(%) n(%) 
Location 
    Pile 299 (22.3) 38 (11.3) 61 (18.2) 77 (23.0) 123 (36.7) 
    Adjacent 816 (60.9) 178 (53.0) 195 (58.2) 213 (63.6) 230 (68.7) 
    Secure 709 (52.9) 134 (39.9) 155 (46.3) 185 (55.2) 235 (70.2) 
    Wash station 113 (8.4) 22 (6.6) 28 (8.4) 33 (9.9) 30 (9.0) 
    Landfill 87 (6.5) 20 (6.0) 26 (7.8) 30 (9.0) 11 (3.3) 
    Outside 650 (48.5) 173 (51.5) 203 (60.6) 162 (48.4) 112 (33.4) 
Time period 
    T1 242 (18.0) 39 (11.6) 40 (11.9) 66 (19.7) 97 (29.0) 
    T2 458 (34.2) 72 (21.4) 83 (24.8) 131 (39.1) 172 (51.3) 
    T3 762 (56.8) 128 (38.1) 175 (52.2) 224 (66.9) 235 (70.2) 
    T4 921 (68.7) 157 (46.7) 203 (60.6) 252 (75.2) 309 (92.2) 
    T5 699 (52.1) 112 (33.3) 186 (55.5) 175 (52.2) 226 (67.5) 
    T6 437 (32.6) 48 (14.3) 113 (33.7) 132 (39.4) 144 (43.0) 
    T7 221 (16.5) 19 (5.7) 56 (16.7) 68 (20.3) 78 (23.3) 
Job task 
    Security 747 (55.7) 127 (37.8) 179 (53.4) 206 (61.5) 235 (70.2) 
    Traffic control 448 (33.4) 76 (22.6) 112 (33.4) 121 (36.1) 139 (41.5) 
    Emergency responseb 181 (13.5) 24 (7.1) 22 (6.6) 59 (17.6) 76 (22.7) 
    Demolitionc 75 (5.6) 25 (7.4) 21 (6.3) 22 (6.6) 7 (2.1) 
    Engineering 89 (6.6) 24 (7.1) 18 (5.4) 26 (7.8) 21 (6.3) 
    Otherd 419 (31.3) 114 (33.9) 99 (29.6) 102 (30.5) 104 (31.0) 
Respiratory protection 
    SCBAe 10 (0.8) 3 (0.9) 1 (0.3) 2 (0.6) 4 (1.2) 
    PAPRf 9 (0.7) 2 (0.6) 2 (0.6) 4 (1.2) 1 (0.3) 
    Cartridge 297 (22.2) 68 (20.2) 66 (19.7) 85 (25.4) 78 (23.3) 
    Two-strap dust mask 316 (23.6) 91 (27.1) 62 (18.5) 85 (25.4) 78 (23.3) 
    One-strap dust mask 385 (28.7) 48 (14.3) 79 (23.6) 112 (33.4) 146 (43.6) 
    Unknown type 115 (8.6) 12 (3.6) 26 (7.8) 26 (7.8) 51 (15.2) 

Quartiles based on dust/debris algorithm.

a

Participants can be included in multiple sub-categories for each major category listed, so percentages will not add to 100%.

b

Includes hand digging, firefighting, search and rescue and bucket brigade.

c

Includes torch cutting, heavy equipment operation and gas-powered tool operation.

d

Including, but not limited to, counselors, food service workers, air sampling, utility work and transportation.

e

Self-contained breathing apparatus respirator.

f

Powered air-purifying respirator.

Table 4 also shows a larger percentage of participants in the highest quartile engaged in emergency response job tasks that would have involved direct contact with pulverized building material and smoke plumes (hand digging, search and rescue, etc.). In fact, compared to the lowest exposure group, over three times the number of participants in the highest exposure group performed these tasks. Furthermore, the second highest exposure group had ∼2.5 times the number of participants performing these tasks, compared to the lowest exposure group. Participants in the highest quartile used respiratory protection more frequently than those in other quartiles (data not shown). However, a large percentage of participants in the highest exposure quartile used one-strap dust masks at some time, which provide essentially no protection (Centers for Disease Control and Prevention, 2002). Compared to the lowest exposure group, about three times more participants in the highest exposure group used a one-strap dust mask.

Table 5 shows correlation statistics between the original algorithm and those with various substitutions from the sensitivity analysis. Each of the substitutions was highly correlated to the original algorithm. For the dust instantiation, the Spearman correlation coefficients were >0.98 with the exception of the back-extrapolation substitution; however, this substitution was still highly correlated (r = 0.86). For the smoke instantiation, all the correlation coefficients were ≥0.90.

Table 5.

Sensitivity Analysis

Algorithm substitution Dust Smoke 
Algorithma 1.00 1.00 
Meanb 0.99 0.92 
Back extrapolationc 0.86 0.97 
Individual pointsd 0.98 0.90 
95% Respirator frequencye 0.99 0.99 
90% Respirator frequencyf 0.99 0.99 
90%/10% Respirator frequencyg 0.99 0.99 
60%/40% Respirator frequencyh 0.99 0.99 
Respirator frequency 95%/60%/40%i 0.99 0.99 
Average frequency of respirator usej 0.99 0.99 
Average frequency of respirator use by typek 0.99 0.99 
Algorithm substitution Dust Smoke 
Algorithma 1.00 1.00 
Meanb 0.99 0.92 
Back extrapolationc 0.86 0.97 
Individual pointsd 0.98 0.90 
95% Respirator frequencye 0.99 0.99 
90% Respirator frequencyf 0.99 0.99 
90%/10% Respirator frequencyg 0.99 0.99 
60%/40% Respirator frequencyh 0.99 0.99 
Respirator frequency 95%/60%/40%i 0.99 0.99 
Average frequency of respirator usej 0.99 0.99 
Average frequency of respirator use by typek 0.99 0.99 

Spearman correlation coefficients for the various algorithm substitutions with the algorithm used for analysis.

a

Original algorithm used for analysis.

b

Substitution of the mean of time period/location air monitoring data for median of time period/location air monitoring data.

c

Substitution of time period/location weights for the first three time periods based on the back extrapolation of the median air monitoring data.

d

Substitution of time period/location weights based on the midpoint of the time period when each point is plotted individually by date and a linear regression line is fitted.

e

Substitution of 95% for the definition of ‘always’ for respiratory protection frequency of use, rather than 100%.

f

Substitution of 90% for the definition of ‘always’ for respiratory protection frequency of use, rather than 100%.

g

Substitution of 90 and 10%, respectively, for the definition of ‘more than half the time’ and ‘less than half the time’ for respiratory protection frequency of use, rather than 75 and 25%, respectively.

h

Substitution of 60 and 40%, respectively, for the definition of ‘more than half the time’ and ‘less than half the time’ for respiratory protection frequency of use, rather than 75 and 25%, respectively.

i

Substitution of 95, 60 and 40%, respectively, for the definition of ‘always', ‘more than half the time’ and ‘less than half the time’ for respiratory protection frequency of use, rather than 100, 75 and 25%, respectively.

j

Substitution of the average frequency of respirator use for missing frequency of use values when a type of respirator was selected, rather than 50%.

k

Substitution of the average frequency of respirator use, by type of respirator selected, for missing frequency of use values when a type of respirator was selected, rather than 50%.

The correlation between the two instantiations is 0.7445. Although the two instantiations are correlated, the exposure scores produced by the two instantiations of the algorithm may be different. Individuals may fall into different categories or quartiles when data is categorized depending on which instantiation is used, because two different surrogate compounds are used to compute time period/location weights.

DISCUSSION

We have developed an exposure assessment method for use with existing NYSDOH epidemiological studies of NYS personnel who responded to the WTC disaster. Applications have included random sampling of participants in the highest and lowest quartiles for each instantiation for inclusion in a pilot biomonitoring study of collected biological specimens. In addition, epidemiological studies have explored the use of the exposure scores in regression analyses. Some published studies to date have used basic exposure measurements (Prezant et al., 2002; Feldman et al., 2004; Salzman et al., 2004; Skloot et al., 2004); specifically, the occupational groups in these studies (firefighters, police first responders and ironworkers) were characterized by time of arrival and/or duration on-site. In another study, firefighters were characterized by time of arrival, as well as number of workdays at the site, unit assignment (surrogate for task) and a dichotomous variable (present versus not present) (Edelman et al., 2003). A more detailed exposure assessment was made by Wolff et al. in a study of exposures among pregnant women near the WTC site. This group of researchers developed an exposure index based on estimated time spent in five zones around the WTC and plume reconstruction modeling (Wolff et al., 2005). Our method includes a combination of quantitative components, including objective environmental measurements and self-reported individual-specific exposure data. The algorithm incorporates several aspects of exposure in one measure and can be used to rank individuals who had very different patterns of exposure, such as high-level, short-term exposures and low-level, long-term exposures.

The approach we have developed for characterizing relative exposure among study participants, while not directly applicable to other existing studies or data sets, includes useful concepts that could be applied by future investigators conducting studies under similar circumstances, e.g. long-term, chronic exposures with or without short-term intense exposure. These situations could be accidental or intentional disasters such as building collapses, chemical plant or oil refinery accidents or fires. This method could potentially be applied to occupational settings as well with some modifications. For example, the locations could be within a workplace, time periods could be years or pre- and post-workplace modifications and time period/location weights could be 8-h maximums or annual averages for any exposure of concern that is monitored. In a post-disaster scenario, adequate occupational and environmental exposure monitoring data may not be available or may not be directly applicable for populations being studied. Some investigators may not have access to sophisticated plume modeling and reconstruction methods (Gilliam et al., 2005; Wolff et al., 2005). Many epidemiological studies designed retrospectively have exposure assessment limitations, such as quantitative data that were collected for another purpose. However, by carefully collecting self-reported data on the general characteristics of population exposures, and combining that data with the most pertinent, publicly available exposure data, researchers may better characterize relative exposure levels among their study participants. While less optimal than direct individual exposure monitoring or some modeling approaches, this may be the most practical approach in many post-disaster research scenarios. However, there are scenarios where this method would not be the best choice, such as an acute, short-term highly toxic chemical release where the need for long-term monitoring may not exist.

From late September 2001 onward, fine particle mass levels in the WTC area were generally within the limits of the National Ambient Air Quality Standard in effect at that time, when averaged over a 24-h period (Thurston and Chen, 2002). However, from 26 September until the fires burned out in late December, smoke plumes that emitted fine particles and gases were noticeable on some days, and debris removal continued to resuspend particles at Ground Zero (Lioy et al., 2006). Plume impacts in October 2001 delivered a different type of aerosol exposure and those exposed ‘could, in a few hours, be subject to materials in amounts and composition they would not have to endure in years of typical ambient conditions’ (Cahill et al., 2004). Thus, although we cannot adequately account for exposures during the earliest time periods, when the highest levels of exposure would have occurred, we believe the method described does capture important exposures that were an issue from late September 2001 onward. This is pertinent for our study cohort because NYS response personnel generally arrived on-site somewhat later than first responders (Mauer et al., 2007).

There are three basic components of the algorithm: time period/location weights, PPE usage and duration on-site. One factor that appears to strongly influence the exposure score is the duration on-site. The mean total time spent at the site steadily and sharply increases by exposure quartile. However, using this method, even a person at the site for months could have a low exposure score if they consistently used highly protective respiratory equipment. Conversely, someone who was at the site for a short period of time with no respiratory protection could be in the highest quartile. Within a given time period, location on the pile carries the most weight for calculating a score.

While the magnitude of the attenuation factor for SCBA respirators, and to a lesser extent PAPR respirators, suggests that these may have a very large influence on the exposure score, this is not reflected in this population because of the extremely low proportion of use. Although one-strap dust masks were used most frequently overall, they do not provide adequate protection (Centers for Disease Control and Prevention, 2002) and thus they do not attenuate the exposure score.

The compounds chosen as surrogates for exposure are intended to represent the complex mixture of exposures contained in dust (calcium) and smoke (total hepta-CDDs). Considering the complexity of exposures, we acknowledge that this may oversimplify the nature of the exposures. WTC dust certainly contained settled combustion products (Lioy et al., 2002) and plume aerosols contained dust particles such as concrete and gypsum (Cahill et al., 2004). We believe that, given the nature of the available data, these were the most representative compounds available to help distinguish between building material-related dust and combustion products in smoke.

Except for the pile, each exposure location included two or more monitoring stations. The use of multiple monitoring stations, particularly in the secure area and outside the secure area, could help account for varying conditions within a given exposure location. This might have helped account for factors such as demolition activities, materials transportation and meteorological conditions. However, it is acknowledged that a limitation of this exposure assessment method is that it does not directly account for such conditions.

The switch from rescue to recovery efforts occurred in mid-October, which falls in the middle of time period 5. Heavy equipment may have disturbed the pile more than earlier hand digging, thus increasing the amount of dust in the air. This may particularly affect the dust scores, and is likely the cause of the peak in values for time period 5, on the pile (Table 2). This may create an information bias if these activities were only limited to the area very near the monitor or only on the monitoring days. We do not believe this is the case because these activities were likely to occur at many locations on the pile and for many days given the nature of this disaster. This peak is not noted in any other locations for time period 5. This suggests that the other exposure locations, overall, were not affected significantly by this activity.

Using a self-administered questionnaire to assess work patterns is also a limitation. These data were collected on average 11 months (range 8–26 months) after the event, potentially contributing to recall bias. To improve reporting of work history at the WTC site, we divided the study into seven time periods corresponding to calendar dates with major divisions by weeks or months after the first 2 days. We also provided specific choices for locations, type of PPE and frequency of PPE use. Furthermore, exposure information was asked first in the questionnaire.

Information on the use of PPE was collected by time period rather than specific day, location or task. This is a limitation, especially in later time periods that are longer, where more than one task and location were more likely to be indicated. Some bias could be introduced because individuals may have used protective equipment only while they were in certain locations. In addition, we did not have any information on respirator fit testing. It is likely that at least some individuals using respirators were not fit tested and therefore were not adequately protected. This would result in an overestimate of their respiratory protection and an underestimate of their exposure.

Air monitoring data coinciding with our exposure locations were unavailable for the first three time periods in the algorithm because air monitoring did not begin in all study locations until 23 September 2001. Therefore, we equated the earliest time periods with time period 4, which was the first time period for which adequate data were available. This likely resulted in a significant underestimate of exposures in the early time periods, particularly in the period immediately after the collapse of the towers. There are some participants who would be impacted by this underestimate because they were present on-site either part of the time or exclusively during the early periods. Many participants would not be affected by this because they were present on-site only during the later time periods. Overall, this cohort arrived somewhat later than first responders such as firefighters and police. A study of NYS WTC responders (Mauer et al., 2007) reported that only one-third were on-site during the first 2 days and only 57% had arrived by 16 September. Work assignments actually peaked between 17 and 30 September (time period 4).

Rain that occurred on 14 and 25 September washed much of the resuspendable material from outdoor surfaces and caused the fires to weaken (Lioy et al., 2006). Because we equated the first three time periods with time period 4 (17–30 September), the first occurrence of rain would not impact our results. However, the rain event on 25 September occurred during time period 4 and may have resulted in lower median time period/location weights that might not accurately reflect higher exposure levels present before that date. This could have resulted in further underestimates of exposures for time periods 1–4.

The algorithm does not take into account job tasks. Although this information was obtained by time period for each individual, it was not associated with a specific number of days or hours, and more than one job task could be selected for each time period. The majority of these responders participated in traffic control, security and supervisory activities in all time periods. Because we do not have personal monitoring or task-specific quantitative data, we chose to exclude this information from the algorithm. This could result in an underestimate of exposures for certain individuals. However, we believe that by developing time period/location-specific weights, some of the variation in exposure will be captured. This is because certain tasks are inherently associated with certain locations. For example, hand digging is most strongly associated with the pile, for which the time period/location weights were higher than other areas.

CONCLUSIONS

We developed a semiquantitative exposure assessment method that provides an exposure score for NYS personnel who worked at the WTC disaster site on or after 11 September 2001. This method allows categorization of NYSDOH study participants by relative exposure levels, including individuals who had very different patterns of exposure, such as high-level, short-term exposures and low-level, long-term exposures, for epidemiological analyses. The two instantiations of the algorithm represent different aspects of overall WTC exposure: dust and smoke. The algorithm accounts for several factors affecting exposure, including time period and duration on-site, location of work and PPE used. It also utilizes quantitative data, specifically air monitoring data from pertinent locations. Although there are limitations to this exposure assessment method, the overall concepts employed could be valuable in planning for future post-disaster research studies with a long-term chronic exposure component when funds are limited and/or access to appropriate modeling data is unavailable.

FUNDING

Centers for Disease Control and Prevention (U1Q/CCU221159-04).

The authors would like to thank Randi Walker and Robert Chinery for their valuable assistance in identifying, reviewing and interpreting available environmental monitoring data sets. They also gratefully acknowledge the assistance provided by Karen Cummings and Nick Pavelchak. The contents are solely the responsibility of the authors and do not necessarily represent the official views of Centers for Disease Control and Prevention.

References

Cahill
TA
Cliff
SS
Perry
KD
, et al.  . 
Analysis of aerosols from the World Trade Center collapse site, New York, October 2 to October 30, 2001
Aerosol Sci Technol
 , 
2004
, vol. 
38
 (pg. 
165
-
83
)
Centers for Disease Control and Prevention. National Institute for Occupational Safety and Health. University of Kentucky
National Agricultural Safety Database
 , 
2002
 
Department of Health and Human Services (NIOSH)
NIOSH Guide to Industrial Respiratory Protection. Appendix E—Respirator Decision Logic Pt. 1
 , 
1987
 
DHHS (NIOSH) Publication No. 87-116. Cincinnati, Ohio: National Institute of Occupational Safety and Health. pp. 13–8
Edelman
P
Osterloh
J
Pirkle
J
, et al.  . 
Biomonitoring of chemical exposure among New York City firefighters responding to the World Trade Center fire and collapse
Environ Health Perspect
 , 
2003
, vol. 
111
 (pg. 
1906
-
11
)
Feldman
DM
Baron
SL
Bernard
BP
, et al.  . 
Symptoms, respirator use, and pulmonary function changes among New York City firefighters responding to the World Trade Center disaster
Chest
 , 
2004
, vol. 
125
 (pg. 
1256
-
64
)
Gilliam
RC
Huber
AH
Raman
S
Metropolitan-scale transport and dispersion from the New York World Trade Center following September 11, 2001. Part II: an application of the CALPUFF plume model
Pure Appl Geophys
 , 
2005
, vol. 
162
 (pg. 
2005
-
28
)
Lioy
PJ
Weisel
CP
Millette
JR
, et al.  . 
Characterization of the dust/smoke aerosol that settled east of the World Trade Center (WTC) in lower Manhattan after the collapse of the WTC 11 September 2001
Environ Health Perspect
 , 
2002
, vol. 
110
 (pg. 
703
-
14
)
Lioy
PJ
Pellizzari
E
Prezant
D
The World Trade Center aftermath and its effect on health: understanding and learning through human-exposure science
Environ Sci Technol
 , 
2006
, vol. 
40
 (pg. 
6876
-
85
)
Mauer
MP
Cummings
KR
Carlson
GA
Health effects in New York State personnel who responded to the World Trade Center disaster
J Occup Environ Med
 , 
2007
, vol. 
49
 (pg. 
1197
-
205
)
McGee
JK
Chen
LC
Cohen
MD
, et al.  . 
Chemical analysis of World Trade Center fine particulate matter for use in toxicologic assessment
Environ Health Perspect
 , 
2003
, vol. 
111
 (pg. 
972
-
80
)
Prezant
DJ
Weiden
M
Banauch
GI
, et al.  . 
Cough and bronchial responsiveness in firefighters at the World Trade Center site
N Engl J Med
 , 
2002
, vol. 
347
 (pg. 
806
-
15
)
Salzman
SH
Moosavy
FM
Miskoff
JA
, et al.  . 
Early respiratory abnormalities in emergency services police officers at the World Trade Center site
J Occup Environ Med
 , 
2004
, vol. 
46
 (pg. 
113
-
22
)
Skloot
G
Goldman
M
Fischler
D
, et al.  . 
Respiratory symptoms and physiologic assessment of ironworkers at the World Trade Center disaster site
Chest
 , 
2004
, vol. 
125
 (pg. 
1248
-
55
)
Thurston
GD
Chen
LC
Risk communication in the aftermath of the World Trade Center disaster
Am J Ind Med
 , 
2002
, vol. 
42
 (pg. 
543
-
4
)
United States Environmental Protection Agency
WTC Response: Environmental Monitoring Data Version 2.0
 , 
2003
 
Wolff
MS
Teitelbaum
SL
Lioy
PJ
, et al.  . 
Exposures among pregnant women near the World Trade Center site on 11 September 2001
Environ Health Perspect
 , 
2005
, vol. 
113
 (pg. 
739
-
48
)

Author notes

The free full text of this article can be found in the online version of this issue.