A new carbon monoxide occupational dosimeter: results from ... - Nature

Journal of Exposure Analysis and Environmental Epidemiology (1999) 9, 546 ± 559 # 1999 Stockton Press All rights reserved 1053-4245/99/$12.00

http://www.stockton-press.co.uk

A new carbon monoxide occupational dosimeter: results from a worker exposure assessment survey MICHAEL G. APTE,a,b DANIEL D. COX,c S. KATHARINE HAMMONDb AND LARA A. GUNDELa a

Indoor Environment Department, Lawrence Berkeley Laboratory, Berkeley, California University of California, School of Public Health, Berkeley, California 94720 c Global Risk Consultants, Oakland, California b

The LBNL / QGI occupational carbon monoxide ( CO ) dosimeter ( LOCD ) , a new, inexpensive CO passive sampler, was field - validated in an occupational exposure assessment study in the Moscone Convention Center ( MCC ) in San Francisco, CA in January, 1997. The LOCD measures time - weighed - average ( TWA ) CO exposures from 10 to 800 parts per million hours ( ppm h; accuracy 20%; precision 10 ppm h ) . This device represents a major improvement over currently available low - cost personal CO monitors. At the MCC, over 1000 workers set up and remove exhibitions. Forty propane - powered forklifts moved materials throughout the 42 000 m2 of exhibit halls. Diesel truck emissions enter the building via three internal underground loading docks. The LOCD was used to measure 154 worker exposures on 3 days. Sampler performance was compared to a standard method at 15 fixed sites. The geometric mean ( GM ) of all 154 exposures was 7 ppm ( geometric standard deviation ( GSD ) = 1.6 ) ; 10% of the exposures was 10 ppm or more. Dock Walkers and Forklift Operators had the highest exposures ( maximum = 34 ppm ) with GM ( GSD ) of 9 ( 1.7 ) and 9 ( 1.6 ) ppm, respectively. Attendants and Installer / Decorators had the lowest exposures with GMs of 6 ( 1.6 ) and 7 ( 1.4 ) , respectively. The Cal / OSHA personal exposure limit for CO is 25 ppm time - weighted average ( TWA ) . Keywords: exposure assessment, industrial hygiene, log - normal distribution, passive sampler, propane forklift, time - weighted average.

Introduction Carbon monoxide ( CO ) exposure is common in all societies. CO is a product of incomplete combustion that is present wherever hydrocarbon fuels are used. The United States CO morbidity and mortality data indicate that acute CO poisoning is a serious public health problem. In 1995, more than 19 000 CO poisoning incidents were reported by the American Association of Poison Control Centers, making it the number one cause of unintentional, non pharmaceutical poisoning in the U.S. (Litovitz et al., 1996 ). Although there is considerable public concern about CO safety, very little is known about the actual extent and distribution of CO exposures in the U.S. ( USEPA, 1991 ).

1. Abbreviations: A, absorbance units; AM, arithmetic mean; ASD, arithmetic standard deviation; Cal / OSHA, California Occupational Safety and Health Administration; DHS, Department of Health Services; GM, geometric mean; GSD, geometric standard deviation; HVAC, heating, ventilation, and air conditioning; IAQ, indoor air quality; LBNL, Lawrence Berkeley National Laboratory; LOCD, LBNL / QGI occupational dosimeter; MCC, Moscone Convention Center; PEL, personal exposure limit; ppm, parts per million; QGI, Quantum Group, Inc.; TWA, time - weighted average. 2. Address all correspondence to: Dr. Michael Apte, MS 90 - 3058, LBNL, 1 Cyclotron Road, Berkeley, CA 94720. Tel.: (510) 486 - 4669. Fax: (510) 486 - 6658. E-mail: [email protected] Received 21 August 1998; accepted 8 February 1999.

One currently available measure of CO toxic exposure is through analyses of death certificates collected by the U.S. Centers for Disease Control ( Cobb and Etzel, 1991 ). These data indicate that the current lifetime risk of unintentional fatal CO poisoning is about 10 4. This risk level is a factor 100 times greater than the U.S. risk level that the Environmental Protection Agency uses to regulate toxic substances such as benzene. This is based on 500± 1000 unintentional CO poisoning fatalities reported in death certificates annually (USDHHS, 1986 ). Although this risk level is high, it is an underestimate because the effects of CO poisoning are easily overlooked or misdiagnosed. Given the magnitude of the public health problem, it is surprising that we know so little about the distribution of CO exposures within our population. One explanation for this knowledge gap is that our CO monitoring technology has been deficient. The costs of using real -time instrumentation and the accuracy of the commercially available diffusion tubes have limited the opportunities for population -based exposure assessment for CO. Lawrence Berkeley National Laboratory ( LBNL ), in collaboration with the Quantum Group, Inc. (QGI, San Diego, CA ), has developed new passive samplers which can be used to characterize total CO exposure and /or indoor environmental CO concentrations (Apte, 1997 ). The LBNL /QGI occupational CO dosimeter (LOCD ) is designed to collect and measure time -weighted average (TWA ) CO exposures

Carbon monoxide dosimeter

over 8- h workshifts. A related device, the LBNL /QGI indoor air quality CO passive sampler, was developed to measure 1- week TWA CO concentrations. These devices have the potential to ``fill in'' the gaps in our knowledge of the population distribution of CO exposures in the U.S. and other countries. The objective of this paper is to present field validation data for the LOCD and the results of the survey where these data were collected. A comparison of performance of the LOCD against that of the DraÈger CO diffusion tube is also presented in order to place the results in the context of this frequently used exposure assessment technology. The study was conducted at the Moscone Convention Center ( MCC ) in San Francisco, CA during the setup of the a computer trade show. This study environment was selected to test the LOCD because the large number of propane - powered forklifts used at the MCC assured that CO exposures would be elevated and broadly distributed. Issues regarding CO contaminated environments from, and CO poisonings due to the use of forklifts in indoor spaces have been discussed elsewhere (USEPA, 1991; Fawcett et al., 1992a,b; Fleming and Opheim, 1992; Ely et al., 1995; McCammon et al., 1996 ). The study was conducted on January 3, 5, and 6, 1997 in collaboration with Crawford Risk Control Services ( Crawford, Oakland, CA ), an industrial hygiene firm contracted by MCC management. Through this collaboration, measurements using the LOCD were collected in parallel with those collected by Crawford. In this survey, the LOCD was used in three modes: (1 ) to measure TWA CO exposures on workers who were also monitored by Crawford using conventional CO passive samplers; (2 ) to sample in parallel with real - time CO monitors that Crawford placed on a number of workers; and ( 3) to measure time -averaged fixed - site CO concentrations in parallel with air samples collected in gas -tight bags by LBNL. Ventilation of the MCC poses some unique challenges. The main exhibition areas of the center are underground, with three major loading docks internal to the structure. Considerable concern has been expressed regarding the quality of air, including CO levels, in the MCC during periods when convention exhibits are setup and removed ( Blackwell, 1997; Katz and Osorio, 1997 ) . During peak work periods, some 40 propane -powered forklifts are operated nearly continuously throughout the building. Additionally, large numbers of diesel trucks, driven through an interior tunnel system, pull up to the interior docks to move materials in and out of the building. A small number of gasoline - and propane -powered utility lifts are operated intermittently during the decoration of the convention halls. Some measures to improve the indoor air quality of the building were already in place prior to the study reported here (Katz and Osorio, 1997 ). These improvements Journal of Exposure Analysis and Environmental Epidemiology (1999) 9(6)

Apte et al.

included fitting of catalytic converters to the forklifts, modifications in the building ventilation system, and gaining cooperation from the truck drivers in minimizing unnecessary engine operation. The implementation of these changes suggests that the Moscone Center was committed to the ongoing process of worker exposure reduction. Since the study presented here was conducted, additional exposure control measures have been implemented and exposures have been reduced further. The current Federal OSHA permissible exposure limit (PEL ) for CO is set as a TWA of 50 ppm for an 8- h shift (OSHA, 1993 ). The Cal /OSHA PEL for CO is 25 ppm TWA over an 8- h workshift ( CAL /OSHA, 1997 ) . Methods Measurement Methods The LOCD Figure 1 is an exploded diagram depicting the LOCD. The LOCD was designed as a small (1.3 cm1.3 cm4.5 cm ) , lightweight, intrinsically safe unit which could be clipped unobtrusively to the lapel of the worker. The principle of operation and laboratory testing of the device is fully discussed elsewhere (Apte, 1997 ). In brief, the device consists of a small palladium ± molybdenum based CO sensor designed into a compact diffusion tube sampler contained in a disposable spectrophotometric cuvette. The TWA CO exposure is calculated from the difference between pre - and post -exposure 700 nm spectrophotometric measurements of sensor absorbance. Silica gel contained in the device is used to desiccate the sample. The LOCD was found to operate in the exposure range of 10± 800 parts per million hours (ppm h ) with an accuracy of 20% and a precision of 10 ppm h, and to be insensitive to humidity up to 90% relative humidity. A slight temperature dependency requires that a separate calibration factor be used when the device is operated at temperatures below 158C. Post -exposure sensor regeneration ( reversal of sensor chemistry ) was found to be negligible for 24 h, but significant over 1 week ( Apte, 1997 ) . A set of 15 potential chemical interferents including different organic and inorganic molecular species were tested, representing a fairly large range of the types of gas phase pollutants commonly found in occupational and residential environments (Apte, 1997 ). These tests were demanding in terms of the high interferent concentrations. With two exceptions, the dosimeters did not show a significant or practical effect from the exposures. The two exceptions were 100 ppm h nitric oxide (a strong negative interference from exposure to NO but not CO, but no observed effect with concurrent CO +NO exposure ), and 200 ppm h ethylene (having a strong positive bias with and without CO ). 547

Apte et al.


Figure 1. Diagram of the LOCD and an expanded view of its internal components.

For this study, 85 LOCDs were assembled with lapel clips. Each LOCD was labeled with a unique identification number and placed in the spectrophotometer for replicate 700 nm absorbance measurements. Prior to these measurements, the spectrophotometer had been adjusted to read 0.000 absorbance units (A ) with an LOCD containing no sensor. The average of the initial absorbance measurements of the LOCDs was about 1.5A. The LOCD sensor responds to CO exposure with a linear change in absorbance ( 1 ppm h CO = 331A ). During the 3 days of the study, the LOCDs were reused each day. They could be reused reliably until the capacity of the sensor has been reached ( an absolute absorbance of about 2.5 ±3.0A ), or until the silica gel desiccant in the devices has been depleted (blue indicator in gel turned clear ). Five LOCDs were used as field blanks (unexposed ) each day. Of the remaining 80 LOCDs, about 55 were exposed as personal monitors 548

clipped to the lapels ( breathing zone ) of the study participants. In addition, 10 were paired with real -time dosimeters worn on the waist on a subset of the participants and 15 were attached, in sets of three each, to five bag samplers for fixed - site measurements within the MCC. At the time of deployment, the LOCDs were uncapped and the sample identification numbers and deployment times were recorded in conjunction with the anonymous participant information. At the end of the workshift sampling period ( approximately 8 h) , they were retrieved from the participants, capped, and sampling finish times were recorded. LOCD Analysis The final absorbance readings of all of the dosimeters were measured at LBNL within 3 h of the end of measurement. CO exposures were calculated by dividing the measured change in absorbance (dA ) by Journal of Exposure Analysis and Environmental Epidemiology (1999) 9(6)


Apte et al.

the empirically derived response slope with temperature adjustment (see Apte, 1997 ).

a DraÈger tube and an LOCD were placed on the lapel of each participant.

The Bag Sampler The bag sampler is a simple device used to collect a sample of CO -laden air into an inert gas sample bag over a period of time. The sampler draws at a constant rate so that the concentration of CO in the bag at any time is the average of the sampled bulk -air concentration over that time. Since CO is a non- reactive gas, the sample is not subject to wall loss due to surface reactions. The bag samplers used in this study were designed and constructed at LBNL. They were outfitted with peristaltic pumps (Masterflex2, Cole Parmer, Niles, IL ) with a flow 3 rate setting of about 10 cm min 1. The bag samplers were built into a small plastic suitcase and were powered externally using 110 VAC. The internal cavity of the suitcase was large enough to hold an inflated 10- liter air sampling bag (Air Sampling Bag, Tedlar, SKC, Inc., Eighty Four, PA ) . The inlet tubing of the pump was connected via a screened bulkhead in the side of the sampler case. The outlet of the pump was attached to a valved fitting on the sampling bag. Tedlar air sampling bags were purged twice with dry pure air and evacuated in preparation for sampling. During this study, the bag samplers with prepared bags were placed at a selected fixed site within the MCC and power was provided via an extension cord. Sampling commenced when the bag valve was opened, allowing a constant flow of ambient air to enter the bag. Bag samplers were analyzed using a gas -filter correlation CO analyzer ( Thermo Environmental Model 48 ) within 24 h of collection. This analyzer was calibrated prior to each use and is documented to be accurate to 1%. The bag sampler and CO analyzer measurement combination was considered to be the ``Gold Standard'' for this study. Unfortunately, two of the 15 bag samples collected leaked or failed to fill, causing a loss of data.

Real -time Datalogging Personal Monitors The STX70 datalogging CO monitor (Industrial Scientific Corp., Oakdale, PA ) was used by Crawford for measuring personal exposures in real -time. This device uses an electrochemical sensor which must be calibrated daily to maintain optimum performance. The internal datalogger in the instrument collects and stores a CO exposure profile at sampling rates as frequent as 1 Hz. Its lower limit of detection is 1 ppm, and the monitoring range is 0 to 999 ppm. These monitors were calibrated each day prior to use with CO - free air and 100 ppm CO, and they were set to record CO concentrations every minute. A LOCD was taped to each STX70 unit before each shift. The paired devices, weighing about 200 g, were attached to the belts of the workers. Each of the 10 participants who wore a real - time datalogger also wore a DraÈger diffusion tube and an LOCD in the breathing zone. At the completion of each sampling session, the collected data were uploaded to a computer for analysis. TWA CO exposures were calculated from the data for the sampling period.

DraÈger Diffusion Tubes The DraÈger diffusion tube ( DraÈgerwerk, LuÈbeck, Germany ) is a common device for measurement of workplace CO exposure. It contains an adsorbent that darkens upon exposure to CO, and the tube has a graduated scale printed on it representing CO exposure in parts per million hours, with a minimum graduation of 50 ppm h. Breaking the glass seal at the inlet end of the tube deploys it. The device is not recommended for exposures times beyond 8 h. CO exposure is determined by comparing the length of a stain that develops within the tube against the graduated scale. Workers typically wear the tubes throughout the work shift. These devices are easy to use and have been shown to have a linear response in the laboratory. However, they have been found to underestimate CO exposure and have statistically significant humidity effects (Hossain and Saltzman, 1989; Valerio et al., 1997 ). In this study, both Journal of Exposure Analysis and Environmental Epidemiology (1999) 9(6)

Industrial Hygiene Study Design MCC Physical Characteristics The MCC occupies two city blocks in downtown San Francisco, California. The main exhibition hall spaces of the North and South sections are one floor below ground and are interconnected ( Figure 2) . Overall, the MCC has a 41 000 m2 (442 000 ft 2 ) of exhibit space and 15 000 m2 ( 160 000 ft 2 ) of meeting space. The main underground exhibition hall spaces of North and South Halls are approximately 16 900 m2 ( 181 000 ft 2 ) and 24 200 m2 (261 000 ft 2 ), respectively. A tunnel system leading to the subterranean hall level provides truck access to three loading docks. The Red Dock, serving the North Hall, contains slots for about ten single trailer rigs to serve as back up. The Blue and Green Docks service the South Hall, with slots for about five single rigs at each dock. The North and South Halls have separate ventilation systems. The truckway tunnel and Red Dock (North ) have an air exhaust capacity of 1800 m3 min 1 ( 62 000 cfm ) and an air supply capacity of 1300 m3 min 1 (45 000 cfm ). The Green and Blue docks (South ) each has an air exhaust capacity of 900 m3 min 1 ( 32 000 cfm ) and an air supply capacity of 600 ( 22 000 cfm ) . The heating, ventilation, and air conditioning (HVAC ) system of the North Hall provides up to 6000 m3 min 1 (206 000 cfm ) of airflow to the building, using 100% outside air under normal operating conditions. The HVAC system of South Hall can provide up to 18 000 m3 min 1 (607 000 cfm ) of outside air. The MCC engineering department has stated that 100% outside 549

Apte et al.


Figure 2. Fixed - site locations for bag sampler / LOCD monitoring. Numbers indicate the bag sampler number.

air is used in South hall during exhibition move - in /move out periods ( Katz and Osorio, 1997) . MCC Employer, Union, and Worker Classifications The study was designed to assess the workers' CO exposure as a function of job classification and union membership. Table 1 lists the job categories included in the study, the unions representing the workers, and the number of participants from each category. An initial interview of participants was conducted when they were recruited. Information, elicited via a survey questionnaire at the completion of the personal sampling period, was used to correlate the CO measurements (identified by CO sampler number) with the workers' job classifications and assignments, employer, and union affiliation; and the current day's location of work within the MCC. After the study was completed, this information was compared with the data from the survey 550

questionnaire for quality control. A total of 154 personal samples were collected during the study. The total number of individual participants is unknown since, for confidentiality, name or other identifiers were not used to track the participants. Individuals may have been monitored on one, two or three of the study days. MCC Workforce The Attendants (Table 1) at the MCC were employed by Spectator Management Group ( SMG ) to assist in the upkeep of the facility. They performed such services as security, housekeeping, trash removal, and trash compaction. Depending upon their assigned tasks, they were either stationed at a single location or moved throughout the center during their workshifts. Although many of the workers were located on the exposition hall floors and loading docks in close proximity to forklifts and trucks potentially emitting CO, others were located in areas Journal of Exposure Analysis and Environmental Epidemiology (1999) 9(6)


Apte et al.

Table 1. Job categories for which workers were monitored for CO during 1 / 3, 1 / 5, and 1 / 6 / 1997 at the MCC. Union

Employer

Number of Workshifts Monitoreda

Installer / Decorator

SDAC 510

FDC

49

Walker

SDAC 510

FDC

5

Rigger

SDAC 510

FDC

4

Shop Steward

SDAC 510

FDC

3

Dumpmeister

SDAC 510

FDC

2

Job title Decorators

Teamsters Dock Foreman

Teamsters 85

STC

12

Forklift Operator

Teamsters 85

STC

17

General Foreman

Teamsters 85

STC

7

Handyman Truck Driver

Teamsters 85 Teamsters 85

STC STC

6 1

Attendant / Security / Management Attendant Supervisor

SEI 14 Management

SMG FDC, STC

Desk Worker

Management

SMG

33 11 4

a

Some participants were monitored on 1, 2 or 3 days so that the actual number of participants was less than the number of workshifts monitored. Note: FDC = Freeman Decorating Co.; STC = Sullivan Transfer Co.; SMG = Spectator Management Group; SDAC 510 = Sign Display and Allied Crafts Local 510.

such as the mezzanine, upper floor meeting rooms, and rest rooms which were distant from direct CO sources. A number of individuals, classified as Desk Workers, were employed by SMG in various desk jobs. These workers were located in various offices located throughout the MCC facility. Some of the Desk Workers were located in offices that were in close proximity to a loading dock. Supervisors were employed by Freeman Decorating Company (FDC ) and Sullivan Transfer Co. (STC ) management to oversee work at the docks and on the exhibition hall floor. The Dock Supervisors' position interacted with forklift and truck operators and were usually in close proximity to engine exhaust. The other supervisors were often on the exhibition hall floor where considerable forklift traffic was present throughout the workshift. The Dock Foreman was a union position parallel to the Dock Supervisor, and General Foreman and Shop Steward were union positions which paralleled their respective supervisor categories in terms of location and interaction with forklift and truck operation. The Installer / Decorator, Handyman, Dumpmeister and Rigger job categories primarily involved work on the MCC exhibition hall floors. Their jobs involved construction of the exhibitions and decorations within the hall. These participants were intermittently in close proximity to forklifts as they moved around the exhibition halls. The Riggers operated gasoline -powered lifts that were used to reach attachment points on the high ceilings and walls of the building. These lifts were potentially an additional CO Journal of Exposure Analysis and Environmental Epidemiology (1999) 9(6)

source within the building. The Dumpmeister was in charge of coordinating the removal of trash from the exhibition sites as they were constructed. The Teamsters performed three jobs. The Forklift Operator operated the Forklifts, while the Walker worked on foot next to the forklifts. Obviously, these workers were in constant proximity to the emissions from the propane powered forklifts. Forklift Drivers and Walkers were observed to work together with the forklift running within the enclosed area of the long diesel truck trailer. The Truck Driver position involved operation of a diesel truck for the MCC, spending considerable time at the loading docks. Participant Selection Participants for personal sampling were solicited from the worker population by union. The goal was selection of 20 Decorators, 15 Teamsters, and 15 Attendants /Security /Management each day. Participants from each category were taken on a first - come, first served basis until the number for each category was reached. The participants were notified of the survey by union and management representatives and offered the opportunity to volunteer. Due to the workers' heightened concern for the MCC air quality, many individuals came forward as volunteers. No strict quota was set for participants from any particular employer, union, or job category. However, due to the nature of the Teamsters' work shifts, it was necessary to seek out these workers as they signed on for the day, and ask them if they would volunteer to participate in the study. Nearly all of those 551

Apte et al.

who volunteered to be a participant were included in the study. The protocol employed in the study included collection of a signed information sheet and consent form prior to enlistment into the survey, and a survey questionnaire to be completed at the end of the workshift. Although the workers' names were collected on the consent form, no personal identification of the participants was made on the survey questionnaire.

Results The level of interest and cooperation from the MCC management, the unions, and the participants was high throughout the study. The workers showed considerable attention to their work environment and concern over how it might be affecting their health. Additionally, the MCC management showed interest in ensuring that the workers were protected from emissions from CO sources. All of the MCC ventilation systems were clearly operating at a high rate during the study, based on the palpable movement of air in many parts of the building. The actual duration of the monitored workshift varied from worker to worker because they did not all come on or leave their shifts at the same time. After approximately 8 h, the monitoring equipment was retrieved from those participants who were still at work. The duration of the sampling period for the LOCDs and DraÈger diffusion tubes averaged 7.8 1 h. One pair of samplers was not retrieved until 10 h had elapsed. Area Measurements: Validation of the LOCD with the Bag Sampler Figure 2 is a map of the underground exhibition hall level of the MCC. It shows the sampling locations for the bag sampler /LOCD data for all 3 days. The temperature in the MCC was colder than 208C ( 688F ) during the study. Detailed temperature measurements were not recorded during the study. Several spot temperature measurements, taken at the industrial hygiene operations desk (Figure 2 ), ranged from 17.88C to 18.98C ( 648F to 668F ). The operations desk was located in an enclosed internal hallway and had less ventilation air supplied, and was observed to be somewhat warmer than the loading docks and interior of the building. The average outside air temperatures for San Francisco, CA on January 3, 5, and 6 were 138C, 98C, and 118C, respectively ( NOAA, 1997). During the study, the ventilation systems were set to supply 100% outside air and were operating at a noticeably high rate, providing a continuous supply of cold air. It is estimated that the indoor temperature in the MCC was between 108C and 198C (408F to 66.28F ) during the study. 552


The bag samplers vs. LOCD data are plotted in Figure 3. The corresponding fixed -site sample locations are shown in Figure 2. The average LOCD data clearly fitted the bag sample data well with essentially no bias: the slope of the fitted regression line was 1.01 (95% confidence interval: 0.92 to 1.1 ). The response was also very linear within the range of measurements ( R 2 = 0.95 ). The level of agreement between the bag samples, analyzed using a CO analyzer, and the average of three LOCDs was within 2 ppm and all but one were within 1 ppm. The highest fixed - site CO measurements were at the docks, the highest of which were observed on the Green Dock on January 3 and 6. The Green Dock bag sampler data ( LOCD data ) were 11 ( 11 ppm ) and 13 (15 ppm ) on these days, respectively. The lowest CO averages were observed in the North and South Halls. The set of three LOCDs attached to bag sample number 5, located at the north end of South Hall, averaged 7 2 ppm ( no bag sampler data are available at this location as the bag sampler failed ). The average absolute difference between all individual LOCD measurements and the bag samplers was 1.2 1.0 ppm. The average absolute difference between the average of three LOCDs measurements and their corresponding bag samplers was 0.7 0.4 ppm. Finally, the average of all 13 pairs of the fixed - site LOCDs and bag samples was not statistically different ( p < 0.01) , being 4.8 3.8 and 4.8 3.5 ppm, respectively. Personal Monitoring of Workers Using the LOCD The distribution of all personal CO exposures measured in this study using the LOCD can be seen in Figure 4. The long right - hand ``tail'' of high CO concentrations is characteristic of a log -normal distribution. The data were verified to fit a log -normal model using a graphical method (Becker et al., 1988 ). The log -transformed data were quite linear ( R 2 = 0.98) , indicating that the data are approximately log -normally distributed and that geometric statistics are probably appropriate for representing these data. Log - normality in the distribution of exposures to air pollutants in space and time arises from the multiplicative interaction of a series of random variables such as source, ventilation and worker mobility (Rappaport, 1991 ). The following discussion will focus on the geometric statistics. However, the arithmetic statistics are also included in the tables. In most cases, only a small difference ( 1 ppm ) was found between the geometric mean and arithmetic mean. Statistics presented from the monitoring of worker exposures at the MCC using the LOCD include arithmetic mean (AM ) and standard deviation (ASD ) , geometric mean ( GM ) and standard deviation ( GSD ), maximum observed CO TWA, and number of workers monitored on each day. Journal of Exposure Analysis and Environmental Epidemiology (1999) 9(6)


Apte et al.

Figure 3. LOCD vs. Bag Sampler data collected at fixed - sites throughout the MCC during the CO exposure study. Air samples were collected in Tedlar Bags over an 8 - h period, and analyzed using a Gas Filter Correlation CO analyzer. Three LOCDs were deployed at the site of each bag sampler. The error bars represent 1 standard deviation about the mean LOCD value. Numbers indicate MCC sampling site. Data collected on: ( o ) 1 / 3 / 97, ( ~ ) 1 / 5 / 97 and ( 8 ) 1 / 6 / 97.

Personal Dosimetry by Date Table 2 presents a summary of the LOCD TWA personal monitoring data for 3 monitoring days in January, 1997. Over the 3 days of the study, 154

workshifts were monitored using the LOCD; 52, 51, and 51 individuals were monitored on January 3, 5, and 6, respectively.

Figure 4. CO exposure distribution at the MCC during trade - show setup. All job titles for all 3 days of the CO exposure study. Journal of Exposure Analysis and Environmental Epidemiology (1999) 9(6)

553

Apte et al.


Table 2. MCC CO exposure survey summary statistics by date of measurement, job category, and job location. These data were collected using the LOCD and reflect TWA 8 - h workshift concentrations. Job category and job location are sorted by GM in ascending order. Category

Number observed (N)

AMa ( ppm )

ASDa ( ppm )

GMa ( ppm )

GSDa

90th percentile ( ppm )

> 25 ppm (%)

Maximum ( ppm )

By date All 3 days ( all data )

154

8

4

7

1.6

11

1

34

1 / 3 / 97

52

9

5

8

1.6

15

2

34

1 / 5 / 97

51

7

2

7

1.4

10

0

15

1 / 6 / 97

51

8

4

7

1.6

11

0

21

Desk Worker

4

6

3

5

1.8

9

0

10

General Foreman

7

7

3

6

1.5

10

0

11

33

7

3

6

1.6

11

0

17

By job

Attendant Shop Steward Installer / Decorator Supervisor Rigger

3

7

1

7

1.1

7

0

8

49 11

7 8

3 2

7 8

1.4 1.3

10 10

0 0

16 11

4

8

3

8

1.4

10

0

11

12

9

5

8

1.7

16

0

21

Handyman

6

10

6

8

1.8

19

0

20

Dumpmeister

2

9

0

9

1.0

8

0

9

17

10

7

9

1.6

17

6

34

Walker

5

10

6

Truck Driver

1

18

NA

All other indoor areas

13

7

North Hall

45

7

South Hall All over

24 32

Blue Dock Red Dock Green Dock

12

Dock Foreman

Forklift Operator

9

1.7

19

0

21

18

NA

NA

0

18

2

6

1.3

8

0

9

3

6

1.5

10

0

15

8 8

2 4

7 7

1.3 1.7

10 12

0 0

12 17

3

8

4

7

1.9

12

0

12

25

9

6

8

1.5

12

4

34

13

6

12

1.7

21

0

21

By location

a

AM = arithmetic mean, ASD = arithmetic standard deviation, GM = geometric mean, and GSD = geometric standard deviation.

The GM of all the workers' TWA CO exposures for the 3 days was 7 ppm (GSD = 1.6) . Recall that the 8- h TWA Cal /OSHA PEL= 25 ppm and the Federal -OSHA PEL=50 ppm. Roughly 10% of the participants had workshift average exposures above 10 ppm. One worker (1% ) had an exposure above the Cal / OSHA, and no worker was exposed above the 50 ppm Federal -OSHA PEL. The highest daily GM and highest individual exposure values both occurred on January 3, when the GM was 8 ppm ( 1.6) and the maximum TWA was 34 ppm ( Table 2). On January 3, the 90th percentile of exposures was 15 ppm, and one individual (2% ) was exposed above 25 ppm ( Table 2). Personal Dosimetry by Job Category Table 2 also presents the dosimetry data summarized by job category. The highest GM was observed in the group of Forklift Operators. This group of workers had a GM of average shift exposures of 9 554

ppm (1.6) , and the maximum 8 -h exposure was 34 ppm. The 90th percentile of Forklift Operator exposures was 17 ppm, and one of these workers ( 6% of Forklift Operators ) had a measured 8 -h exposure in excess of the 25 ppm 8 -h PEL. Dock Foreman and Walker and Handyman job categories had the second highest maximum observed 8 - h TWAs of 20 ±21 ppm ( 90th percentile was 19 ppm ). From Table 2, it is evident that the Dock Foreman, Walkers, and Handyman categories [ GMs of 8 (1.7 ) and 9 ppm (1.7 ), 8 ppm (1.8 ), respectively ] had similar exposures to those of the Forklift Operators [ GM =9 ppm ( 1.6) ] . The Dumpmeister and Supervisor categories also had similar exposure means but lower variability and maximum values. The combined data for Dock Foreman and Walker and Handyman jobs were not statistically different from the Forklift Operators ( Student's two -tailed t-test, p =0.89 ). Although only one workshift Journal of Exposure Analysis and Environmental Epidemiology (1999) 9(6)


measurement was made of the exposure of a Truck Driver, the TWA exposure for this worker was 18 ppm. Workers in the Installer/Decorator and General Foreman job categories were exposed to lower TWA CO concentrations, both with a GM of 7 ppm (1.4 ). The maximum 8 -h TWA for the Installer /Decorators was 16 ppm. The Attendants had a similar exposure distribution with a GM of 6 ppm ( 1.6) and a maximum TWA worker exposure of 17 ppm. Attendants' exposures were not significantly different from Installer/ Decorators' (Student's two -tailed t- test, p= 0.89 ). The job category with the lowest exposures was the Desk Workers with a GM of 5 ppm (1.8) , and a maximum of 10 ppm for one participant. Student's two -tailed t -tests comparing the combined Dock Foreman, Walker, Handyman, and Forklift Operator categories against the Installer/ Decorators showed that they were significantly different ( p =0.005 ). Personal Dosimetry by Location Table 2 also presents the exposure study by location. The majority of workers ( 71% ) worked predominantly in one of the following areas throughout their work shift: North Hall, South Hall, Blue Dock, Green Dock, or Red Dock. A number of participants ( 13% ) reported that they worked predominantly in an area other than those just listed, which included the Mezzanine and Esplanade areas of the South MCC structure. Additionally, 32 of the participants (21% ) reported that they worked in several of the locations listed above, or throughout the building during their work shifts. The Dock Workers appeared to have the highest workshift average exposures. The GM TWA at the Green Dock was 12 ppm ( 1.7 ), with a maximum 8 -h TWA exposure of 21 ppm. The highest single exposure occurred at the Red Dock where one worker was exposed to a TWA of 34 ppm. The CO distributions in the North and South Halls were not dissimilar with GMs of 6 (1.5 ) and 7 ppm ( 1.3) for North and South Halls, respectively. No workers' exposures in North Hall or South Hall exceeded 12.5 ppm. Participants working in the areas such as the Mezzanine and Esplanade appear to have the lowest exposures with a GM of 6 ppm ( 1.3) and a maximum TWA CO exposure of 9 ppm. Those individuals who reported that they worked ``all over'' had a GM of TWA exposures of 7 ppm and a maximum for a worker of 17 ppm. Real-time Personal Monitoring of Workers Table 3 presents a comparison of exposure measurements made using the real -time electrochemical monitors, with LOCD attached to a real -time monitor, and a LOCD attached to the lapel of the participant wearing the real -time monitor. With one exception, the real -time profiles are not included here. There are a number of disagreements between the real -time data and the LOCD measurements. Journal of Exposure Analysis and Environmental Epidemiology (1999) 9(6)

Apte et al.

The averages of all 29 pairs of usable real -time and LOCD TWA data were 3.4 and 6.3 ppm, respectively. An analysis of the data shows that the real -time TWA was greater than 2 ppm lower than the LOCD 70% of the time, and 5 or more ppm lower 28% of the time. Based upon visual inspection of the real - time data, these monitors' calibrations appear to have been subject to a calibration error, causing a number of the instruments to give falsely low averages. A number of the TWA values calculated from the data were low enough to be inconsistent with the bag sample, DraÈger tube and LOCD measurements in the MCC during the days of this study. The LOCD measurements are not likely to be in error to the extent of their disagreement with the real time monitors because the LOCD indicated consistent agreement with the bag sampler measurements in the MCC during the same sampling periods. As discussed above, the average LOCD measurements were within 1 ppm of the bag samplers on all but one comparison, in which the difference was 2 ppm. In contrast, the differences between the LOCD and the TWA real -time measurements were as great as 13 ppm. A comparison of the waist -level LOCD vs. the lapel LOCD measurements indicates that the TWA CO levels at the lapel ( breathing zone ) were several parts per million higher in most cases. For seven pairs, the lapel samples exceeded the waist samples by 5 ppm or more. The lapel measurements were 8.5 4.5 ppm vs. 6.3 3.5 at waist height. Although the reason for this is unknown, it is possible that it is caused by a combination of a vertical stratification of local CO levels caused by the thermal buoyancy of the hot exhaust gases, and by the close proximity of the forklift exhaust outlets to the Forklift Operators. This speculation may be strengthened by the observation that four of the five Forklift Operators' waist LOCD measurements were greater than or equal to the lapel measurement while only three of 24 of the other worker categories' measurements followed this pattern. It is possible that this is because the seated Forklift Operators were positioned above the exhaust outlet of their machines so that both the waist and the lapel were in the plume of undiluted exhaust. Whatever the cause of the waist / lapel height discrepancy, this observation indicates that it is important to measure worker CO exposure at breathing level or to adjust for the difference when interpreting data measured at other levels. Figure 5 depicts a real -time CO exposure profile of one of the participants in the study ( job category, Walker; location, Green Dock; date, 1/ 6/97; on Table 3 ). This monitoring session is likely representative of the types of exposure profiles expected from the MCC workers; however, the profile presented here had the highest observed 8 -h average CO concentration of those with real -time monitors: 16 ppm. The data in the figure have been smoothed by plotting 5- minute average concentrations. In 555

Apte et al.


Table 3. TWA CO exposures measured at the MCC for 30 volunteer participants wearing real - time CO monitors, LOCD attached to the real - time monitor, and an LOCD attached to the lapel of the worker wearing the real - time monitor. Worker Job Classification

Location

TWA CO Real - time ( ppm )

Attendant

All over

0

2

3

1 / 5 / 97

Attendant

North Hall

2

3

8

1 / 3 / 97

Attendant

North Hall

1

2

3

1 / 5 / 97

Attendant

North Hall

4

5

6

1 / 5 / 97

Attendant Attendant

North Hall North Hall

3 1

3 3

15 3

1 / 6 / 97 1 / 6 / 97

Attendant

Red Dock

6

17

1 / 3 / 97

Attendant

South Hall

2

7

8

1 / 6 / 97

Attendant

S. Mezzanine

1

3

3

1 / 3 / 97

Dock Foreman

Red Dock

1

5

8

1 / 5 / 97

Dock Supervisor

Green Dock

0

5

8

1 / 3 / 97

Dock Supervisor

Green Dock

6

10

15

1 / 5 / 97

Forklift Operator Forklift Operator


3 6

6 7

lost 6

1 / 3 / 97 1 / 3 / 97

Forklift Operator

North Hall

5

7

12

1 / 3 / 97

Forklift Operator

North Hall

11

9

4

1 / 5 / 97

Forklift Operator

North Hall

3

8

6

1 / 6 / 97

Forklift Operator

South Hall

2

10

10

1 / 5 / 97

General Foreman

North Hall

0

3

5

1 / 5 / 97

Handyman

Red Dock

1

6

5

1 / 6 / 97

Installer / Decorator Installer / Decorator


1 0

13 3

7 7

1 / 3 / 97 1 / 5 / 97


North Hall

0

5

5

1 / 6 / 97


South Hall

0

3

8

1 / 5 / 97


South Hall

8

9

10

1 / 6 / 97

Manager

All over

1

7

8

1 / 3 / 97

Rigger

North Hall

5

8

11

1 / 3 / 97

Rigger

North Hall

0

3

5

1 / 6 / 97

Walker Walker

Green Dock Green Dock

16 15

13 15

21 20

1 / 6 / 97 1 / 6 / 97

Failed

the case of this worker, a 5- minute average peak CO concentration of about 55 ppm was observed. The 1 -minute peak reached about 160 ppm. The LOCD that was attached to this real -time monitor measured a TWA of 13 ppm and the LOCD attached to the worker's lapel measured a TWA of 21 ppm. Comparison of Parallel LOCD and DraÈger Diffusion Tube Exposure Measurements Figure 6 presents a plot comparing 8 -h TWA LOCD vs. DraÈger exposure measurements conducted by Crawford for the 136 instances where both types of dosimeters were worn simultaneously by participants. The LOCD measurements were a subset of the 154 from which the data above are presented. Ten additional DraÈger tubes that had been paired with LOCDs were lost because they dropped from their 556

TWA CO LOCD with Real - time ( ppm )

TWA CO LOCD on Lapel ( ppm )

Date

lapel clips during the workshift. Unfortunately, one DraÈger tube sample that was lost was the mate to the highest LOCD measurement of 34 ppm worn by a Forklift Operator. Eight LOCDs were deployed on participants without paired DraÈger tubes. The DraÈger tube data in Figure 6 fall into discrete groups relating to the graduated scale printed on the tubes. The lowest graduation on the tubes is 50 ppm h, which corresponds to an 8 -h TWA of 6.3 ppm (i.e., 50 ppm h /8 h =6.3 ppm ). The discrete levels of DraÈger data below 6.3 ppm indicate attempts at visual interpolation between zero and 50 ppm h. The overall scatter of Figure 6 shows that the correlation between the DraÈger data and the LOCD data was quite poor. A regression line for all 136 data points had a slope of 0.63 (R 2 = 0.37) . A regression of the DraÈger data below 6.3 ppm Journal of Exposure Analysis and Environmental Epidemiology (1999) 9(6)


Apte et al.

Figure 5. Real - time CO exposure profile MCC during a trade - show setup. Crawford, Inc. data.

and their corresponding LOCD data had a slope of 0.61 ( R 2 = 0.44) . When only the paired data with DraÈger tube

TWA values of 10 ppm or more were considered, the slope was 0.80 (R 2 =0.58) . The average absolute value of the

Figure 6. Comparison between parallel 8 - h TWA LOCD and DraÈger diffusion tube CO measurements from the MCC CO exposure study. The dashed vertical line represents the DraÈger diffusion tube limit of detection of 6.3 ppm. The dotted diagonal line is a 458 line representing 1:1 corresponding between the two measurement methods. Journal of Exposure Analysis and Environmental Epidemiology (1999) 9(6)

557

Apte et al.

difference between the DraÈger and the LOCD was 3.1 2.5 ppm. This can be compared to the similar statistics for the LOCD comparison to the bag samplers presented above, which showed that the average difference between bag samples and individual LOCD measurements was 1.2 1.0 ppm. The above comparison indicates that attempts to interpolate below the 50 ppm h minimum graduation of the DraÈger tubes did not yield accurate information. The slope of the DraÈger ± LOCD relationship indicates that the DraÈgers' measurements were about 60% of the LOCD values. This improved slightly for the DraÈger measurements above 10 ppm where the DraÈger values were about 20% lower than the LOCD. The DraÈger tube underestimated all but one of the ten paired samples where CO concentration was greater than 15 ppm, and three of these DraÈger tubes underestimated the LOCD by 70%. The reason for these underestimates is not clear; however, other researchers noted that the DraÈger CO diffusion tubes underestimate, on average, by about 20% ( Valerio et al., 1997 ). The underestimate of concentrations well above the limit of detection is disconcerting and suggests that the DraÈger tubes may not be well-suited for quantitative CO exposure assessment. However, as a qualitative indicator of CO levels, the DraÈger tubes were adequate in this study since they did not misrepresent exposures relative to the regulatory standards.

Discussion Both fixed - site measurements and personal sampling indicate that the highest exposures occurred at the docks, particularly, the Green Dock. In general, the personal exposures appeared to be higher than the area measurements. This is not surprising since worker activities tend to include time spent in close proximity to the forklifts, whereas the fixed -site samplers were located safely out of the way of the forklift traffic. The CO concentrations observed at the fixed - sites may be considered the baseline to which workers in the monitored area were exposed. In contrast, personal exposures could include additional CO exposure due to more direct exposure to engine exhaust. CO sources are present in the MCC, and the potential exists for CO exposures above the PEL. One of the 155 measured TWA exposures exceeded the 25 ppm Cal /OSHA PEL (34 ppm ). There does not appear to be a large difference between the exposure distributions of the 3 days of the study, although those on January 3 were slightly higher. In contrast, there do appear to be large differences in exposure distributions between different job categories and different locations. The forklift and truck exhaust are the source of the CO in the MCC. The observed variation in exposures can be explained through understanding how the 558


various categories affect worker interaction with these machines and their exhaust. By job category, those workers who directly interacted with the forklifts and trucks are the highest exposed. This includes the Forklift Driver, the Walker, Truck Driver, and Dock Worker. One work task was observed which may result in the highest CO exposure: a forklift was driven into the trailer of a big rig truck for the purpose of unloading materials. This activity constitutes operation of an internal combustion engine within an enclosed space. The Forklift Operator, or a Walker on foot alongside a forklift, inside the truck may be repeatedly exposed to very high CO ( > 100 ppm ) levels. The job categories with the lowest exposures appeared to be Attendant, Installer /Decorator, and General Foreman. Workers in these categories were mostly on the floor in North and South halls. The LOCD provided considerably more accurate CO exposure data than the DraÈger tubes. The overall accuracy of TWA measurements of the DraÈger tubes appeared to be about 3 ppm in the range of exposures, whereas the LOCD was accurate to about 1 ppm. The average of three LOCD samplers was in almost perfect agreement with the bag samplers. In contrast, the DraÈger tube measurements consistently underestimated the TWA CO concentration even when it was over 10 ppm Ðthey were 60% of those measured by the LOCD on average. Although the DraÈger tube measurements did not provide quantitatively correct exposure measurements, they did serve as a qualitative indicator of the MCC CO levels. An issue not discussed elsewhere in this report is worker exposures to engine pollutant emissions other than CO. Other pollutants including nitrogen dioxide, particles, or volatile organic compounds, emitted from diesel or propane engines, may cause irritation to the respiratory systems and mucous membranes of exposed workers. It is possible that although the MCC ventilation rate was sufficient to protect against excessive CO levels, it may have been insufficient for removing these other compounds to safe and non irritating levels.

Conclusions This field study has validated the use of the LOCD for occupational CO exposure assessment. The LOCD was able to withstand the rigors of workplace sampling without failing. A total of 154 8- h personal samples were collected on workshifts over 3 days. Exposure distributions were calculated with an estimated precision of 1 ppm. None of the LOCD failed and only one was lost. The LOCDs were used to compare results with the DraÈger diffusion tube, a commercial industrial hygiene tool for monitoring workplace CO exposures. The comparison suggests that the Journal of Exposure Analysis and Environmental Epidemiology (1999) 9(6)


DraÈger device read about 40% low overall, and 20% low for observed TWA concentrations 10 ppm. Operations in the MCC exposed workers to CO levels above those normal for non- workplace conditions, although nearly always lower than the Cal /OSHA standard. Seemingly small changes in the operation of forklifts and trucks, or the ventilation system may have a large impact on CO exposures. If an increase in forklift emissions or poorer ventilation patterns were to occur, CO exposures could cross the threshold from ``compliance'' to ``exceedence'' of occupational health and safety standards. On the other hand, careful maintenance and improvement of conditions nearly always assures exposures below the PEL. It is a tribute to the MCC operations that given the large number of forklifts and trucks that operate in the facility, the workers' CO exposures were, with one exception, not observed to exceed the exposure standards. This is particularly true given the task of ventilating a structure with an internal, underground loading dock. Nonetheless, an ongoing effort must be made to ensure that CO levels are controlled. This task will require continual vigilance on the part of the MCC building operators. Finally, one may speculate that exposures to CO may be of greater concern at other indoor facilities that have not undergone the scrutiny to the extent of the MCC. Acknowledgments We would like to thank Agnes Bodnar and Tim Arronson and the industrial hygiene staff at Crawford Risk Control Services for their help on this project. Thanks are due to the entire workforce and management ( SMG ) at the MCC as well as the City and County of San Francisco for their willing participation and permission to conduct this study. This work was supported, in part, by the Office of Building Technology and the Office of Energy Research, Laboratory Technology Research Program, of the U.S. Department of Energy under contract no. DE - AC03 - 76SF0098. References Apte M.G. A population - based exposure assessment methodology for carbon monoxide: development of a carbon monoxide passive sampler and occupational dosimeter. Doctoral Dissertation, University of California, Berkeley, LBNL - 40836, Lawrence Berkeley National Laboratory, Berkeley, CA 94720, 1997. Becker R.A., Chambers J.M., and Wilks A.R. The New S Language. A Programming Environment for Data Analysis and Graphics. Wads-

Journal of Exposure Analysis and Environmental Epidemiology (1999) 9(6)

Apte et al.

worth and Brooks / Cole Advanced Books and Software, Pacific Grove, CA, 1988. Blackwell S. Moscone Maladies. San Francisco Bay Guardian, San Francisco, CA, March 5, 11, 1997. CAL / OSHA. Title 8 Ð General Industry Safety Orders, Section 5155 Ð Airborne Contaminants. California Code of Regulations, Register 97 No. 22, 5 - 30 - 97, Sacramento, CA, 1997. Cobb N., and Etzel R.A. Unintentional carbon monoxide - related deaths in the United States, 1979 through 1988. Journal of the American Medical Association 1991: 166: 659 ± 663. Ely E.W., Moorehead B., and Haponik E.F. Warehouse workers' headache: emergency evaluation and management of 30 patients with carbon monoxide poisoning. American Journal of Medicine 1995: 98 ( 2 ) : 145 ± 155. Fawcett T.A., Moon R.E., Francia P.J., Mebane G.Y., Theil D.R., and Piantadosi C.A. Warehouse workers headache. Carbon monoxide poisoning from propane - fueled forklifts. Journal of Occupational Medicine 1992a: 34 ( 1 ) : 12 ± 15. Fawcett T.A., Moon R.E., Francia P.J., Mebane G.Y., Theil D.R., and Piantadosi C.A. Letters to the editor: warehouse workers' headache Ð the author replies. Journal of Occupational Medicine 1992b: 34 ( 9 ) : 871 ± 872. Fleming J.L., and Opheim G.S. Letters to the editor: warehouse workers' headache. Journal of Occupational Medicine 1992: 34 ( 9 ) : 872. Hossain M.A., and Saltzman B.E. Laboratory evaluation of passive colorimetric dosimeter tubes for carbon monoxide. Applied Industrial Hygiene 1989: 4: 119 ± 125. Katz E., and Osorio A.M. Letter to J. Moerschbaecher, Convention Facilities, City and County of San Francisco. Department of Health Services, Occupational Health Branch, Berkeley, CA, 1997. Litovitz T.L., Felberg L., White S., and Klein - Schwartz W. 1995 Annual report of the american association of poison control centers toxic exposure surveillance system. American Journal of Emergency Medicine 1996: 14 ( 5 ) : 587 ± 537. McCammon J.B., McKenzie L.E., and Heinzman M. Carbon monoxide poisoning related to the indoor use of propane - fueled forklifts in Colorado workplaces. Applied Occupational and Environmental Hygiene 1996: 11 ( 3 ) : 192 ± 198. NOAA. Global summary of the day data set on NOAA World Wide Web site. National Oceanic and Aeronautic Administration, Hyattsville, MD, 1997. OSHA. CFR Part 1910.1000 Air Contaminants, Table Z - 1; Amended by Fed. Reg. 58:35308, 35340 ( June 30, 1993 ) ; corrected by Fed. Reg. 58:40191 ( July 27, 1993 ) , U.S. Department of Labor, Occupational Safety and Health Administration, 1993. Rappaport S.M. Assessment of long - term exposures to toxic substances in air. Annals of Occupational Hygiene 1991: 35 ( 1 ) : 61 ± 121. USDHHS. In: Annually Published Reports: Vital Statistics of the United States, 1984 ± 1992: Volume II. Mortality, Part A. U.S. Department of Health and Human Services, Public Health Service, National Center for Health Statistics, Hyattsville, MD, 1986 ± 1992. USEPA. Air Quality Criteria for Carbon Monoxide. EPA - 600 / 8 - 90 / 045F. USEPA, Research Triangle Park, NC, 1991. Valerio F., Pala M., Lazzarotto A., and Balducci D. Preliminary evaluation, using passive tubes, of carbon monoxide concentrations in outdoor and indoor air at street level shops in Genoa ( Italy ) . Atmospheric Environment 1997: 31 ( 17 ) : 2871 ± 2876.

559