F. Hernandez, T. Hall, M. Phillips, R. Clinkenbeard, OUHSC, Oklahoma City, OK.
If representative data is an underlying assumption of the methodology used to draw an association between an effect and an outcome, it is important for researchers to determine if the data is representative. The purpose of this project is to investigate if individuals selected for sample collection in a historical data set were representative of their job classification. The database came from a chemical manufacturing plant, stratified by year and job classification. Fourteen strata, which had adequate numbers of samples (range 98-345), were selected for analysis; these encompassed two job classifications across 7 years. The expected distribution of samples among the individual workers in each job was calculated based on the hypothesis that sampling was random. The observed distributions of samples were compared with the expected distributions using a Chi Square goodness-of-fit test. In the job classification of monomer operator, the observed distributions were significantly different from the expected distribution in 6 of the 7 years. The monomer mechanic job classification did not show a statistically significant difference between the observed and expected distributions in any of the 7 years. The job classification of monomer operator had more employees and more samples than the job classification of monomer mechanic and consequently had higher statistical power. The conclusion reached is that the sampling of monomer operators was not random. The analysis for the monomer mechanic classification failed to reject the null hypothesis that sampling was random.
H. Basara, T. Hall, P. Jones, University of Oklahoma, Oklahoma City, OK; N. Esmen, University of Illinois, Chicago, IL; G. Marsh, J. Buchanich, University of Pittsburgh, Pittsburgh, PA; G. Marsh, J. Buchanich, University of Pittsburgh, Pittsburgh, PA.
Personal exposure data collected during industrial hygiene surveys are typically the only source of data available for estimating historical exposures. This type of data has been used for an exposure reconstruction without regard to what this data represented. Typically, this type of data has been collected to determine compliance with OSHA PELs, exposure related tasks, and/or where engineering controls were needed. When using this type of happenstance information, the distribution of the exposure data could not be assumed to approximate a randomly generated sampling distribution. The purpose of this study was to examine whether there is any difference between the means of a randomly sampled distribution of exposure data and the happenstance sample collected from industrial hygiene records. In this study, a random sample of chloroprene exposure data from two sites (Louisville and Ponchartrain) and four occupations (monomer/polymer mechanics and monomer/polymer operators) was selected from the existing industrial hygiene measurement data set. In this data set there were approximately 25,000 individual exposures estimates collected over a 20-year period. Exposure means estimated from the generated sample distribution for Louisville monomer operators and mechanics was 4.5 and 1.7 ppm for b-chloroprene, whereas the exposure means from the collected data set indicated 6.5 and 3.6 ppm. Based on analysis of these data, there was no significant difference between the sampling population and the data collected for industrial hygiene purposes. This may be explained by the frequency of exposure measurements in the data set because the data had a higher frequency in the lower range of exposures. Addition-ally, the similarity in population means may be attributed to overlap in tasks and work environments.
P. Jones, T. Hall, M. Phillips, R. Lynch, H. Basara, University of Oklahoma, Oklahoma City, OK; N. Esmen, University of Illinois, Chicago, IL; R. Leonard, DuPont, Newark, DE.
Exposure to ß-chloroprene during specific process-related tasks in the synthesis of synthetic rubber was estimated using a physical-chemical modeling approach. These task-specific exposures were then used to estimate cumulative exposure for groups of workers who performed these same tasks. The model-based approach to exposure estimation allowed for estimation of historical exposures in the absence of industrial hygiene data. The extent to which modeled exposures to process effluents represented actual exposures has been a question of interest. A database of 550 personal exposure estimates with tasks and time recorded from the period 1988-1996 was available from a recently completed epidemiologic study. The task information was extracted from the sampling forms and used in physical-chemical models to estimate exposure. The tasks modeled included process sampling in two different areas of the plant, in addition to draining and clearing of lines and equipment. Exposure estimates were presented as 8-hour time-weighted averages and were adjusted to account for the exposure time that was represented as the time respiratory protection was used. Task exposure estimates from the industrial hygiene database for cleaning line filters (screens) ranged from <10 ppm ß-chloroprene to more than 150 ppm ß-chloroprene as time-weighted averages. Exposure estimates derived from the physical-chemical modeling approach generally agreed with these estimates. However, the modeling approach demonstrated considerable variation, GSD > 4. Task-exposure estimates derived from personal exposure estimates also indicated considerable variation with a GSD of >5. Results of analysis demonstrated the utility of the physical-chemical modeling approach. Further, results indicate that exposure estimates derived using this approach adequately approximate actual exposures.
T. Hall, H. Basara, P. Jones, University of Oklahoma, Oklahoma City, OK; N. Esmen, University of Illinois, Chicago, IL; G. Marsh, J. Buchanich, A. Youk, University of Pittsburgh, Pittsburgh, PA.
A model-based approach for exposure reconstruction was used to develop exposure classes (five exposure classes ranging from 0.005 to >100 ppm ß-chloroprene) for workers within a facility population based on task and process information. Each exposure class had width boundaries that were one order of magnitude. Median exposure values from each exposure class were assigned to each individual for each year/occupation. These estimates were then used to calculate cumulative exposures and intensities of exposure for each employee across time. This approach was used in spite of the availability of more than 25,000 individual exposure estimates for individual workers. The selection of the exposure estimation methodology was based on observation that the available industrial hygiene sampling data was not collected for epidemiologic purposes and, therefore, could not be relied on to provide accurate estimations of worker exposure distribution parameters. Over or underestimation of exposure can result in uncontrolled exposure misclassification that impacts epidemiological analysis in unpredicted ways. In this analysis, results were from the model-based exposure assignment for 500 randomly selected workers in randomly selected occupations with actual exposure estimates. The worker selection criterion was that each subject was sampled at least 25 times/year over at least 5 years. Annual median exposure estimates were compared against class median exposure assignments from the model-based approach. Results of this study indicated that the model-based exposure approach overestimated the measured exposure in more than 50% of examined cases. However, these results indicated that median exposure estimates resulting from actual exposure data fell within the assigned exposure class more than 90% of the time. These data supported the model-based approach for exposure assignment and strongly suggested that exposure misclassification was within acceptable limits (for this study 10%).
J. Coble, M. Dosemeci, P. Inskip, NCI, Rockville, MD.
A variety of exposure metrics have been suggested for assessing exposure to ELF magnetic files in epidemiologic studies; however, a biological mechanism through which magnetic fields might alter or disrupt cellular processes has not been identified, and therefore, the appropriate exposure metric to use in epidemiologic studies has not been established. A set of exposure metrics for assessing occupational exposure to magnetic fields was developed for a population-based, case control study of brain cancer using information obtained from work history questionnaires supplemented with job modules and combined with a job exposure matrix. Detailed work histories were taken for the 1545 subjects with a total of 8536 jobs and 568 different occupational codes. Supplemental job modules were administered to 1012 subjects for 2043 of the jobs to collect detailed information about their work practices. For these jobs, an algorithm was used to calculate an exposure “score” based on the amount of time spent working with and their general proximity to electrical powered equipment. For each subject, exposure metrics were developed by combining the estimated exposure intensity for each job in their work histories with the duration of employment. The relative frequency of subjects categorized into different exposure groups based on the following metrics are compared: (1) cumulative occupational exposure intensity, (2) average occupational exposure intensity, (3) peak occupational exposure intensity, and (4) duration of occupational exposure > 4 mG. Magnetic field strength in occupational environments have substantial temporal and spatial variability that leads to large differences in exposure intensity between individuals with the same or similar occupation. The advantages and limitations of using job modules with an algorithm to assess occupational exposure to magnetic fields for epidemiology analysis are presented.
S. Nobile, C. Cigna, S. Francese, M. Patrucco, Politecnico di Torino DITAG, Torino, Italy; F. Lembo, A. Giglietta, SPreSAL ASl 14 VCO, Verbania, Italy; F. Lembo, A. Giglietta, SPreSAL ASl 14 VCO, Verbania, Italy.
A preliminary analysis of the health and safety conditions of workers in the dimension stone shops pointed out that hearing loss and silicosis should be considered the most serious expected health impairments. Because the stone shop situation appears quite different when compared with other manufacturing industries, a research program was developed in cooperation of the university and the National Safety Bureau to set up a suitable and practical approach to the measurement techniques and procedures on which the evaluation of the workers exposure can be based. The main problems are the definition of a sample duration suitable for both statistical significance and the available analysis techniques; the difficult environmental conditions at the shops, e.g., due to dampness, vibrations, and air dispersed oil mix; the remarkable variations in the duration of the various operations carried out; and the different pollutant emission in terms of rate and characteristics where different rock is treated (even mineralogical in the case of dust and of level and frequency of the emitted noise). The measurement campaigns were carried out at some Northwestern Italian shops where a number of different stone blocks are treated and there is a a wide range of different machines and tools. Dust concentration and noise levels were recorded during the various operations (e.g., bushhammering, sanding, polishing.) with both stationary and personal sampling lines; the environmental (temperature, humidity, and air velocity) and side conditions (type and number of machines operating in the proximity of the measurement site, types of worked stone, exhaust systems where present, etc.) were also recorded. The results are summarized, the most common problems pinpointed, and the basics of a guideline for an effective approach to the measurement techniques and procedures on which to draw the evaluation of the workers exposure to noise and airborne dust are discussed.
Posted May 30, 2006