Formaldehyde Exposure in U.S. Industries from OSHA Air Sampling Data

Abstract

National occupational exposure databanks have been cited as sources of exposure data for exposure surveillance and exposure assessment for occupational epidemiology. Formaldehyde exposure data recorded in the U.S Integrated Management Information System (IMIS) between 1979 and 2001 were collected to elaborate a multi-industry retrospective picture of formaldehyde exposures and to identify exposure determinants. Due to the database design, only detected personal measurement results (n = 5228) were analyzed with linear mixed-effect models, which explained 29% of the total variance. Short-term measurement results were higher than time-weighted average (TWA) data and decreased 18% per year until 1987 (TWA data 5% per year) and 5% per year (TWA data 4% per year) after that. Exposure varied across industries with maximal estimated TWA geometric means (GM) for 2001 in the reconstituted wood products, structural wood members, and wood dimension and flooring industries (GM = 0.20 mg/m³. Highest short-term GMs estimated for 2001 were in the funeral service and crematory and reconstituted wood products industries (GM = 0.35 mg/m³. Exposure levels in IMIS were marginally higher during nonprogrammed inspections compared with programmed inspections. An increasing exterior temperature tended to cause a decrease in exposure levels for cold temperatures (−5% per 5°C for T < 15°C) but caused an increase in exposure levels for warm temperatures (+15% per 5°C for T >15°C). Concentrations measured during the same inspection were correlated and varied differently across industries and sample type (TWA, short term). Sensitivity analyses using TOBIT regression suggested that the average bias caused by excluding non-detects is approximately 30%, being potentially higher for short-term data if many non-detects were actually short-term measurements. Although limited by availability of relevant exposure determinants and potential selection biases in IMIS, these results provide useful insight on formaldehyde occupational exposure in the United States in the last two decades. The authors recommend that more information on exposure determinants be recorded in IMIS.

Keywords:

• IMIS
• mixed-effect models
• occupational exposure databank

ACKNOWLEDGMENTS

The authors thank Bruce Beveridge from OSHA for providing access to IMIS formaldehyde data. They would also like to thank Jan Erik Deadman for his helpful editorial comments. Jérôme Lavoué was supported by the Quebec Association for Occupational Hygiene, Health and Safety (AQHSST) and the Institut de Recherche Robert-Sauvé en Santé et en Sécurité du Travail (IRSST).

Notes

^A Source: National Climatic Data Center, National Environmental Satellite, Data, and Information Service, National Oceanic and Atmospheric Administration, Asheville, N.C.

^A Non-detects excluded.

^B Sample size.

^C Sample GM (mg/m³).

^D Sample GSD.

^E Percentage of values above the OSHA PEL (0.9 mg/m³); the statistic includes both short-term and TWA data because of limitations in the database design.

^A The linear model is the reference.

^B Dataset obtained by taking all NDs as TWA measurements.

^C Dataset created by distributing NDs between all measurement types according to their respective proportions in the “detected” dataset.

^A Approximate 95% confidence interval.

^A Sample size (in the dataset of detected values).

^B Prediction of the final linear model, GM (mg/m³) estimated for year 2001, programmed visit, average effect of state, and temperature at the median temperature: 12°C.

^C Estimated GM (mg/m³) for the same conditions as in Footnote B but with the TOBIT model, with the NDs distributed in all categories of sample type (dataset B).

^D Category not present in the TOBIT model.

^A Coefficient estimate.

^B Standard error of the coefficient.

^C p-value from a t distribution.

^D Corresponds to the categories: nonprogrammed inspection, TWA measurements, state OH, industry 7261, year 1979, and temperature (-16.89°C).

^E The variable year was modified to have a minimum of 0.

^F The coefficient has to be multiplied either by 0 (no effect) or the number of years past 1987 (modification of the temporal trend).

^G The variable temperature (°C) was modified to have a minimum of 0.

^H The coefficient has to be multiplied either by 0 (no effect) or the number of degrees past 15°C (modification of the linear trend).

^I SIC major division A.