Abstract Details

Title: National occupational exposure databases, are they biased? Insight from the US and France.
Authors: Jerome Lavoue; Philippe Sarazin; Laurel Kincl; Igor Burstyn; Gautier Mater
Session: Large Scale Occupational Exposure Databases: Applications in Research and Prevention - St. Patrick, 18 Oct 2016, 10:30 AM - 11:45 AM

Background & Objectives

Occupational exposure databases (OEDB) reflecting monitoring by national agencies can aid in the prevention of occupational diseases. However, exposure levels found in these OEDBs may not be representative of typical workplace exposures. We studied several national OEDBs and explored the issue of bias. The objective is to present the insights gained across past and ongoing projects on two US and two French OEDBs.

Methods

The IMIS, set up in 1979, contains exposure measurements taken by the U.S. Occupational Safety and Health Administration (OSHA) inspectors to verify compliance with permissible exposure limits (PEL). Supplementary data containing all laboratory analyses performed at OSHA’s central laboratory is included in the Chemical Exposure Health Database (CEHD). In France, COLCHIC, set up in 1986, has data from prevention and research activities, and SCOLA, set up in 2006, has data from compliance activities. Statistical models were used to explore associations between ancillary variables and exposure levels in IMIS. IMIS and CEHD were linked to evaluate factors determining inclusion in IMIS of samples present in CEHD. COLCHIC and SCOLA were linked to explore differences in exposure levels by reason for sampling.

Results

As of 2012, IMIS, CEHD, COLCHIC and SCOLA contained approximately 800 000, 1.9M, 850 000, and 150 000 records, respectively, covering hundreds of chemical agents. Statistical modelling of IMIS, conducted on 511 047 records and 77 agents, showed exposure was associated with history of compliance-related penalties. Linkage between CEHD and IMIS, performed on 588 818 records and 78 agents, showed that non-detects as well as records that are chemicals part of analytical panels (e.g. metals) were less likely to be recorded in IMIS. Across 17 common agents with >500 samples in COLCHIC and SCOLA (2007-2012), the ratio of the median concentration in COLCHIC to the median concentration in SCOLA was on average almost four-fold.

Conclusions

Many factors intervene in the process leading to a measurement being taken and recorded in an OEDB. We observed associations between several of them with exposure levels in 4 different OEDBs, suggesting bias is present, albeit to an unknown extent due to the lack of a gold standard.