Application of Pattern Mining Methods to Assess Exposures to Multiple Airborne Chemical Agents in Two Large Occupational Exposure Databases from France

Our plain-English summary of Application of Pattern Mining Methods to Assess Exposures to Multiple Airborne Chemical Agents in Two Large Occupational Exposure Databases from France, by J.F. Sauvé, A. Emili, G. Mater (International Journal of Environmental Research and Public Health, 2022).

The study

• Authors: J.F. Sauvé, A. Emili, G. Mater
• Journal: International Journal of Environmental Research and Public Health
• Published: 2022;19(3)
• Cited by: 3 articles, as counted by NIH iCite
• Read it: PubMed · Publisher · Free full text

Abstract

Surveys of the French working population estimate that approximately 15% of all workers may be exposed to at least three different chemical agents, but the most prevalent coexposure situations and their associated health risks remain relatively understudied. To characterize occupational coexposure situations in France, we extracted personal measurement data from COLCHIC and SCOLA, two large administrative occupation exposure databases. We selected 118 chemical agents that had ≥100 measurements with detected concentrations over the period 2010-2019, including 31 carcinogens (IARC groups 1, 2A, and 2B). We grouped measurements by work situations (WS, combination of sector, occupation, task, and year). We characterized the mixtures across WS using frequent itemset mining and association rules mining. The 275,213 measurements extracted came from 32,670 WS and encompassing 4692 unique mixtures. Workers in 32% of all WS were exposed to ≥2 agents (median 3 agents/WS) and 13% of all WS contained ≥2 carcinogens (median 2 carcinogens/WS). The most frequent coexposures were ethylbenzene-xylene (1550 WS), quartz-cristobalite (1417 WS), and toluene-xylene (1305 WS). Prevalent combinations of carcinogens also included hexavalent chromium-lead (368 WS) and benzene-ethylbenzene (314 WS). Wood dust (6% of WS exposed to at least one other agent) and asbestos (8%) had the least amount of WS coexposed with other agents. Tasks with the highest proportions of coexposure to carcinogens include electric arc welding (37% of WS with coexposure), polymerization and distillation (34%), and construction drilling and excavating (34%). Overall, the coexposure to multiple chemical agents, including carcinogens, was highly prevalent in the databases, and should be taken into account when assessing exposure risks in the workplace.

In plain English

This study analyzed personal air measurements taken in French workplaces between 2010 and 2019 and stored in two national occupational exposure databases, COLCHIC and SCOLA. The authors kept 275,213 measurements above the limit of quantification, covering 118 chemical agents including 35 carcinogens. They grouped them into 32,670 "work situations", each a combination of industrial sector, job title, task and year. Asbestos measurements from SCOLA were left out because the task was not systematically recorded. Using data-mining methods first built to study which items shoppers buy together, the authors looked for agents found together. At least two agents were detected in 10,547 work situations (32%), with a median of three. 4403 work situations had at least two carcinogens. The most common combination was ethylbenzene with xylene, followed by cristobalite with quartz (1417 work situations). Only 8% of work situations with asbestos also had another agent detected; wood dust was lowest at 6%. The authors concluded that these methods can help identify common coexposure patterns. They noted that the databases are not a random sample of French workplaces.

We wrote this plain-English summary from the full article with the help of AI, and a second AI model checked each sentence against the article. It is not the authors' wording; the abstract and the article are their own account.

What the study did and found

• Design: Analysis of two French occupational exposure databases using pattern mining methods (frequent itemset mining and association rules mining) to find chemical agents measured together
• Who was studied: 275,213 personal air measurement records above the limit of quantification, for 118 chemical agents including 35 carcinogens, from French workplaces, 2010 to 2019. About 48% of records were below the limit of quantification and were excluded.
• Exposure: Airborne chemical agents, including asbestos, quartz, cristobalite, wood dust, metals and solvents. Coexposure meant at least two agents detected in the same work situation, where a work situation is a combination of sector, job title, task and year. Asbestos measurements from SCOLA were not included.
• Main results:
  • Of 32,670 work situations, 10,547 (32%) had measurable levels of at least two different agents, with a median of three agents per work situation
  • 4403 work situations were exposed to at least two different carcinogens (median 2, maximum 14)
  • Ethylbenzene with xylene was the most prevalent combination, with a support of 14.7% among work situations exposed to at least two agents
  • Quartz and cristobalite were found together in 1417 work situations
  • Asbestos and wood dust had the lowest proportions of work situations also exposed to another agent: 8% for asbestos and 6% for wood dust. The median across the 118 agents was 88%.
  • In the carcinogen co-occurrence analysis, asbestos was among the agents with fewer than 10 work situations exposed to at least one other agent
  • Excluding quartz and cristobalite, the most prevalent carcinogen combinations were hexavalent chromium with lead (368 work situations), MIBK with ethylbenzene (325) and benzene with ethylbenzene (314)
  • Among tasks, arc welding had the highest proportion of work situations with coexposures (70%), followed by offset and screen printing (65%)
  • 97.5% of work situations with detectable cristobalite also had detectable quartz, the association rule with the highest confidence
• The authors' conclusion: The authors concluded that market basket analysis methods can be useful for identifying prevalent coexposure patterns in databases of industrial hygiene measurements
• Limitations the authors state:
  • The databases do not represent a random sample of French workplaces
  • SCOLA covers only agents with binding regulatory exposure limits, so coexposures to other agents in the same workplace are less likely to be assessed
  • No measurement of an agent in a work situation does not mean there was no exposure
  • The authors did not consider whether agents in mixtures share similar health effects, except for carcinogens

Topics

Medical terms in this study, explained in our Lexicon: Occupational Exposure, IARC.

Related studies

• Estimated number of cancers attributable to occupational exposures in France in 2017: an update using a new method for improved estimates — C. Marant Micallef, H. Charvat, et al. (2023)
• An estimate of cancers attributable to occupational exposures in France — P. Boffetta, P. Autier, et al. (2010)
• Cancers in France in 2015 attributable to occupational exposures — C. Marant Micallef, K.D. Shield, et al. (2019)
• Occupational exposures and cancer: a review of agents and relative risk estimates — C. Marant Micallef, K.D. Shield, et al. (2018)
• Choice of rating method for assessing occupational asbestos exposure: study for compensation purposes in France — C. Gramond, P. Rolland, et al. (2012)
• Estimating the social cost of respiratory cancer cases attributable to occupational exposures in France — H. Serrier, H. Sultan-Taieb, et al. (2014)

Browse more research in International Journal of Environmental Research and Public Health, or return to medical research.