Reference GSD Values

Hello fantastic Expostats community,

I am looking to reference a “typical” value for workplace GSDs in my manuscript and have taken a look at some sources for guidance. These are CEN 2018, Buringh 1991, Kromhout 1993, and Symanski 2006. Buringh and Kromhout seem like good sources with typical GSDs of 2.5 and 2.7 respectively, but for Jerome, I was unsure how you got the median GSD of 2.5 from Symanski as referenced in your PDC?

Interestingly, the AIHA Blue Book references 2.5 quite a bit but doesn’t include an explanation or source as best I can find.

Thanks all!

Hello Mike,

I think 2.5 as typical workplace variability results from it being a “round” number mixed with data from the papers I mentionned in my PDCs, which tend to suggest slightly lower values. Here’s below my visual summary of the available studies (some results I might have obtained by transforming/treating the originally presented data).

Additionnaly, a former PH.D student summarized more recent data reported from France at a conference from the Canadian association for research on work ands health (CARWH) in 2002. This was only a preliminary analysis of partial data (not even bayesian) but to me it is mostly reassuring as to the relevance of the older studies above : public dropbox link to pdf

Does that help ?

PS: here are the full refs for all this :

Ouedraogo D., Pfitzer, H., Chouvet, M., Lavoué, J. (2022) Updating available information on typical exposure variability in the workplace. Canadian Association for Research in Work and Health (CARWH) 2022 conference. The Changing World of Work, Health and Research. September 15-16th, 2022 (Virtual).

Buringh, E., & Lanting, R. (1991). Exposure variability in the workplace : Its implications for the assessment of compliance. American Industrial Hygiene Association journal, 52(1), 6‑13.

Kromhout, H., Symanski, E., & Rappaport, S. M. (1993). A comprehensive evaluation of within- and between-worker components of occupational exposure to chemical agents. The Annals of occupational hygiene, 37(3), 253‑270.

Symanski, Elaine, Maberti, S., & Chan, W. (2006). A meta-analytic approach for characterizing the within-worker and between-worker sources of variation in occupational exposure. The Annals of occupational hygiene, 50(4), 343‑357. Meta-Analytic Approach for Characterizing the Within-Worker and Between-Worker Sources of Variation in Occupational Exposure | Annals of Work Exposures and Health | Oxford Academic

Thank you so much Jerome. I really really appreciate your time and these resources.

In the Symanski study, they reported everything as R0.95. Did you do some kind of correction on the R0.95 values to get to the median GSD of 2.5? If so, could you explain how? And what reference value did you use for the overall median?

I’m going to take a guess at Table 4: R0.95 between workers, by job by location(b), 50%ile: 4.8 and then some factor applied to that to convert R0.95 to GSD[84%ile/50%ile] (I’m sure my stats ignorance is showing here)

Re: the poster, the ITGA database is impressive! Did they have this volume of data as a result of the French law that aligns with CEN 689? As wasn’t that promulgated in 2008?

Is there a concern with using a minimum of n=3 per SEG for this analysis? (I ask because I did the same for my manuscript), and because of our very limited sampling, most datasets were n=3.
Was there a paper written for this? I would love to read it if so.


For the Symanski paper : It is from medians in Table 4, where the median values for R0.95(b) and R0.95(w) are provided. I realize though that my n is wrong, it is 306 rather than 571.

The relation of R with GSD ( b, w, or global) is R = exp[2X1.96*ln(GSD)].

R0.95 is the ratio of the 97.5th centile to the 2.5th centile

Table 4 gave me ln(GSDb) = 0.406 and ln(GSDw) = 0.83

Then I get to GSD total : exp( sqrt( 0.406^2+0.83^2)) = 2.52

For the poster : Certainly there was a jump in sampling after the 2010 regulation ( EN689 is only in 2018), but ITGA has always sampled much. They started as a lab for former French coal mines.

For n=3 we didn’t see a noticeable effect of sample size on gsd values (but wait for the peer reviewed paper!). This abstract was preliminary (not even bayesian), paper is not finished unfortunately…a MSc student might help with it part time over the next year : fingers crossed !.

Fantastic, thank you Jerome! I look forward to reading that paper.