Hello Michael,

Basically, all information come from Fig 4 in the NIOHS document you mention, which is also in NIOSH OESSM manual from 1977 (77-173) as technical appendix L

```
Based on the attached evaluations, it looks like this is only accurate at very low GSDs, like <1.3? But that is not typical of occupational exposures, correct?
```

Yes, one day below 0.5*OEL is only reassuring for very small GSDs. Most available publications report median GSDs of 2.4-2.5. As an illustration, in Kromhout et al, out of 165 GSD values only 2 had GSDs below 1.3.

Kromhout, H, E Symanski, and S M Rappaport. “A Comprehensive Evaluation of Within- and between-Worker Components of Occupational Exposure to Chemical Agents.” The Annals of Occupational Hygiene 37, no. 3 (June 1993): 253–70.

For typical GSDs, the fraction of the OEL to use would be closer to 0.1, even less. If you look at the British-Dutch and European guideline and the French regulation, they use a screening scheme as follow : take 3 samples. If all 3 are below 0.1OEL, green. If any value >OEL, red. If no decision, do the probabilistic sampling/interpretation procedure.

You can check various screening schemes directly on expostats, e.g. enter 8 with an OEL of 100.

Doesn’t this depend on the number of samples collected? Like if I collect 5 samples I am much more likely to get an exceedance of 50% of the PEL, as compared to collecting 1 sample?

Yes it depends on n for several points :

The screening threshold ( e.g. 0.5 vs 0.1) : the higher n, the higher the fraction, this is clear in EN689 where thresholds for various sample sizes are provided.

Indeed, when n increases the probability of getting one above the OEL increases. This has been described as a disincentive to sample. However, this is only if you use a bad decision scheme. One thing to keep in mind is that with a realistic sample size (e.g. 6), Any one value above the OEL out of 6 will suggest a problem (easy to see as it suggests an exceedance fraction of 16%). This is of course assuming all data is representative. So in a way datasets with one value above the OEL are easy to diagnose. The issue is that many samples of 6 values with none above the OEL actually correspond to overexposure.

```
Depending on the exceedance fraction, 50% of the PEL seems to be a poorly performing screening test, no? Paul Hewitt modeled this and found that for an EF of 25%, 6 samples would be needed in order to have 90% confidence that the action limit would be exceeded.
```

Indeed. But to be precise it depends for what purpose : one value below 0.5*OEL cannot be used to show compliance with any confidence. However (again appendix L ), except for low GSDs, one value above 0.5*OEL seems like a fairly good indicator of non compliance.

I haven’t done super extensive sims, but I like the French/British/Duch scheme : start with 3 ( accidents always being possible with only 1), stopping right there will be possible if all 3 are <0.1OEL, or any value is >OEL.

```
Should the action level be used as an exposure limit in expostats? My intution is no, because its basis is a screening step that the PEL is exceeded more than 5% of the time, which is the criteria we are assessing anyway. It is similar to the screening criteria in EN 689 like if n=3 is <10% of the OEL, then the exposure is acceptable.
```

exactly : the screening value is in itself a statistical extrapolation procedure. So it shouldn’t be used to modify the OEL

In a way you could even avoid using the thresholds/comparisons and put your initial measurements in expostats : if overexposure risk < 30 or 5 depending on your preference, green. If not, collect additional samples (still using any value above OEL as an indicator of RED).

One thing I have not worked on yet : above what value of overexposure risk at the initial assessment is it very unlikely that it might go below 30 or 5 when going for the full sample.