Variability in exposure measurements

Whenever an occupational hygienist carries out an air sampling survey, either for a group of workers carrying out the same tasks in a similar way, or for the same individual worker carrying out the same job on different days, it’s inevitable that a range of results will be obtained.  However, although most people would probably accept that it would be unreasonable to expect all the results to be the same, in my experience many people, including health and safety professionals don’t fully appreciate just how much variation there is in worker exposures.

To illustrate this, let’s look at an example.

The following results were obtained from a survey for sulphuric acid mist during anodising of metal components. This is an electrolytic process where sulphuric acid to create an protective or decorative oxide layer on the surface of  the metal. The process is called “anodising” because the component forms the anode electrode of an electrical circuit.

As you can see, there is quite a wide spread, ranging from 0.04 mg/m³ to 0.20 mg/m³ . The highest result is 5 times higher than the lowest. Many people find this surprising, but it’s not untypical for sampling results. You might also have noticed that the results are not evenly distributed across the range. This can be seen more easily if we plot them on a histogram

Here’s another set of results from the same facility from a survey carried out at a later date

And plotted as a frequency histogram:

Again we can see a similar pattern but with a larger spread – the highest result is seventeen times higher than the lowest.

The pattern evident in these two examples is not untypical for sampling results. What we can see is that

1. there is a relatively wide spread of results
2. the majority are at the lower end of the range
3. there are a small number of results that considerably higher than the majority

Statistically, we typically find that the sample results for a group of “similarly exposed” workers conform to a log-normal distribution. This is a skewed distribution with the majority of results bunched together,  but with a small, though significant, number of very high results.

There is often a tendency to consider these low frequency high results as untypical and the temptation is to exclude them, particularly if they exceed the relevant exposure limit. Sometimes that may be valid – if a good reason can be found for rejecting them. But in many cases the high results are simply part of the natural variation in exposures and can’t be ignored.

There are other important implications too. Particularly for interpreting the results from surveys where only a few samples were taken. It is more likely that the results will be from the lower end of the distribution as the higher results have a low frequency.

Understanding variation is important for occupational hygienists as it has major implications for how we design our surveys  and interpret the results.

Advertisement