A significant part of occupational hygiene activities is spent measuring worker’s occupational exposures levels. Considerable spatial and temporal variability is usually observed in most exposure assessment surveys, frequently with 10 fold variations in exposure intensity in conditions apparently similar. This has historically represented an important challenge to the interpretation of measured levels with regard comparison with occupational exposure limits (OEL). Progressively developed during the last 2 decades, there now exists a consensus framework for the analysis of exposure levels with regards to exposure limits. Within this framework exposure levels are assumed to follow, at least approximately, a lognormal distribution. Several parameters from the underlying distribution and deemed associated with health risk are estimated from a number of measurements and compared to the OEL.

These developments, although permitting a better assessment of risk compared to historical approaches, have not been widely adopted by occupational hygiene practitioners. Hence, they involve notions of statistics not usually taught in traditional education programs. Moreover they require calculations not usually feasible in common tools such as calculators or spreadsheet programs. While some specific tools have been developed over the years, usually through volunteer initiatives, most are lacking in several areas, be they accessibility, functionality, user-friendliness or complexity. In addition, uncertainty in parameter estimates has mostly been taken into account through formal hypothesis tests or the calculation of confidence intervals, the results of which are not easily conveyed to decision makers, hampering the ability of practitioners to efficiently communicate risk. Finally, available tools are standalone, and are not easily amenable to integration within an existing data management structure.

The Webexpo projects aims at improving current practices in the interpretation of occupational exposure levels through the creation of a library of algorithmic solutions to frequently asked risk assessment questions in occupational hygiene. Most of these questions require the estimation of parameters from one or several distributions. This will be performed based on Bayesian statistics which permit to provide answers in probabilistic terms (e.g. what are the odds that….), facilitating risk communication, and allow to tackle methodological issues traditionally rarely taken into account such as the data reported as not detected (an increasing concern). The list of question to be solved will be established through consensus with a group of international experts. In brief we envision 4 main exposure models for the Bayesian estimation:

1. Estimating parameters from one
distribution: This is the traditional similar exposure group approach. The
measurements are assumed to come from a distribution of exposures shared by a
group of workers. As an illustration, this model will permit to answer the
question: what is the probability that unmeasured exposures for this group exceed
the OEL more than 5% of the time.

2. Estimating parameters from a model
involving a one level hierarchy. This model extends the 1^{st} model by
permitting to estimate to what extent a group of workers do or do not share
similar exposures. The global exposure variability is split into within and
between worker variability. It is possible to assess the group risk but also
whether some individual workers might experience higher risk than the group. As
an illustration, this model will permit to answer the question: Although group
exposure seems acceptable, what is the probability that a random worker might
experience exposure exceeding the OEL more that 5% of the time.

3. Estimating parameters from a model
where the exposure distribution is affected by an external factor. This will permit to assess for example to
what extent an intervention to control exposure was successful, or whether
exposure depends on specific tasks. As an illustration, this model will permit
to answer the question: what is the probability that this intervention reduced
exposure by a factor of at least 2.

4. Estimating parameters from a model
combining situations 2 and 3. This will represent the most complex model. As an
illustration, this model will permit to answer the question: what is the
probability that a random worker performing task A might experience exposure
exceeding the OEL twice more often than a random worker performing task B.

Each of the previous models, estimated for both the lognormal and normal distributions, will include several modules to allow for maximum flexibility for future users of the algorithms: a data entry module, a module for estimating distributional parameters (the core Bayesian estimation module), a numerical data interpretation module, a graphical output module, and a module permitting to supplement the measurement data with external information, such as prior expert opinion.

International experts as well as a panel of practitioners will be consulted to enrich the core functionalities included and provide feedback and suggestions on how to most efficiently convey statistical uncertainty to non-statisticians.

The outcome of the project will be a library of algorithms written in several languages that will cater for very diverse uses: statistical language, web-based language, and language for standalone offline applications. The code and documentation will be publicly available to allow users to build their own applications. In addition, in a secondary knowledge transfer project, the library will be used as a basis for the creation of a free web-based application.

The Webexpo project will ultimately result in the availability for the IH community of a comprehensive toolbox for the interpretation of occupational exposure levels, with the added flexibility for users to build or adapt their own software instead of using a new one. This toolbox design will be the result of a large consensus about which functionalities are the most useful, and the various tools will all be based on the same powerful calculation engine.

The Webexpo project has been funded in 2015 for the period 2015-2018 by the Institut de recherche Robert-Sauvé en santé et en sécurité du travail.