TEXAS: a Tool for EXposure ASsessment. Modèles statistiques pour l’estimation de l’exposition professionnelle aux agents chimiques

Occupational exposure measurements of airborne substances realized by chemical laboratories of the French Health Insurance are gathered and stored into a database. The so called COLCHIC database contains more than 900.000 measurements of 800 different substances issuing from samples realized into French industries during the last 25 years. Using these results is an important issue for the knowledge and the prediction of occupational exposures. In this study, statistical models predicting the atmospheric concentration of 26 substances are built using more than 19.000 measurements. For a given substance, the models use determinants of exposure as inputs such as the worker’s job, the effective task, the type of local exhaust ventilation or the business activity of the enterprise are found. The models allow estimating the geometric mean and the geometric standard deviation of the distribution. These parameters are used for building an exposure profile, which is combined with the hazard related with the substance, materialized by the limit value. A bayesian network is used for integrating this combination for displaying a probability of risk. A decision rule is proposed based on the interpretation of this probability for qualifying the situation: “Well controlled”, “Controlled”, “Poorly controlled”. On this basis, 62% of estimations by the models are correct, 36% of estimations are approximate and only 2% are wrong. The result of this study is a pragmatic software tool, the Texas software: a Tool for Exposure Assessment. It allows the industrial hygienist to obtain the ordre of magnitude of the risk management level of airborne chemical substances depending on simple determinants of exposure collected at the workplace.