TEXAS: a Tool for EXposure ASsessment-Statistical models for estimating occupational exposure to chemical agents.

Measurements of occupational exposure to chemical agents are performed by sampling and analyzing workplace atmospheres. In France, this is done by the industrial hygienists of the prevention network of the Social Security Service, who collect and then enter the data in the COLCHIC database. More than 900000 measurements performed in French companies over the past 25 years have been collected. Using this amount of data is major challenge for obtaining knowledge and predicting occupational exposures. This study presents the way in which statistical models are built and used on the basis of almost 19000 recent measurements of 26 frequent chemical substances. For a given substance, the models use 13 exposure determinants as inputs, such as the task performed, the occupation of the operator or the type of process employed. The models permit to estimate two parameters: the geometric mean and geometric standard deviation. These parameters are used to build an exposure profile. By combining them with the limit value, an exposure index is estimated using a Bayesian network. A decision rule based on the interpretation of this probability is proposed to qualify the predicted situation as 'well-controlled situation', 'controlled situation', and 'poorly controlled situation'. On the basis of this decision rule, 62% of predictions are true for all substances confounded, an average of 36% of predictions are approximate and only 2% of them are wrong. The result of this study led to the development of a pragmatic software tool named TEXAS, tool for exposure assessment, which enables industrial hygienists to obtain a rapid estimation of the level of exposure control as a function of simple determinants of work situations.

[1]  Hans Kromhout,et al.  Design of measurement strategies for workplace exposures , 2002, Occupational and environmental medicine.

[2]  Trevor Hastie,et al.  The Elements of Statistical Learning , 2001 .

[3]  Farman A Moayed,et al.  Application of artificial neural network models in occupational safety and health utilizing ordinal variables. , 2011, The Annals of occupational hygiene.

[4]  R. Vincent,et al.  COLCHIC-occupational exposure to chemical agents database: current content and development perspectives. , 2001, Applied occupational and environmental hygiene.

[5]  David J. Hand,et al.  A Simple Generalisation of the Area Under the ROC Curve for Multiple Class Classification Problems , 2001, Machine Learning.

[6]  J. Mulhausen,et al.  A strategy for assessing and managing occupational exposures , 1998 .

[7]  Isabelle Guyon,et al.  An Introduction to Variable and Feature Selection , 2003, J. Mach. Learn. Res..

[8]  K Teschke,et al.  Studying the determinants of exposure: a review of methods. , 1999, American Industrial Hygiene Association journal.

[9]  Ian H. Witten,et al.  The WEKA data mining software: an update , 2009, SKDD.

[10]  D K Burns,et al.  The HSE National Exposure Database--(NEDB). , 1989, The Annals of occupational hygiene.

[11]  Nir Friedman,et al.  Probabilistic Graphical Models , 2009, Data-Driven Computational Neuroscience.

[12]  Juha Laitinen,et al.  Default values for assessment of potential dermal exposure of the hands to industrial chemicals in the scope of regulatory risk assessments. , 2006, The Annals of occupational hygiene.

[13]  P. Sottas,et al.  An empirical hierarchical Bayesian unification of occupational exposure assessment methods , 2009, Statistics in medicine.

[14]  S M Rappaport,et al.  An exposure-assessments strategy accounting for within- and between-worker sources of variability. , 1995, The Annals of occupational hygiene.

[15]  Richard A. Olshen,et al.  CART: Classification and Regression Trees , 1984 .

[16]  D. Steinberg.  CART: Classification and Regression Trees , 2009 .

[17]  Andrew P. Bradley,et al.  The use of the area under the ROC curve in the evaluation of machine learning algorithms , 1997, Pattern Recognit..

[18]  P O Droz,et al.  Multimodel inference and multimodel averaging in empirical modeling of occupational exposure levels. , 2009, The Annals of occupational hygiene.

[19]  Hans Kromhout,et al.  Advanced REACH Tool (ART): overview of version 1.0 and research needs. , 2011, The Annals of occupational hygiene.

[20]  J. Lavoué,et al.  Formaldehyde Exposure in U.S. Industries from OSHA Air Sampling Data , 2008, Journal of occupational and environmental hygiene.

[21]  Leo Breiman,et al.  Random Forests , 2001, Machine Learning.

[22]  Marek J. Druzdzel,et al.  SMILE: Structural Modeling, Inference, and Learning Engine and GeNIE: A Development Environment for Graphical Decision-Theoretic Models , 1999, AAAI/IAAI.

[23]  Wouter Fransman,et al.  Use of the MEGA exposure database for the validation of the Stoffenmanager model. , 2012, The Annals of occupational hygiene.

[24]  R Stamm,et al.  MEGA-database: one million data since 1972. , 2001, Applied occupational and environmental hygiene.

[25]  Hans Kromhout,et al.  The Advanced REACH Tool (ART): incorporation of an exposure measurement database. , 2013, The Annals of occupational hygiene.

[26]  Carol Rice,et al.  A Source of Exposure Data for Occupational Epidemiology Studies , 1990 .

[27]  Raymond Vincent,et al.  Occupational exposure to mineral fibres: analysis of results stored on colchic database. , 2007, The Annals of occupational hygiene.

[28]  Anders L. Madsen,et al.  LAZY Propagation: A Junction Tree Inference Algorithm Based on Lazy Evaluation , 1999, Artif. Intell..

[29]  R. Vincent,et al.  Assessment of occupational exposure to chemicals by air sampling for comparison with limit values: the influence of sampling strategy. , 2014, The Annals of occupational hygiene.

[30]  Leo Breiman,et al.  Bagging Predictors , 1996, Machine Learning.

[31]  Structural equation models in occupational health: an application to exposure modelling , 2011, Occupational and Environmental Medicine.

[32]  Ron Kohavi,et al.  Supervised and Unsupervised Discretization of Continuous Features , 1995, ICML.