CLASSIFICATION OF OCCUPATIONAL INJURY CASES USING THE REGRESSION TREE APPROACH

Occupational safety and illness surveillance has made a great effort to spread a safety culture to all workplaces and a great deal of progress has been made in finding solutions that guarantee safer working conditions. This paper analyses occupational injury data in order to identify specific risk groups and factors that in turn could be further analyzed to define prevention measures. A technique based on rule induction is put forward as a non-parametric alternative tool for analyzing occupational injury data which specifically uses the Classification And Regression Tree (CART) approach. Application of this technique to relevant workrelated injury data collected in Italy has been encouraging. Data referring to 156 cases of injury in the period 2000-2002 were analyzed and lead to the factors that most affect work-related injuries being identified. According to the literature, up to the time of writing computer-intensive nonparametric modeling procedures have never been used to analyze occupational injuries. The aim of this paper is to use a real world application to illustrate the advantages and flexibility of applying a typical non-parametric epidemiological tool, such as CART, to an occupational injury study. This application can provide more informative, flexible, and attractive models identifying potential risk areas in support of decision-making in safety management.