prediction of workers pulmonary disorder exposed to silica dust in stone crushing workshops using logistic regression and artificial neural networks techniques

the work exposure conditions such as dust concentration, exposure time, use of respiratory protection devices and smoking status are effective to cause pulmonary function disorder. the objective of this study was prediction of pulmonary disorders in workers exposed to silica dust using artificial neural networks and logistic regression. a sample of 117 out of 150 workers employed in the stone crushing workshops placed in hamadan province, in the west of iran, was selected based on simple random approach. information about occupational exposure histories were collected using a questionnaire. to assess the pulmonary disorder status in the workers exposed to silica dust based on the spirometry indices as well as the workers characteristics the prediction models of artificial neural networks and logistic regression were employed using the spss software version 16. measurements of pulmonary function indices of the studied workers showed that the indices for workers having pulmonary disorder versus the others were statistically significant (p <0.01). the results of the obtained models showed that the artificial neural networks and the logistic regression had a high performance for prediction of pulmonary disorder status. however, the developed neural networks model had a better performance than the logistic regression model in viewpoint of sensitivity, specificity, kappa statistic and the area under roc curve. the neural networks prediction model was more accurate compared with the logistic regression. in this regards, the developed prediction model can be used as a helpful tool and guideline by occupational health experts for evaluating workers exposure conditions and determining the health priorities and control measures in the stone crushing workshops.