P.1.11 'Creation of classification model using machine learning; to detect dysphonia work-related'. (April 2019)

• Record Type: Journal Article
• Authors: Gilbert, Natalia; Assar, Rodrigo; Martinez, Rodrigo
• Is Part Of: Occupational and environmental medicine. Volume 76(2019)Supplement 1
• Page Start: A80
• Page End: A80
• Publication Date: 2019-04
• Subjects: Medicine, Industrial -- Periodicals; Environmental health -- Periodicals; 616.980305
• DOI: 10.1136/OEM-2019-EPI.216
• Languages: English
• ISSNs: 1351-0711

Abstract

Abstract : The underdiagnosis of occupational disease causes severe damage to the health system. The classification of a disease as a professional is based on the decision on whether the present labor factors are sufficient for the generation of the disease, and this function is carried out by a qualified professional or committee. Occupational dysphonia is one of the 5 most frequent occupational diseases in Chile, whose condition impact on the labor productivity and the quality of life of the patient. Today there are no unified criteria among the occupational qualification decisión makers to decide on the sufficient of laboral factors of occupational dysphonia disease. Computerized systems have been developed to support clinical diagnosis decision-making process; among these, Machine Learning methods have been used to simulate the reasoning of the expert from the analysis and identification of complex patterns in large databases, so in this study it is suggested that the creation of a dysphonia classification model is possible employing Machine Learning tools. For this purpose, 103 cases obtained from patients with qualification results cause by dysphonia was analize in relation to the number of variables studied and their distribution for the observation of the characteristics that give identity to the groups studied. Subsequently, different classification models were developed using Machine Learning and the one that presented the best performance was chosen. Statistical analyzes show that of the 6 models of Machine Learning elaborated, Random Forest was the one that presented the best performance (accuracity=0.83 and Kappa value=0.61), variables that manage to establish identity to each group represent 26.5% of the total of studied variables. The results in this work show the potential of the use of computer tools can be useful as a support tool for diagnosis of occupational disease.