Abstract
OBJECTIVE: Workers' diseases and injuries are often highly related to work. However, due to limited resources and unclear work relatedness, workers' compensation insurance cannot cover all diseases or injuries among workers. This study aimed to estimate the status and probability of disapproval from national workers' compensation insurance using basic information from Korean workers' compensation system. METHODS: The compensation insurance data for Korean workers consists of personal, occupational, and claims data. We describe the status of disapproval by workers' compensation insurance according to the type of disease or injury. A prediction model for disapproval by workers' compensation insurance was established by applying two machine-learning methods with a logistic regression model. RESULTS: Among 42 219 cases, there were significantly higher risks of disapproval by workers' compensation insurance for women, younger workers, technicians, and associate professionals. We established a disapproval model for workers' compensation insurance after the feature selection. The prediction model for workers' disease disapproval by the workers' compensation insurance showed a good performance, and the prediction model for workers' injury disapproval showed a moderate performance. CONCLUSION: This study is the first attempt to demonstrate the status and prediction of disapproval by workers' compensation insurance using basic information from the Korean workers' compensation data. These findings suggest that diseases or injuries have a low level of evidence of work relatedness or there is a lack of research on occupational health. It is also expected to contribute to the efficiency of the management of workers' diseases or injuries.