Prediction of suicidal ideation in shift workers compared to non-shift workers using machine learning techniques

Abstract

Background: Shift work can affect sleep and increase the risk of suicide. This study attempted to predict suicidal ideation according to shift work by using machine learning techniques.

Methods: We analyzed a total of 43,095 data conducted by using the 10-year Korean National Health and Nutrition Examination Survey (KHANES). Shift workers and daytime workers were categorized and analyzed using random forest (RF) and decision tree (DT) techniques of machine learning techniques.

Results: Shift workers were more than twice as likely to have suicidal ideation as daytime workers. The RF model showed an accuracy of 91.6% for shift workers and 98% for daytime workers. In the DT technique, the rate of suicidal ideation was the highest among shift workers (82.7%) when they were depressed and had an EuroQol-5 Dimension (EQ-5D) score of less than 0.71.

Limitations: Shift work type was evaluated questionnaire and based on screening data, it was not possible to reflect recent changes in the work type and we evaluated for only suicidal ideation for suicide risk factors.

Conclusion: The variables influencing the suicide risk of shift workers and daytime workers differ. In the case of shift workers, negative factors such as depression and low quality of life are risk factors for suicide. Efforts are needed to reduce risk factors through administrative and policy interventions to manage workers' health by early screening.

Keywords: Daytime workers; Depression; Machine learning; Shift workers; Suicide.