A recent study by researchers at Osaka Public University has found that relying solely on the results from workplace 'stress checks' is inadequate for accurately predicting individuals who may take long-term sick leave due to mental health conditions like depression. This research analyzed data from over 230,000 stress assessments and applied machine learning techniques, concluding that a combination of various health-related metrics, including health checkup results, may be necessary for creating effective predictive models. The findings were published in an international academic journal, emphasizing a need for improvement in predictive accuracy in mental health management within corporate environments.

The 'stress check' system involves employees responding to a 57-item questionnaire assessing their feelings of work-related stress and anxiety. Currently, Japanese law mandates that companies with over 50 employees conduct these checks annually, with plans for extending this requirement to smaller companies by May 2028. These assessments are intended to support the mental health of workers and improve the overall workplace environment through proactive measures such as medical consultations based on the results.

Despite the initial optimism surrounding the stress checks as a tool for early detection of potential long-term sick leaves, diverse previous research efforts have failed to reach a consistent agreement on the effectiveness of these assessments. This inconsistency highlights the complexities of mental health evaluation and the importance of integrating a broader range of health data to enhance the predictive capability regarding employee well-being and sick leave in companies.