Machine learning in the analysis of mental health at work: a scoping review

This scoping review aimed to assess the role of machine learning in workplace mental health research by systematically analyzing existing studies to understand current methodologies, applications, and trends.

We conducted a comprehensive search across multiple databases, including EBSCO, Scopus, ProQuest, Web of Science, PsycINFO, IEEE, and ACM, screening a total of 5600 abstracts. Altogether, we analyzed 92 journal articles, conference papers, and book chapters published before September 2025.

Since 2020, there has been a notable increase in publications on the topic. Studies have mainly employed cross-sectional designs (73%) and workplace questionnaires (51%) targeting specific occupational groups (67%), particularly from Asia excluding China (41%). Supervised learning methods, such as Random Forest and Neural Networks, have been frequently utilized to investigate conditions like depression, burnout, and anxiety. Most studies predicting mental health at work using machine learning are currently conducted by data scientists as single-measurement studies, whereas longitudinal studies from medicine, epidemiology, social sciences, or behavioral sciences are comparatively rare. In the context of machine learning, prediction denotes the model's ability to infer outcomes based on input data. However, most publications do not systematically analyze the temporal dynamics of mental health or forecast mental health outcomes from an epidemiological perspective.

The application of machine learning in occupational mental health research remains in its preliminary stages, with a primary focus on methodology and computer science. The review highlights the necessity for interdisciplinary collaboration to fully leverage the potential of machine learning in advancing occupational health research.