Mental health and suicidal ideation in construction: using Bayesian network analysis to target multiple psychosocial hazards and their combined effects on mental health

This research addresses growing concern about psychosocial hazards in the building and construction industry and aims to inform more nuanced and targeted approaches to psychosocial risk mitigation. Using Bayesian network analysis with data from 2,454 construction workers recruited across 24 sites in Queensland, Australia, the study examined conditional interdependencies among workplace psychosocial hazards, general mental health, and suicide ideation. Following data preprocessing and model validation, approximately 47% of workers scored above the applied GHQ threshold, indicating poorer general mental health and elevated psychological distress. Role conflict and supervisor relationship conflict emerged as root nodes in the learned network structure. Sensitivity analysis identified past suicide ideation, role overload, role conflict, supervisor relationship conflict, low change consultation, and low procedural justice as influential parameters associated with general mental health. Scenario analyses indicated that simultaneously modelling lower role overload, higher praise and recognition, and greater job control increased the estimated probability that a worker would experience typical general mental health from 53% to 77%, exceeding the improvements associated with modelling changes to any single psychosocial hazard in isolation. The model also identified conditional relationships among past suicide ideation, general mental health, and future suicide ideation. By identifying interconnected psychosocial hazards and model-based priorities for further investigation, the findings may assist employers, regulators, and industry stakeholders in prioritising areas for psychosocial risk management and suicide risk mitigation. Future research should replicate the model across jurisdictions and evaluate related interventions using longitudinal and pre- and post-intervention designs.