Dynamic Bayesian network analysis of the social determinants of mental health.

Mental disorders contribute substantially to the global burden of disease, accounting for up to 16.5% of all years of healthy life lost due to disability and premature mortality. Epidemiological evidence indicates that mental health problems are associated with a diverse range of demographic, social, and economic factors, referred to collectively as social determinants; however, the causal mechanisms underlying these associations are widely recognized to be complex and are only incompletely understood. Here, we use recently developed structure learning methods for Bayesian networks and high-quality panel data from Australia to construct a provisional dynamic network model of the causal dependencies connecting a broad selection of social determinants and mental health. This provisional causal model identifies a heterogeneous set of proximate risk-modifying factors (direct causes), including subjective financial well-being, community connectedness, loneliness, and general health, that mediate the individual-level mental health effects of all remaining variables included in our analyses. Simulation analyses indicate that ideal preventive interventions targeting people's sense of financial security, local community engagement, and loneliness have the greatest capacity to improve population mental health outcomes, while significant reductions in the prevalence of mental health problems may also be achieved by promoting physical well-being and participation in volunteer or charity work and paid employment. We conclude that policies such as a Job Guarantee that are capable of simultaneously altering multiple adverse (or protective) social and economic exposures are likely to be critical in effectively addressing the substantial personal and societal costs of mental health-related disability.