Presentation (CTI)

OBJECTIVES: Mental health literacy (MHL) is a multidimensional construct of growing importance for institutional and population-level public health planning. However, the causal relationships among its modifiable dimensions remain poorly understood, limiting their use for evidence-informed policy making and decision support. This study aims to learn the data-driven causal structure among key MHL dimensions using observational data and to identify relationships that are robust across multiple causal discovery methods.
METHODS: Seven causal structure learning algorithms, representing constraint-based (PC, FCI) and score-based (Tabu, GES) approaches, as well as hybrid approaches (MMHC, RSMAX2, H2PC), were applied to six continuous, modifiable MHL predictors derived from institutional survey data. Within-algorithm stability was evaluated using nonparametric bootstrap resampling, retaining edges with high support. Robustness across methods was assessed using a triangulation strategy to derive a consensus causal structure, with edge direction assigned through directional voting.
RESULTS: Several relationships among the six MHL dimensions were consistently identified across algorithms and bootstrap samples, forming a stable consensus causal structure at the population level. Only edges with bootstrap support greater than 0.7 and agreement across at least half of the algorithms were retained. While some relationships exhibited directional uncertainty inherent to observational causal discovery, a subset of directed dependencies remained invariant across methods and significance thresholds.
CONCLUSIONS: Although the core dimensions of mental health literacy are well recognized, their causal interconnections are not sufficiently understood. This study moves beyond purely theory-driven models and captures population-specific relationships among MHL components by directly learning the underlying structure from the data. The resulting causal structure offers an interpretable foundation for comparing, prioritizing, and designing interventions, thereby strengthening evidence-informed decision support and policy development in mental health literacy.