From Causal Discovery to InterventionSimulation: Modeling Precariousness and Depression in the HELIUS Study

This study investigates how the internal structure of depression–precariousness dynamics shapes responsiveness to external support. Using cross-sectional data from the HELIUS study, we first apply cycle-capable causal discovery algorithms to estimate directional dependencies between financial stress, domain-specific precariousness, and depressive symptoms. Our multi-resolution analysis—spanning both aggregate and symptom-level constructs—highlights financial stress as a plausible upstream driver of depression, with symptoms such as sleep disturbance and depressed mood emerging as initiators of broader social vulnerability. To translate these structural insights into dynamic predictions, we construct a family of analytically calibrated linear stochastic models. Simulations reveal that variations in internal structure can substantially alter system responses to financial stress reduction. Some models show rapid but transient improvements in depression, while others adjust more slowly yet sustain gains over time. Precariousness tends to respond more immediately across all configurations, though its trajectory also shifts subtly with structural variation. Together, these findings demonstrate that intervention outcomes depend not only on the magnitude of external input but also on latent properties of the system. More broadly, this study illustrates how combining causal discovery with dynamical modeling supports principled, simulation-based exploration of intervention scenarios—grounded in data, yet able to probe counterfactual structure. This approach offers a scalable framework for diagnosing hidden system constraints and informing targeted public health strategies in complex psychosocial domains.