Connecting precariousness and depression: From causal discovery to intervention simulation

Precariousness and depression are closely linked, but the potential causal structure of this relationship and the extent to which feedback between them may shape intervention responses remain unclear. Using cross-sectional data from more than 21,000 adults in the multi-ethnic HELIUS study, we use constraint-based causal discovery methods that allow for latent confounding and feedback to generate hypotheses about candidate graph structures linking precariousness domains and depressive symptoms. Because the data are cross-sectional and many edge orientations are not uniquely determined, we summarize the findings in terms of their stability across specifications and resampling, and treat suggested directions as tentative, data-compatible hypotheses rather than identified causal effects. Across specifications and algorithms, financial stress is the precariousness domain most consistently linked to depression in the causal discovery summaries. Directional edge marks are less stable: in FCI summaries, the edge pattern is more often consistent with financial stress preceding depression, whereas CCI and PC tend to leave the same relation unresolved. In addition, several symptom-level configurations suggest that depressive symptoms may contribute to downstream social precariousness. We then use a calibrated simulation model to examine how depression and precariousness could co-evolve under intervention scenarios across different feedback strengths compatible with the observed data. In the simulation model, slower but more persistent improvement follows intervention under stronger internal feedback, whereas faster but less durable change follows under weaker feedback. These findings suggest that intervention responses may depend not only on the size of the external change but also on the feedback structure linking precariousness and depression. Longitudinal and quasi-experimental studies are needed to clarify directionality, estimate feedback strengths, and assess the policy relevance of these patterns. • Causal discovery maps candidate structure linking precariousness and depression • Financial stress emerges as the most consistently recovered link to depression • Cross-method disagreement in edge directions keeps causal ordering unresolved • A reduced simulation model examines responses to reduced financial stress • Stronger feedback slows improvement but makes it more persistent over time