A causal machine learning approach to disentangling the pressure-satisfaction pathways in depression
Fuente:
arXiv
Saved in:
| Main Author: | |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866909987319578624 |
|---|---|
| author | Nan, Zhaojin |
| author_facet | Nan, Zhaojin |
| contents | Purpose: Prior research has established perceived pressure and life satisfaction as important correlates of depression, yet their causal interplay remains insufficiently identified. This study aims to disentangle whether satisfaction acts as an independent protective factor or operates by buffering pressure, and to identify population-specific risk profiles across students and workers. Methods: We applied a causal machine learning framework to harmonized data from India, China, and Malaysia (total N=28,243). We integrated random forests and logistic regression with Causal Mediation Analysis and Causal Forests. To resolve theoretical ambiguity regarding directionality, we validated the causal pathway between pressure and satisfaction using numerical simulation benchmarks. Results: Pressure emerged as the dominant predictor across all cohorts. Simulation-validated analysis confirmed a causal pathway flowing from Life Satisfaction -> Pressure -> Depression, rejecting the reverse hypothesis. Satisfaction mitigated depression partially through pressure reduction (proportion mediated around 15.1%), rather than functioning exclusively as a direct mechanism. A distinct developmental reversal was observed: younger age predicted vulnerability in students, whereas older age predicted risk in workers. Causal forests further revealed that the depressogenic impact of pressure was significantly amplified in students with high anxiety. Conclusion: Pressure acts as the proximal bottleneck for depression risk, while life satisfaction functions as an antecedent buffer. These findings challenge uniform risk models by highlighting age-related context dependence and suggest that precision interventions targeting stress reduction and high-anxiety subgroups offer the most effective pathway for breaking the causal cycle of depression. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2512_06654 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | A causal machine learning approach to disentangling the pressure-satisfaction pathways in depression Nan, Zhaojin Applications Purpose: Prior research has established perceived pressure and life satisfaction as important correlates of depression, yet their causal interplay remains insufficiently identified. This study aims to disentangle whether satisfaction acts as an independent protective factor or operates by buffering pressure, and to identify population-specific risk profiles across students and workers. Methods: We applied a causal machine learning framework to harmonized data from India, China, and Malaysia (total N=28,243). We integrated random forests and logistic regression with Causal Mediation Analysis and Causal Forests. To resolve theoretical ambiguity regarding directionality, we validated the causal pathway between pressure and satisfaction using numerical simulation benchmarks. Results: Pressure emerged as the dominant predictor across all cohorts. Simulation-validated analysis confirmed a causal pathway flowing from Life Satisfaction -> Pressure -> Depression, rejecting the reverse hypothesis. Satisfaction mitigated depression partially through pressure reduction (proportion mediated around 15.1%), rather than functioning exclusively as a direct mechanism. A distinct developmental reversal was observed: younger age predicted vulnerability in students, whereas older age predicted risk in workers. Causal forests further revealed that the depressogenic impact of pressure was significantly amplified in students with high anxiety. Conclusion: Pressure acts as the proximal bottleneck for depression risk, while life satisfaction functions as an antecedent buffer. These findings challenge uniform risk models by highlighting age-related context dependence and suggest that precision interventions targeting stress reduction and high-anxiety subgroups offer the most effective pathway for breaking the causal cycle of depression. |
| title | A causal machine learning approach to disentangling the pressure-satisfaction pathways in depression |
| topic | Applications |
| url | https://arxiv.org/abs/2512.06654 |