Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Synthesizing the Counterfactual: A CTGAN-Augmented Causal Evaluation of Palliative Care on Spousal Depression

Pietro Grassi, Roberto Molinari, Chiara Seghieri, Daniele Vignoli

Abstract

Spousal bereavement severely deteriorates mental health. While palliative care benefits dying patients, its "stress-buffering" effect on survivors' depression remains empirically elusive due to acute small-$N$ constraints in longitudinal dyadic data. This study evaluates the causal impact of palliative care on bereaved spouses while introducing Synthetic Data Generation (SDG) to resolve sample attrition in quasi-experimental designs. Using SHARE panel data, we augment the sparse treated cohort via a Conditional Tabular GAN, anchoring synthetic trajectories to empirical baseline constraints to preserve causal pathways. A Matched Difference-in-Differences estimator applied to the high-fidelity augmented dataset evaluates the treatment effect. Results reveal a non-linear psychological response. Palliative care initially exacerbates acute depressive symptoms at the time of loss ($β_0 = 0.218,\ p < 0.05$), reflecting the intense emotional confrontation of the intervention. However, a sustained stress-buffering effect emerges in subsequent periods ($β_2 = -0.763,\ p < 0.01$), indicating an accelerated long-term recovery compared to standard care. Estimates are highly robust to unobserved confounding (Oster's $δ> 1$). Substantively, we advocate for reconceptualizing end-of-life care as a dyadic public health intervention. Methodologically, we establish SDG as a robust analytical tool capable of powering fragile quasi-experiments in longitudinal social surveys.

Synthesizing the Counterfactual: A CTGAN-Augmented Causal Evaluation of Palliative Care on Spousal Depression

Abstract

Spousal bereavement severely deteriorates mental health. While palliative care benefits dying patients, its "stress-buffering" effect on survivors' depression remains empirically elusive due to acute small- constraints in longitudinal dyadic data. This study evaluates the causal impact of palliative care on bereaved spouses while introducing Synthetic Data Generation (SDG) to resolve sample attrition in quasi-experimental designs. Using SHARE panel data, we augment the sparse treated cohort via a Conditional Tabular GAN, anchoring synthetic trajectories to empirical baseline constraints to preserve causal pathways. A Matched Difference-in-Differences estimator applied to the high-fidelity augmented dataset evaluates the treatment effect. Results reveal a non-linear psychological response. Palliative care initially exacerbates acute depressive symptoms at the time of loss (), reflecting the intense emotional confrontation of the intervention. However, a sustained stress-buffering effect emerges in subsequent periods (), indicating an accelerated long-term recovery compared to standard care. Estimates are highly robust to unobserved confounding (Oster's ). Substantively, we advocate for reconceptualizing end-of-life care as a dyadic public health intervention. Methodologically, we establish SDG as a robust analytical tool capable of powering fragile quasi-experiments in longitudinal social surveys.

Paper Structure

This paper contains 21 sections, 1 equation, 5 figures, 5 tables.

Table of Contents

  1. Introduction
  2. Theoretical Framework
  3. Palliative Care and the Widowhood Effect: A Stress-Buffering Mechanism
  4. The Challenge of Causal Inference in Small-N Longitudinal Settings
  5. Data Augmentation for Causal Inference
  6. Data and Methodology
  7. Data Source and Analytical Sample
  8. Synthetic Data Generation: Integrating the CTGAN Architecture
  9. Causal Inference: Matched Difference-in-Differences
  10. Results
  11. Pre-Generation Diagnostics and the Restoration of Common Support
  12. Evaluation of Synthetic Data Generation
  13. Causal Estimates: The Effect of Palliative Care
  14. Discussion
  15. The Non-Linear Trajectory of Grief and Stress-Buffering
  16. ...and 6 more sections

Figures (5)

  • Figure 1: End-to-End Synthetic Data Generation Pipeline. The process integrates the core CTGAN machinery within a tailored longitudinal causal framework. After temporal alignment and long-to-wide transformation (1), a static structural skeleton is extracted (2) to condition both the generator and the critic (3). The generated synthetic dynamic trajectories (4) are explicitly anchored back to reality via Skeleton Injection (5). Finally, wide-to-long reconstruction applies deterministic temporal rules (6), and post-processing domain clamping (7) ensures logical and clinical boundaries (e.g., clipping EURO-D scores) before DiD estimation.
  • Figure 2: Pre-generation diagnostics: study estimates of the Matched Difference-in-Differences model applied to the raw observational SHARE sample. The statistically significant coefficients in the pre-treatment periods ($\tau \le -2$) highlight the critical violation of the parallel trends assumption, driven by residual selection bias and a fundamental lack of common support in the small-$N$ dataset.
  • Figure 3: Density distributions of key covariates. The comparison between the original observational data (Real) and synthetic data generated via CTGAN, TVAE, and TTVAE demonstrates CTGAN's superior ability to capture highly skewed clinical variables, such as the EURO-D depression score.
  • Figure 4: Event-Study Estimates of Palliative Care on Survivor's EURO-D Score (CTGAN + Matched DiD). Error bars represent 95% confidence intervals.
  • Figure A1: Heterogeneity by Gender. Event-Study Estimates of Palliative Care on Survivor's EURO-D Score for Female (above) and Male (below) sub-samples.