Causal Diffusion Models for Counterfactual Outcome Distributions in Longitudinal Data

2026-04-14

AI Summary

This work addresses the challenges of time-varying confounding bias and uncertainty quantification in longitudinal causal inference. We propose the Causal Diffusion Model (CDM), which, to our knowledge, is the first to employ denoising diffusion probabilistic models for generating full counterfactual outcome distributions. CDM leverages a residual denoising architecture and a relational self-attention mechanism to capture complex temporal dependencies and multimodal trajectory structures. Notably, it achieves robust prediction and calibrated uncertainty quantification without requiring explicit deconfounding strategies such as inverse probability weighting or adversarial balancing. Experiments on a pharmacokinetic–pharmacodynamic tumor growth simulator demonstrate that CDM improves distributional accuracy by 15–30% in terms of 1-Wasserstein distance over existing methods, while maintaining state-of-the-art or comparable point estimation accuracy (RMSE) even under high confounding.

Technology Category

• Machine Learning: Causal Learning
• Reasoning under Uncertainty: Causality
• Computer Vision: Diffusion Models for Vision

Application Category

• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs

Abstract

Predicting counterfactual outcomes in longitudinal data, where sequential treatment decisions heavily depend on evolving patient states, is critical yet notoriously challenging due to complex time-dependent confounding and inadequate uncertainty quantification in existing methods. We introduce the Causal Diffusion Model (CDM), the first denoising diffusion probabilistic approach explicitly designed to generate full probabilistic distributions of counterfactual outcomes under sequential interventions. CDM employs a novel residual denoising architecture with relational self-attention, capturing intricate temporal dependencies and multimodal outcome trajectories without requiring explicit adjustments (e.g., inverse-probability weighting or adversarial balancing) for confounding. In rigorous evaluation on a pharmacokinetic-pharmacodynamic tumor-growth simulator widely adopted in prior work, CDM consistently outperforms state-of-the-art longitudinal causal inference methods, achieving a 15-30% relative improvement in distributional accuracy (1-Wasserstein distance) while maintaining competitive or superior point-estimate accuracy (RMSE) under high-confounding regimes. By unifying uncertainty quantification and robust counterfactual prediction in complex, sequentially confounded settings, without tailored deconfounding, CDM offers a flexible, high-impact tool for decision support in medicine, policy evaluation, and other longitudinal domains.

Problem

• counterfactual outcomes
• longitudinal data
• time-dependent confounding
• uncertainty quantification
• causal inference

Innovation

• Causal Diffusion Model
• Counterfactual Inference
• Longitudinal Data
• Denoising Diffusion Probabilistic Models
• Temporal Confounding