A Unified Framework for the Transportability of Population-Level Causal Measures

2025-05-19

AI Summary

Existing causal inference methods primarily focus on transporting risk differences across populations, neglecting widely adopted clinical effect measures—such as risk ratios, odds ratios, and numbers needed to treat—thereby limiting clinical relevance and policy applicability. This paper proposes the first unified transportability framework for both absolute and relative first-order population-level causal effects. It systematically characterizes their non-equivalent transport behavior under covariate distribution shift. Leveraging conditional exchangeability, we develop a double-identification strategy and a novel semiparametric estimator. We establish theoretical guarantees of consistency and asymptotic normality. Monte Carlo simulations and real-world data analyses demonstrate that our method achieves significantly lower bias and variance, and greater robustness, compared to state-of-the-art alternatives. The framework thus enables reliable, multi-dimensional translation of causal evidence into clinical decision-making and health policy.

Technology Category

• Reasoning under Uncertainty: Causality
• Machine Learning: Causal Learning
• Knowledge Representation and Reasoning: Action, Change, and Causality

Application Category

• Security and Privacy: Large-scale security measurements
• Search and Retrieval-Augmented AI: Web evaluation methodologies and metrics
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking

Abstract

Generalization methods offer a powerful solution to one of the key drawbacks of randomized controlled trials (RCTs): their limited representativeness. By enabling the transport of treatment effect estimates to target populations subject to distributional shifts, these methods are increasingly recognized as the future of meta-analysis, the current gold standard in evidence-based medicine. Yet most existing approaches focus on the risk difference, overlooking the diverse range of causal measures routinely reported in clinical research. Reporting multiple effect measures-both absolute (e.g., risk difference, number needed to treat) and relative (e.g., risk ratio, odds ratio)-is essential to ensure clinical relevance, policy utility, and interpretability across contexts. To address this gap, we propose a unified framework for transporting a broad class of first-moment population causal effect measures under covariate shift. We provide identification results under two conditional exchangeability assumptions, derive both classical and semiparametric estimators, and evaluate their performance through theoretical analysis, simulations, and real-world applications. Our analysis shows the specificity of different causal measures and thus the interest of studying them all: for instance, two common approaches (one-step, estimating equation) lead to similar estimators for the risk difference but to two distinct estimators for the odds ratio.

Problem

• Transport diverse causal measures beyond risk difference
• Enable generalization of treatment effects across populations
• Unify framework for first-moment causal effect transportability

Innovation

• Unified framework for transporting causal measures
• Identification under conditional exchangeability assumptions
• Classical and semiparametric estimators evaluation