Causal decomposition analysis aims to identify risk factors (referred to as “mediators”) that contribute to social disparities in an outcome. Despite promising developments in causal decomposition analysis, current methods are limited to addressing a time-fixed mediator and outcome only, which has restricted our understanding of the causal mechanisms underlying social disparities. In particular, existing approaches largely overlook individual characteristics when designing (hypothetical) interventions to reduce disparities. To address this issue, we extend current longitudinal mediation approaches to the context of disparities research. Specifically, we develop a novel decomposition analysis method that addresses individual characteristics by (a) using optimal dynamic treatment regimes (DTRs) and (b) conditioning on a selective set of individual characteristics. Incorporating optimal DTRs into the design of interventions can be used to strike a balance between equity (reducing disparities) and excellence (improving individuals’ outcomes). We illustrate the proposed method using the High School Longitudinal Study data.

中文翻译：

---

时变中介的因果分解分析：设计个性化干预措施以减少社会差异


因果分解分析旨在识别导致结果中社会差异的风险因素（称为“中介因素”）。尽管因果分解分析取得了有希望的发展，但目前的方法仅限于解决固定时间的中介因素和结果，这限制了我们对社会差异背后的因果机制的理解。特别是，在设计（假设的）干预措施以减少差异时，现有方法在很大程度上忽视了个体特征。为了解决这个问题，我们将当前的纵向中介方法扩展到差异研究的背景下。具体来说，我们开发了一种新颖的分解分析方法，通过（a）使用最佳动态治疗方案（DTR）和（b）对一组选择性的个体特征进行调节来解决个体特征。将最佳 DTR 纳入干预措施设计可用于在公平（减少差距）和卓越（改善个人成果）之间取得平衡。我们使用高中纵向研究数据来说明所提出的方法。