Department of Maternal and Child Health, Gillings School of Global Public Health, University of North Carolina at Chapel Hill, 135 Dauer Drive, Chapel Hill, NC, 27599-7445, United States; Injury Prevention Research Center, University of North Carolina at Chapel Hill, 137 East Franklin Street, Suite 500, Chapel Hill, NC, 27599-7505, United States.
Department of Epidemiology, Gillings School of Global Public Health, University of North Carolina at Chapel Hill, 135 Dauer Drive, Chapel Hill, NC, 27599-7445, United States.
Child Abuse Negl. 2019 May;91:78-87. doi: 10.1016/j.chiabu.2019.02.011. Epub 2019 Mar 2.
Child maltreatment research involves modeling complex relationships between multiple interrelated variables. Directed acyclic graphs (DAGs) are one tool child maltreatment researchers can use to think through relationships among the variables operative in a causal research question and to make decisions about the optimal analytic strategy to minimize potential sources of bias.
The purpose of this paper is to highlight the utility of DAGs for child maltreatment research and to provide a practical resource to facilitate and support the use of DAGs in child maltreatment research.
We first provide an overview of DAG terminology and concepts relevant to child maltreatment research. We describe DAG construction and define specific types of variables within the context of DAGs including confounders, mediators, and colliders, detailing the manner in which each type of variable can be used to inform study design and analysis. We then describe four specific scenarios in which DAGs may yield valuable insights for child maltreatment research: (1) identifying covariates to include in multivariable models to adjust for confounding; (2) identifying unintended effects of adjusting for a mediator; (3) identifying unintended effects of adjusting for multiple types of maltreatment; and (4) identifying potential selection bias in data specific to children involved in the child welfare system.
Overall, DAGs have the potential to help strengthen and advance the child maltreatment research and practice agenda by increasing transparency about assumptions, illuminating potential sources of bias, and enhancing the interpretability of results for translation to evidence-based practice.
儿童虐待研究涉及对多个相互关联的变量之间复杂关系进行建模。有向无环图（DAG）是儿童虐待研究人员可以用来思考因果研究问题中操作变量之间关系并做出关于最佳分析策略决策的一种工具，以最小化潜在的偏差来源。
本文旨在强调 DAG 在儿童虐待研究中的实用性，并提供一个实用的资源，以促进和支持 DAG 在儿童虐待研究中的使用。
我们首先提供了与儿童虐待研究相关的 DAG 术语和概念概述。我们描述了 DAG 的构建，并定义了 DAG 上下文中的特定类型的变量，包括混杂因素、中介因素和共发因素，详细说明了每种类型的变量如何用于告知研究设计和分析。然后，我们描述了 DAG 可能为儿童虐待研究提供有价值见解的四个具体场景：（1）确定要包含在多变量模型中的协变量以调整混杂因素；（2）识别调整中介因素的意外影响；（3）识别调整多种虐待类型的意外影响；（4）识别儿童福利系统中涉及的儿童数据中的潜在选择偏差。
总体而言，DAG 有可能通过提高对假设的透明度、阐明潜在的偏差来源以及增强结果的可解释性，从而有助于加强和推进儿童虐待研究和实践议程。
