Research Project 1

总结/摘要-项目1 预防儿童和青少年虐待因果数据科学中心项目1 （CHAMP中心）将涉及确定许多类别的预测和因果模型， 虐待暴露及其后果，并制定一套可扩展的决策 支持工具（DST）。这些DST将通过以下方式指导决策： 能够防止虐待和减轻与虐待有关的 结果。鉴于对非因果/非病因因素的目标进行干预， 改善健康或适应功能，这些决定要求从业者可以 自信地：1）确定导致健康或功能障碍的病因，2）针对 与特定儿童和家庭的这些损伤相关的外部或内部病因因素。 当病因复杂且必须进行干预时，这变得尤其具有挑战性。 定制，以精确地针对不同的原因，因为往往是真正的儿童和家庭， 符合与虐待有关的干预标准。项目1的目标是确定模型， 制定一套DST来指导此类决策。我们将通过应用先进的因果数据来做到这一点 科学方法，以几个大的，相关的数据集，以得出有效的因果模型，为几个 虐待暴露类别（性虐待、身体虐待、忽视）和虐待- 相关结果（创伤后应激障碍、抑郁、自残、攻击、物质 滥用、肥胖、功能障碍）。我们将把这些分析与机器学习结合起来 预测分类，以确定哪些预测模型可以准确地将儿童分类为风险 类别我们将与资源中心的成员密切合作， 传播和外展核心，使所有相关利益攸关方（领域 专家，初级保健儿科医生，儿童福利工作者，行为健康临床医生，机构 管理员、孩子和家庭成员），以确保Project 1开发的工具 有效、实用、可靠、安全。项目1开发的一个DST将构成基础 项目2将在现场试验中测试该工具。项目2的见解对项目2至关重要。 1迭代地改进它开发的DST。

SUMMARY/ABSTRACT – PROJECT 1 Project 1 of the Center on Causal Data Science for Child and Adolescent Maltreatment Prevention (CHAMP Center) will involve determining predictive and causal models for many categories of maltreatment exposures and their consequential outcomes, and developing a scalable set of decision support tools (DSTs) from these models. These DSTs will be designed to guide decisions by practitioners in a position to prevent maltreatment and mitigate the severity of maltreatment-related outcomes. Given that intervention on targets that are non-causal/non-etiological factors cannot improve health or adaptive functioning, such decisions require that that the practitioner can confidently: 1) identify etiological factors that cause health or functional impairments, and 2) target the external or internal etiological factors relevant for these impairments for a particular child and family. This becomes especially challenging when etiologies are complex and interventions must be customized to precisely target diverse sets of causes, as is often true for children and families who meet criteria for maltreatment-related interventions. The goal of Project 1 is to determine models and to develop a set of DSTs to guide such decisions. We will do this by applying advanced causal data science methods to several large, relevant data sets to derive valid causal models for several categories of maltreatment exposure (sexual abuse, physical abuse, neglect) and maltreatment- related outcomes (post-traumatic stress disorder, depression, self-harm, aggression, substance abuse, obesity, functional impairment). We will integrate these analyses with machine learning predictive classification to identify which predictive models can accurately classify children into risk categories. We will work closely with members of the Resource Core to develop the models and of the Dissemination & Outreach Core to engage the full spectrum of relevant stakeholders (domain experts, primary care pediatricians, child welfare workers, behavioral health clinicians, agency administrators, children, and family members) to ensure that the tools Project 1 develops are effective, practical, trustworthy, and safe. One of the DSTs that Project 1 develops will form the basis for Project 2, which will test this tool in a field trial. Insights from Project 2 will be essential for Project 1 to iteratively improve on the DSTs it develops.