Developing Personalized Predictive Models of Aggression

开发个性化的攻击性预测模型

基本信息

批准号：
10662685
负责人：
Colin Vize
金额：
$ 18.02万
依托单位：
UNIVERSITY OF PITTSBURGH AT PITTSBURGH
依托单位国家：
美国
项目类别：
财政年份：
2023
资助国家：
美国
起止时间：
2023-03-15 至 2028-02-29
项目状态：
未结题

来源：
https://reporter.nih.gov/project-details/10662685
关键词：
Address Adult Affect Age Aggressive behavior Anger Behavior Behavior Control Behavior Disorders Biological Cellular Phone Complex Computing Methodologies Data Data Reporting Detection Development Diagnosis Diagnostic Etiology Family Future Goals Heart Rate Heterogeneity Individual Intervention Life Machine Learning Measurement Mental disorders Mentors Mentorship Methodology Methods Modeling Motivation National Institute of Mental Health Nature Participant Patient Self-Report Pattern Personality Traits Persons Phenotype Physiological Positioning Attribute Prevention Process Protocols documentation Psychological Models Psychopathology Public Health Research Research Design Research Ethics Resources Risk Risk Factors Sampling Scientific Inquiry Scientist Selection Criteria Social Environment Surveys System Technology Testing Time Training Universities Youth adaptive intervention aggression prevention analytical method behavioral phenotyping career data streams economic impact effective therapy emotion regulation forest gradient boosting high risk population improved inclusion criteria innovation machine learning method mobile sensing novel personalized intervention personalized medicine personalized predictions predictive modeling predictive tools programs prospective psychologic random forest real time monitoring recruit research study sensor skills smartphone based device theories wearable device young adult

项目摘要

PROJECT SUMMARY/ABSTRACT Aggressive behavior is a transdiagnostic indicator of both youth and adult psychiatric disorders and a significant public health concern due to the direct harms to victims and its broader economic impact. Nonetheless, prediction of aggressive behavior is challenging due to significant variability in how, why, and when people act aggressively. This heterogeneity impedes efforts to establish etiological factors, identify biological substrates, and develop uniformly effective treatments. Though theories of aggression emphasize that it is a context-dependent, dynamic interpersonal behavior, research rarely attempts to study aggression in the contexts where it normally occurs and is most consequential (i.e., daily life). The current project seeks to improve on past research by studying the transdiagnostic mechanisms of aggression using novel analytic and measurement methodology necessary for pursuing a personalized medicine approach in aggressive behavior research and prevention. This project will use real-time data capture in conjunction with state-of-the-art analytic methods to deconstruct the heterogeneous behavioral phenotypes that relate to aggression. To achieve this, we will use relevant passively-sensed and self-reported data via smartphones from a sample of young adults (age=18-30; N=150) diagnosed with mental and behavioral disorders and at elevated risk for aggression. Data will be collected over the course of a 3-week ambulatory assessment protocol. We will apply machine learning methods capable of uncovering and modeling the complex dynamic processes observed in aggression at the level of each individual (i.e., personalized models) to prospectively predict aggressive urges and behavior. The results will pave the way for scalable just-in-time adaptive interventions tailored to an individual’s specific antecedents of aggression. The proposed study will contribute to NIMH Strategic Priorities 3.2 by 1) focusing on personalized models that can accommodate the complex topography of aggression and its antecedents and 2) applying innovative computational approaches (i.e., machine learning) to multiple streams of data (passively sensed, self-report) to identify potential just-in-time intervention targets for aggressive individuals. The comprehensive research and training plan detailed in this proposal will allow this candidate to address the primary research questions of the proposal and develop the expertise necessary to be an independent scientist. Specifically, this candidate will receive training in 1) personalized models of psychopathology and aggression; 2) methods for carrying out EMA-based studies and modeling intensive longitudinal data; and 3) collecting, processing, and predictive modeling with passive sensor data. This candidate has assembled a team of expert mentors (Wright, Jacobson) and consultants who possess the expertise to supervise the project and provide the training necessary to support the candidate in his development as an independent scientist. The expertise of the mentorship team, and the resources offered by the University of Pittsburgh, place the candidate in an ideal position to achieve his training, research, and career goals.

项目概要/摘要攻击性行为是青少年和成人精神疾病的跨诊断指标由于对受害者的直接伤害及其更广泛的经济影响，引起了重大的公共卫生问题。然而，由于攻击行为的方式、原因和方式存在显着差异，预测攻击行为具有挑战性。当人们表现出攻击性时，这种异质性会阻碍确定病因、识别病因的努力。尽管攻击理论强调生物底物，并开发一致有效的治疗方法。由于攻击行为是一种依赖于情境的、动态的人际行为，因此研究很少尝试研究攻击行为它通常发生且最重要的环境（即日常生活）。通过使用新颖的分析和方法研究攻击性的跨诊断机制，改进了过去的研究在攻击行为中追求个性化医疗方法所必需的测量方法该项目将使用实时数据捕获与最先进的分析相结合。解构与攻击性相关的异质行为表型的方法要实现这一目标，我们将通过智能手机使用来自年轻人样本的相关被动感知和自我报告数据（年龄=18-30；N=150）被诊断患有精神和行为障碍并且具有较高的攻击性数据。将在为期 3 周的动态评估方案过程中收集数据，我们将应用机器学习。能够揭示和模拟在攻击行为中观察到的复杂动态过程的方法每个人的水平（即个性化模型）来前瞻性地预测攻击性冲动和行为。结果将为根据个人具体情况量身定制可扩展的及时适应性干预措施铺平道路拟议的研究将有助于 NIMH 战略重点 3.2 1) 重点关注个性化模型可以适应复杂的攻击地形及其前因， 2）将创新的计算方法（即机器学习）应用于多个数据流（被动地感知、自我报告）来识别攻击性个体的潜在及时干预目标。本提案中详述的全面研究和培训计划将使该候选人能够解决以下问题提案的主要研究问题并发展成为独立机构所需的专业知识具体来说，该候选人将接受 1) 个性化精神病理学模型和攻击性；2) 进行基于 EMA 的研究和建模密集纵向数据的方法；该候选人组装了一个使用无源传感器数据进行收集、处理和预测建模的模型。拥有监督项目专业知识的专家导师（Wright、Jacobson）和顾问团队并提供必要的培训以支持候选人发展为独立科学家。导师团队的专业知识以及匹兹堡大学提供的资源使候选人处于实现培训、研究和职业目标的理想位置。