Establishing an Artificially Intelligent Framework for Improving Therapeutic Alliance with Obese African American Youth and Caregivers through Multimodal Monitoring of Empathetic Accuracy and Interper

项目摘要/摘要。肥胖是所有种族/民族的重大公共卫生问题 群体、年龄、性别和地理区域，但一些群体受到的影响比其他群体更不成比例。与非西班牙裔白人青少年相比，少数民族青少年，特别是AA，肥胖的可能性不成比例。这项拟议的研究旨在改进由社区卫生工作者(CHWS；治疗师)为患有肥胖的非裔美国人(AA)青少年(AAAO)提供的全面、多成分的基于家庭的行为干预(FIT家庭)。此应用程序扩展了遵循肥胖相关行为干预试验(Orbit)的研究计划，以开发行为治疗，利用我们最近使用创新机器学习(ML)模型所做的工作，彻底检查会话内治疗过程，特别是CHW同理心在治疗AAAO及其主要照顾者的治疗联盟(TA)的发展和维护中的作用。TA被认为是治疗的基本要素。心理辅导员由三个特征组成：1.治疗师与患者之间的情感纽带；2.患者与治疗师对治疗目标的认同；3.治疗任务的认同。TA是一种动态的、双向的和协同的治疗元素，其含义是跨越客户和治疗师的种族、年龄和性别的差异。治疗师的共鸣对于建立和维持一个强大的TA至关重要，其结果是由治疗联盟调解的。与特定的干预措施相比，同理心在治疗结果中解释了更多的差异，并一致地预测客户的变化。不幸的是，分析关键的促进性人际关系技能，如治疗师的同理心，在技术上是具有挑战性的，因为另一个人的行为意图是微妙的、上下文相关的(例如，治疗师和患者在个性、背景、文化价值观等方面不同)、互动性(双向和协同性)；并且可以通过多种方式(语调、生理学)来表达。为了更好地理解同理心，特别是在跨种族的背景下，需要经验方法和分析工具来识别实际临床遭遇中产生行为变化的同理心行为。我们将通过两个与同理心和治疗结果相关的指标来实施同理心：1)同理心准确性，定义为CHW对患者状态的准确理解；以及2)人际生理同步性(IPS)，定义为治疗师和患者IPS之间随时间的相关性(即心率[HR]、HR变异性、皮肤电反应、语音特征)。本研究使用这两种移情测量方法来理解移情、治疗联盟和治疗结果之间的关系。此外，这项研究将探索共情效应如何基于种族和治疗师-客户二元组的性别一致性/不一致性而变化。对于未来的临床翻译，收集的音频和生理数据将被用来建立不同程度共情的M-L分类模型。然后，这个模型可以用来为治疗师的培训提供实时反馈，以改善他们的共情反应。

PROJECT SUMMARY/ABSTRACT. Obesity is a significant public health concern among all racial/ethnic groups, ages, gender, and geographic areas, but some groups are more disproportionately affected than others. Ethnic minority adolescents, especially AAs, are disproportionately likely to be obese relative to non-Hispanic White adolescents. The proposed study is designed to improve a comprehensive multicomponent family- based behavioral intervention (FIT Families) delivered by Community Health Workers (CHWs; therapists) for African American (AA) adolescents with obesity (AAAO). This application expands a program of research that has followed the Obesity Related Behavioral Intervention Trials (ORBIT) for developing behavioral treatments, by capitalizing on our recent work using innovative machine-learning (ML) models to thoroughly examine within-session therapy processes, specifically the role of CHW empathy in the development and maintenance of the therapeutic alliance (TA) while treating AAAO and their primary caregivers. TA is considered an essential element of therapy. The TA is composed of three features: 1. Emotional bond between therapist and client; 2. Agreement between the client and therapist on the goals of treatment; and 3. Agreement on the tasks of therapy. TA is a dynamic, bidirectional, and synergistic element of therapy with implications across differences in race, age, and gender of the client and therapist. Therapist empathy is critical in establishing and maintaining a strong TA, with outcomes mediated by the therapeutic alliance. Empathy accounts for more variance in therapy outcome than specific interventions and consistently predicts client change. Unfortunately, analysis of crucial facilitative interpersonal skills such as therapist empathy is technically challenging as the intent of another's behavior is nuanced, contextual (e.g., therapists and client differ in personality, background, cultural values, etc.), and interactive (bidirectional and synergistic); and can be expressed in multiple ways (voice tone, physiology]). To better understand empathy, especially in a cross-race context, requires empirical approaches and analytical tools that can discern empathetic behaviors in actual clinical encounters that produce behavior change. We will operationalize empathy through two measures that are each associated with empathy and treatment outcomes: 1) Empathic accuracy, defined as the CHW's accurate understanding of the client's state; and 2) Interpersonal physiological synchrony (IPS), defined as the correlation between therapist and client IPS over time (i.e., heart rate [HR], HR variability, galvanic skin response, voice features). This study uses these two empathy measures to understand the relationship between empathy, therapeutic alliance, and treatment outcomes. Further, the study will explore how empathy effects change based on the race and gender concordance/discordance of therapist-client dyads. Toward future clinical translation, the audio and physiological data collected will be used to build a M-L classification model for differing degrees of empathy. This model could then be used to provide real-time feedback for therapists training to improve their empathic responses.