Integrating Health Records, Genomic, and Social Data to Stratify Adolescent Depression Risk

项目摘要 在美国，五分之一的青少年在18岁之前会经历抑郁发作。早期预防 可以抵消一生的发病率，包括工作和社会损害，物质使用和自杀行为。 在人群水平上预防青少年抑郁症的一个关键步骤是有效地检测个体 谁能从有针对性的干预中获益最多。然而，已知的风险因素（例如，阈下症状， 认知方式，人际关系因素）往往没有广泛的评估，直到年轻人提出 对于精神病护理，传统研究中建立的前瞻性风险筛查工具仍然很差 在临床环境中大规模实施，提供者可能无法常规收集或整合 补充措施。来自主要卫生系统的大规模常规电子健康记录（EHR） 这是一个克服这些先前限制的强大机会，但尚未被青少年利用。 抑郁症，往往缺乏环境和遗传数据，可以告知病因的理解和风险 分层K 08职业发展奖的总体目标是利用大规模的EHR数据， 将基因组和社会决定因素联系起来，以加强对高风险青年的系统识别 抑郁症在现实世界中的健康状况。在这个项目中，候选人将开发和验证一个新的 从美国主要医疗保健系统中识别青少年抑郁症病例的表型算法 包含超过600万人长达20年的纵向EHR数据的国家（目标1）; 全面评估基于EHR的青少年的一系列潜在的社会和基因组决定因素 抑郁症（目标2）;并应用现代统计和机器学习方法来训练和评估一个初始的 基于常规EHR数据的青少年抑郁症前瞻性危险分层模型（目标3）。改善 EHR中青少年抑郁症的表型和分层将促进新的研究途径， 将是后续R级赠款的基础，包括跨卫生系统的外部验证， 的风险分层和临床轨迹模型，并采取简短的预防性干预措施，以提高 那些处于危险之中的人。在精神病学和遗传流行病学的坚实基础和多学科的支持下， 在一个理想的环境中，世界一流的专家团队，候选人将获得新的专业知识，在预测 分析，生物医学信息学（特别是EHR基因组整合），青少年抑郁症和 通过强化指导研究、监督培训和专业发展， 活动该奖项将为候选人提供必要的培训，以发展成为一个完全独立的 临床知情的研究者，具有将数据科学、统计学 遗传学和发育流行病学，为早期抑郁症预防提供可行的策略， 增强韧性。

PROJECT ABSTRACT One in five adolescents in the United States will experience a depressive episode before age 18. Early prevention could offset a lifetime of morbidity including work and social impairment, substance use, and suicidal behavior. A critical step to preventing adolescent depression at a population level is the efficient detection of individuals who could benefit most from targeted intervention. However, known risk factors (e.g., subthreshold symptoms, cognitive styles, interpersonal factors) are often not widely assessed in practice until young people are presenting for psychiatric care, and prospective risk screening tools built in traditional research studies remain poorly implemented at scale in clinical settings where it may not be feasible for providers to routinely collect or integrate additional measures. Large-scale, routine electronic health records (EHRs) from major health systems present a powerful opportunity to overcome these prior limitations but have not yet been harnessed for adolescent depression and often lack environmental and genetic data that may inform etiological understanding and risk stratification. The overall aim of this K08 Career Development Award is to leverage large-scale EHR data with linked genomic and social determinants to enhance the systematic identification of young people at elevated risk of depression in real-world health settings. In this project, the candidate will develop and validate a novel phenotype algorithm for identifying adolescent depression cases from a major healthcare system in the United States containing up to 20 years of longitudinal EHR data for over six million individuals (Aim 1); integrate and comprehensively assess a range of potential social and genomic determinants for EHR-based adolescent depression (Aim 2); and apply modern statistical and machine learning methods to train and evaluate an initial prospective risk stratification model for adolescent depression based on routine EHR data (Aim 3). Improving the phenotyping and stratification of adolescent depression in EHRs will facilitate new avenues of research that will be the basis of subsequent R-level grants that include external validation across health systems, refinement of risk stratification and clinical trajectory models, and brief preventive interventions to enhance resilience in those at risk. Supported by a solid foundation in psychiatric and genetic epidemiology and a multidisciplinary team of world-class experts in an ideal environment, the candidate will acquire new expertise in predictive analytics, biomedical informatics (specifically EHR-exposome-genome integration), adolescent depression and prevention science through intensive mentored research and supervised training and professional development activities. This Award will provide the necessary training for the candidate to develop into a fully independent clinically informed investigator with a translational research program that bridges data science, statistical genetics, and developmental epidemiology to inform actionable strategies for early depression prevention and resilience promotion.