Classification of Stroke Etiology Using Advanced Computational Approaches

项目总结/摘要 此提案是K23指导患者导向研究职业发展奖的申请 旨在准备博士里查夏尔马，医学博士，公共卫生硕士的职业生涯中独立，以病人为导向的学术 血管神经学夏尔马博士是一个早期的职业临床科学家谁拥有助理教授 耶鲁大学医学院的任命。她的长期目标是建立中风临床研究 实验室，应用先进的计算技术，以更好地通知实施循证医学 二级中风预防治疗以及新疗法的发现。要实现这一 Sharma博士的目标是实现她的培训目标：1）提高她在数据科学方面的技能， 处理高维数据集; 2）获得先进的计算技术的设施; 3）开发一个 生物标志物为基础的临床试验方法的基础;和4）获得必要的核心能力， 独立领导一个研究项目。通过这一培训赠款机制，夏尔马博士将有 有机会完成直接相关的正式课程，掌握基础研究方法， 产生一个工作组合，这将是获得独立的研究资金的关键。夏尔马医生有 她的指导团队的全力支持，其中包括心脏病专家和结果研究员哈兰克鲁姆霍尔茨博士 她擅长利用大数据来改善护理;她的共同导师劳伦·桑辛博士是一位血管神经学家 和一位在中风和生物数据调查方面具有专业知识的转化科学家;共同导师辛西娅博士 Brandt是一位急诊医学医生和信息学家，他是电子病历专家 研究;共同导师赵宏宇博士，一位擅长分析来自各种各样的数据的生物统计学家。 来源包括健康记录和生物数据;和共同导师胡曼卡梅尔博士，神经重症监护医生和 多产的中风临床研究者这个导师团队，沿着与合作者和受人尊敬的顾问 委员会将为夏尔马博士提供实现这些目标所需的指导、资源和支持 向学术独立过渡。 每年有近80万例缺血性中风，其中20-40%的中风的根本原因 不明因此，继发性卒中治疗可能不专门针对导致卒中的病理学。 卒中，可能导致复发性、可能更具破坏性的缺血性卒中的风险未减轻。 我们的目标是通过一系列分析来解决这种诊断不确定性。该研究将利用 新方法分析急性缺血性卒中患者的电子健康记录数据， 医疗保健系统（耶鲁-纽黑文医院、马萨诸塞州总医院和退伍军人事务部 美国卫生管理局医院）、蛋白质组学、神经影像学和来自一个 脑卒中患者亚组和临床试验数据，以实现2个特定目标：1）将非隐源性分类 使用EHR输入在住院期间分配给患者的缺血性卒中病因; 2）预测 诊断最初的隐源性患者，其卒中在1年随访时解决为特定病因。

PROJECT SUMMARY/ABSTRACT This proposal is an application for the K23 Mentored Patient-Oriented Research Career Development Award designed to prepare Dr.Richa Sharma, MD, MPH for a career in independent, patient-oriented academic vascular neurology. Dr.Sharma is an early career clinician scientist who holds an Assistant Professor appointment at the Yale School of Medicine. Her long-term goal is to establish a stroke clinical research laboratory that applies advanced computational techniques to better inform implementation of evidence-based secondary stroke prevention treatments as well as discovery of novel therapeutics. In order to accomplish this goal, Dr.Sharma seeks to fulfill her training objectives of: 1) furthering her skills in data science to manage and process high-dimensional datasets; 2) gaining facility with advanced computational techniques; 3) developing a foundation in biomarker-based clinical trial methodology; and 4) acquiring core competencies necessary to independently lead a research program. Through the mechanism of this training grant, Dr.Sharma will have the opportunity to complete directly relevant formal coursework, master fundamental research methodologies, and generate a portfolio of work that will be critical for obtaining independent research funding. Dr.Sharma has the full support of her mentoring team which includes Dr.Harlan Krumholz, a cardiologist and outcomes researcher who specializes in harnessing big data to improve care; co-mentor Dr.Lauren Sansing, a vascular neurologist and a translational scientist with expertise in stroke and investigation of biologic data; co-mentor Dr.Cynthia Brandt, an emergency medicine physician and an informatician who is an expert in electronic medical record research; co-mentor Dr.Hongyu Zhao, an expert biostatistician adept at analysis of data from a myriad of sources including health record and biologic data; and co-mentor Dr.Hooman Kamel, a neurointensivist and a prolific stroke clinical researcher. This mentorship team, along with collaborators and esteemed advisory committee, will provide Dr.Sharma with the guidance, resources, and support necessary to reach these goals and transition to academic independence. There are nearly 800,000 ischemic strokes per year, of which the underlying cause of 20-40% of these strokes is unknown. Consequently, secondary stroke therapies may not specifically target the pathology that led to the stroke, potentially resulting in an unmitigated risk for a recurrent, potentially more devastating ischemic stroke. We aim to address this diagnostic uncertainty through a series of analyses. The proposed research will utilize novel methods to analyze electronic health record data of patients admitted with acute ischemic stroke to 3 healthcare systems (Yale-New Haven Hospital, Massachusetts General Hospital, and the Veterans Affair Health Administration Hospitals in the United States), proteomic, neuroimaging, and survey data from a subgroup of stroke patients, and clinical trial data to achieve 2 specific aims: 1) to classify non-cryptogenic ischemic stroke etiologies assigned to patients during hospitalization using EHR inputs; 2) to predict the diagnosis of initially cryptogenic patients whose strokes are resolved to a specific etiology by 1-year follow-up.