Advanced Development and Dissemination of EMERSE for Cancer Phenotyping from Medical Records

项目摘要 全国癌症中心越来越多地使用电子健康记录（EHR）， 包含非结构化的自由文本笔记的存储库的巨大增长。这些笔记包括临床 这些概念在EHR的其他任何地方都找不到，需要这些概念来描述 患者的特殊“症状”。拥有如此多的电子格式数据， 促进临床和转化癌症研究的机会，但许多研究人员没有 获得实用工具，以协助从这些说明中确定和提取概念。为了解决这个 问题，密歇根大学综合癌症中心信息核心开发的信息 检索工具EMERSE-电子医疗记录搜索引擎-它已经实现了广泛的本地 成功地支持了数百项癌症研究，并导致了120多篇同行评议的出版物。 EMERSE提供了创新的功能，使研究，包括可视化的结果，可重复使用的病人 列表和可共享的搜索项集（“捆绑包”），以促进更多的可重复搜索。这个项目 建议在EMERSE内发展创新能力，并发展基础设施， 向全国的癌症中心传播EMERSE。本方案的软件创新包括（1） 安全地联网EMERSE实例，以实现跨癌症中心的搜索，这对于 罕见的癌症，并与不断增长的国家数据共享计划保持一致;（2）利用网络 EMERSE实例，以存储的搜索词捆绑包的形式在机构之间共享搜索知识， 促进跨研究中心和研究内检索的可重复性;（3）支持检索， 完全识别或去识别的临床记录，取决于用户的证书和预期用途， 这是跨机构搜索所需的;（4）支持包含本地化的自定义词典， 用于查询扩展的本体。技术和协作基础设施，以支持 将广泛传播EMERSE，包括组建EMERSE研究 信息学网络。五个癌症中心，每个中心都有不同的健康信息技术 环境，将加入ERIN的这个项目，并在本地实施EMERSE，使其可用于他们的 研究人员实施工作将错开进行，以便能够从一个地点的实施工作中吸取经验教训， 用于改进文档并应用于未来的站点。实施细节将包括如何 解决监管、机构审查委员会、安全和合规性问题。来自于 这些最初的实施者将被整合到一个“路线图”指南中，以帮助未来的癌症中心和其他 机构可以轻松地实施EMERSE。EMERSE将通过各种方式推广到其他癌症中心。 机制等将从多个方面评估这项工作的影响和有效性， 信息检索度量、可用性、实现、采用、满意度和效用。

PROJECT SUMMARY The increasing use of electronic health records (EHRs) by cancer centers nationwide has led to the tremendous growth of repositories containing unstructured, free text notes. These notes include clinical concepts that cannot be found anywhere else in the EHR, and these concepts are needed to characterize a patient’s specific ‘phenotype’. Having so much data available in electronic format provides unprecedented opportunities for advancing clinical and translational cancer research, yet many researchers do not have access to practical tools to assist with identifying and extracting the concepts from these notes. To address this problem, the University of Michigan Comprehensive Cancer Center Informatics Core developed the information retrieval tool EMERSE—electronic medical record search engine—which has achieved widespread local success in supporting hundreds of cancer studies and has led to over 120 peer-reviewed publications. EMERSE provides innovative features that enable research including visualization of results, reusable patient lists, and shareable sets of search terms (“Bundles”) that promote more reproducible searches. This project proposes to develop innovative capabilities within EMERSE and develop the infrastructure to widely disseminate EMERSE to cancer centers nationwide. The software innovations in this proposal include (1) securely networking the EMERSE instances to enable searches across cancer centers, which is important for rare cancers and is in alignment with growing national data sharing initiatives; (2) leveraging the networked EMERSE instances to share search knowledge between institutions in the form of stored search term Bundles, promoting the reproducibility of searches across sites as well as within studies; (3) supporting search with either fully identified or de-identified clinical notes, depending on the users’ credentials and intended use, which is needed for cross-institutional searching; (4) supporting the inclusion of localized custom dictionaries or ontologies to be used for query expansion. The technical and collaborative infrastructure to support the widespread dissemination of EMERSE will be developed, including formation of the EMERSE Research Informatics Network (ERIN). Five cancer centers, each with varying health information technology environments, will join ERIN for this project and locally implement EMERSE, making it available to their researchers. Implementations will be staggered so that lessons learned from implementing at one site can be used to improve the documentation and be applied to future sites. Implementation details will include how to address regulatory, institutional review board, security, and compliance issues. The collective knowledge from these initial implementers will be integrated into a ‘roadmap’ guide to help future cancer centers and other institutions implement EMERSE with ease. EMERSE will be promoted to other cancer centers through various mechanisms. The impact and effectiveness of this work will be evaluated across multiple dimensions including information retrieval metrics, usability, implementation, adoption, satisfaction, and utility.