Interactive Exploration of Temporal Patterns in Electronic Health Records

电子健康记录中时间模式的交互式探索

• 批准号: 7838863
• 负责人: CATHERINE Plaisant-Schwenn
• 金额: $ 33.75万
• 依托单位: UNIV OF MARYLAND, COLLEGE PARK
• 依托单位国家: 美国
• 财政年份: 2009
• 资助国家: 美国
• 起止时间: 2009-09-30 至 2011-08-31
• 项目状态: 已结题
• 来源: https://reporter.nih.gov/project-details/7838863
• 关键词: Accident and Emergency department; Admission activity; Asthma; Biology; Biomedical Computing; Clinical; Clinical Research; Collaborations; Complex; Data; Data Quality; Data Set; Databases; Diagnosis; Doctor of Medicine; Doctor of Philosophy; Electronic Health Record; Event; Future; General Hospitals; Genes; Healthcare; Hospitals; Imagery; Individual; Informatics; Language; Left; Massachusetts; Medical; Mining; Patient Care; Patients; Pattern; Pharmaceutical Preparations; Pneumonia; Population; Principal Investigator; Process; Recording of previous events; Records; Registries; Reporting; Research; Research Activity; Research Personnel; Solutions; Specific qualifier value; Spottings; System; Testing; To specify; United States National Institutes of Health; Visual; Washington; Work; Workplace; base; blind; comparative effectiveness; data mining; design; effectiveness research; empowered; improved; interest; next generation; novel; prototype; theories; tool

DESCRIPTION (provided by applicant): As the use of electronic health records (EHRs) spreads, large databases of EHR data are becoming available but are underused because there is no easy-to-use interfaces to understand what is available, specific complex temporal queries, review the results, compare groups of patients and rapidly and easily iterate this process. Visual and interactive solutions can dramatically increase the benefits of EHR databases, leading to improved clinical research and patient care. Research activities often require comparing treatments, drugs, patients with or without certain genes, etc. To assembling the groups of patients to be studied require queries that have a temporal component, e.g. "Find patients whose onset of asthma followed within 3 months of treatment of pneumonia". Currently available systems make possible simple queries such as "Find patients who have the diagnoses of asthma and pneumonia" leaving users with the burden of shuffling through large numbers of results in search for the useful data. Specifying useful but more complex temporal queries with SQL or other languages is impossible for most medical researchers, and data mining results are found questionable and hard to interpret as users cannot control the blind mining process and problems with "dirty" data remain unseen and unaddressed. Novel interface designs are needed for 1) interactive query interfaces allowing researchers and clinicians to find data that show temporal patterns of interest in both numerical and categorical data 2) event history operators to align, rank, filter and group by the results visually, allowing researchers and clinicians to see patterns, exceptions, and possibly data quality problems in the data they retrieved 3) powerful comparison tools to explore alternatives (e.g. to conduct comparative effectiveness research), and annotation mechanisms to record findings and prepare reports. We believe that the future of user interfaces is in the direction of larger, information- abundant interactive displays that are easy to use and empower the user to make discoveries while being aware of the quality of the data used in the process. By bridging the worlds of data bases, user interface design and information visualization, the next generation of potent visual analytic tools and work environments can be developed. Visual and interactive solutions for specifying search and reviewing results can dramatically increase the benefits of EHR databases, leading to improved clinical research and patient care. Novel interface designs are needed for 1) interactive query interfaces allowing researchers and clinicians to find data that show temporal patterns of interest in both numerical and categorical data 2) event history operators to align, rank, filter and group by the results allowing researchers and clinicians to see patterns and exceptions in the data they retrieved 3) novel visual summaries of temporal categorical data.

描述（由申请人提供）： 随着电子健康记录 (EHR) 使用的普及，大型 EHR 数据数据库变得可用，但未得到充分利用，因为没有易于使用的界面来了解可用内容、特定的复杂时间查询、审查结果、比较患者组以及快速轻松地迭代此过程。可视化和交互式解决方案可以显着提高 EHR 数据库的优势，从而改善临床研究和患者护理。研究活动通常需要比较治疗、药物、具有或不具有某些基因的患者等。为了组装要研究的患者组，需要具有时间成分的查询，例如“查找肺炎治疗后 3 个月内出现哮喘的患者”。目前可用的系统使简单的查询成为可能，例如“查找诊断为哮喘和肺炎的患者”，这使用户不得不在大量结果中搜索有用数据。对于大多数医学研究人员来说，使用 SQL 或其他语言指定有用但更复杂的时间查询是不可能的，并且数据挖掘结果存在问题且难以解释，因为用户无法控制盲目挖掘过程，并且“脏”数据的问题仍然未被发现和解决。需要新颖的界面设计：1) 交互式查询界面，允许研究人员和临床医生找到显示数字和分类数据中感兴趣的时间模式的数据 2) 事件历史操作符以可视化方式对结果进行对齐、排名、过滤和分组，使研究人员和临床医生能够在他们检索的数据中看到模式、异常和可能的数据质量问题 3) 强大的比较工具来探索替代方案（例如，进行有效性比较 研究），以及记录发现和准备报告的注释机制。我们相信，用户界面的未来是朝着更大、信息丰富的交互式显示器的方向发展，这些显示器易于使用，使用户能够进行发现，同时了解过程中使用的数据的质量。通过连接数据库、用户界面设计和信息可视化的世界，可以开发下一代强大的视觉分析工具和工作环境。用于指定搜索和审查结果的可视化和交互式解决方案可以显着提高 EHR 数据库的优势，从而改善临床研究和患者护理。需要新颖的界面设计：1) 交互式查询界面，允许研究人员和临床医生找到显示数字和分类数据中感兴趣的时间模式的数据；2) 事件历史运算符根据结果进行对齐、排名、过滤和分组，使研究人员和临床医生能够查看他们检索的数据中的模式和异常；3) 时间分类数据的新颖的视觉摘要。