SCH: INT Re-envisioned Chat-assessment for Real-time Investigating of Nursing and Guidance

20年前，美国国家医学科学院(原美国国家医学科学院)发表了开创性的出版物 医学研究所)，错误是人类和跨越质量鸿沟，使全国聚焦于 医疗保健的安全性和质量，但美国的患者结局指数继续落后于其他国家 工业化国家。2009年《美国复苏和再投资法案》要求医保 供应商采用电子健康记录(EHR)系统，导致广泛采用EHR，尽管 主要用于计费目的，而不是研究或质量改进工作。因此，EHR对 卫生保健质量已经趋向于医生效率和指南遵从性的领域。 尽管有大量证据表明，护理质量与急诊患者的预后直接相关 在护理销售方面，护士往往缺乏及时的信息来改善个别患者的预后，以及 患者群体的预后指数在石灰上的变化很慢。广泛采用电子健康记录 在美国的医院，现在允许为医院的所有患者确定结果质量指标 向护士提供实时反馈。质量指标通常只能通过将其他指标拼凑在一起来确定 确定事件发生的信息，例如，挖掘隐藏在护理笔记中的信息。 目标是为护理和指导(关怀)的实时调查开发图表评估， 用于实时报告和预测护理质量指标的自动机器学习系统 住院病人协助护士制定护理计划。CARI NG将向矿山反映算法创新 来自多源、异质数据的序列模式，包括护理叙述，产生了稳健的 预测模型对不确定的标签不敏感，并随着卫生保健实践的变化而演变。 关怀将使用相互连接的张量来表示EHR数据，捕获更高阶的关系，时态 加权，即越新的数据获得更多的权重，并将领域专家反馈纳入 发展。虽然关怀最初将为我们的十家医院开发行业合作伙伴 Emory Healthcare，其灵活的改进将使其他医疗保健机构能够适应。结果 该项目将为护士提供实时可操作的数据，以提高护理质量 立即接收。此外，该系统还可以在卫生信息基础设施中实施 机构级别，集成多规模和多层次的临床、上下文和组织数据 围绕每个患者进行实时报告并整合到预测模型中。

Two decades have lapsed since the seminal publications of the National Academy of Medicine (formerly the Institute of Medicine), To Err Is Human and Crossing the Quality Chasm, cast a national spotlight on health-care safety and quality, yet US patient outcome indices continue to lag behind those in other industrialized countries. The 2009 American Recovery and Reinvestment Act mandated health-care providers adopt electronic health record (EHR) systems, leading to widespread EHR adoption, albeit primarily for billing purposes rather than research or quality improvement efforts. Thus EHR impact on health-care quality has tended to be in the domains of physician efficiency and guideline compliance. Despite a large body of evidence that nursing quality is directly related to patient outcomes in the acute care selling, nurses often lack timely information to use in improving individual patient outcomes, and indices of outcomes across patient populations are slow to budge over lime. Widespread adoption of EHRs in U.S. hospitals now allows determination of outcome quality indicators for all patients in a hospital for real-time feedback to nurses. Quality indicators are often only determined by piecing together other information to determine occurrence of an incident, e.g., exhuming information buried in nursing notes. The goal is to develop Chart-assessment for Real-lime Investigation of Nursing and Guidance (CARING), an automated machine learning system to report and predict nursing quality indicators in real-time for hospitalized patients to assist nurses in care planning. CARI NG will reflect algorithmic innovations to mine sequential patterns from multi-sourced, heterogeneous data including nursing narratives, yielding robust predictive models that are insensitive to uncertain labels and evolve with changes in health-care practices. CARING will represent EHR data using inter-connected tensors, capturing higher-order relations, temporal weighting, i.e., more recent data receives more weight, and incorporating domain expert feedback in development. Although CARING will be developed initially for the ten hospitals of our industry partner Emory Healthcare, its flexible refinement will enable adaptation at other health-care institutions. Outcomes of this project will give nurses actionable data in real time to improve nursing care quality that they do not receive now. Moreover, this system can be implemented into the health information infrastructure at an institutional level, integrating multi-scale and multi-level clinical, contextual, and organizational data surrounding each patient for real-time reporting and incorporation into predictive models.