How Can We Make Invasive Non-Surgical Procedures Safer? Using Big Data to Identify Adverse Events and Opportunities to Mitigate Harm

背景：这是HSR&D IIR第二次提交建议，以过渡Hillary Mull博士。 从她的HSR&D职业发展奖(CDA)项目转向独立的退伍军人健康服务 研究生涯。拟议的工作寻求在穆尔博士成功的CDA项目的基础上，通过调整她的 建立和验证门诊手术到侵入性手术的监测模式的方法 非外科临床专科：介入心脏病学、介入放射学和胃肠道 内窥镜检查程序。这种基于信息学的方法依赖于在中组合文本和结构化数据字段 退伍军人事务部数据仓库(CDW)。穆尔博士的CDA资助的监测研究发现了一种不利的 事件发生率为9%，阳性预测值为85%，显著改善了不良事件检测。 意义/影响：目前，没有积极监测侵入性程序和初步 分析和与一线工作人员的交谈表明，不良事件的发生有一定的频率，并 对病人造成重大伤害。先前的工作发现，在这三个专业的侵入性程序导致后 程序急诊室就诊或住院人数每年超过50,000例。非退伍军人事务部文献 这表明，通过改善临床护理(例如，坚持 抗生素处方指南)。随着护理的日益增多，这一研究领域将变得更加重要 手术室外的过渡。在未被研究的环境中检测和监测不良事件 退伍军人事务部CDW中的现有数据与HSR&D资金优先事项C-医疗保健信息学一致。 创新：与来自全国各地的硬币专家一起，以及运营合作伙伴的支持 从每个临床专业、药房福利管理和退伍军人事务部信息学与计算 基础设施方面，穆尔博士建议应用她的CDA专业知识来建立一个监视系统，以识别入侵 具有可预防的不良事件的非外科手术；这些手术不受任何退伍军人管理局的约束 监视活动。这项工作解决的第二个缺口是缺乏全国可用的数据集捕获 程序性麻醉的使用。我们将使用图表回顾和文本查询数据挖掘方法来获得此结果 信息。我们的IIR工作的高潮将是一个关于不良事件和 潜在的可修改的影响因素，包括程序麻醉数据，可供退伍军人管理局研究人员使用。 具体目标：1)利用17-20财年数据开发和验证监控模型；2)测试监控 系统(应用模型系数，每月执行有限的图表审查)，从21-22财年开始，并进行改进 使用附加的CDW变量的系统；3)测试与可修改过程相关的假设，包括 是否有训练有素的麻醉提供者参与，或者患者是否接受了不适当的抗生素。 方法：我们的样本包括由专家临床医生共同定义的非手术侵入性程序。 调查人员。我们将遵循穆尔博士的CDA工作中概述的方法，将患者、程序、 来自CDW的提供商和设施数据。接下来，我们将审查案例，以确定可预防的事件 并使用图表回顾数据来估计Logistic回归模型预测 不良事件。模型系数将持续应用，以确定可能存在 不良事件对目标图表的审查。监测数据将被用来检验研究假设。 下一步/实施：通过这一IIR，以及在随后由合作伙伴资助的QUERI提案中，Dr。 穆尔和她的团队将建立一个为侵入性非手术设计的不良事件监测系统 可用于评估可修改的护理过程以防止患者受到伤害的程序。通过识别风险 对于可预防的不良事件的因素，我们可以确定我们与运营合作伙伴应该在哪些方面 专注于QUERI资助的QI工作，以提高患者的安全。研究结果将为以下工作提供急需的信息 研究和临床社区继续衡量和改善退伍军人管理局的护理质量。

Background: This is the second submission of an HSR&D IIR proposal to transition Dr. Hillary Mull, Ph.D. from her HSR&D Career Development Award (CDA) project toward an independent VA health services research career. The proposed work seeks to build on Dr. Mull's successful CDA project by adapting her approach to developing and validating a surveillance model for outpatient surgery to invasive procedures in non-surgical clinical specialties: interventional cardiology, interventional radiology and gastrointestinal endoscopy procedures. This informatics-based approach relies on combining text and structured data fields in the VA Corporate Data Warehouse (CDW). Dr. Mull's CDA-funded surveillance research identified an adverse event rate of 9% and had a positive predictive value of 85%, dramatically improving adverse event detection. Significance/Impact: Presently, there is no active surveillance of invasive procedures and preliminary analyses and conversations with frontline staff suggest adverse events occur with some frequency and impose significant patient harm. Prior work found invasive procedures in these three specialties result in post- procedure emergency room visits or hospitalizations exceeding 50,000 cases annually. Non-VA literature suggests half of this utilization may be preventable with improvements in clinical care (e.g., adherence to antibiotic prescribing guidelines). This field of research will become even more important as care increasingly transitions outside the operating room. Detecting and monitoring adverse events in understudied settings using existing data in the VA CDW is consistent with HSR&D funding priority C-Healthcare Informatics. Innovation: Together with experts from COINs around the country and the support of operational partners from each clinical specialty, Pharmacy Benefits Management and VA Informatics and Computing Infrastructure, Dr. Mull proposes to apply her CDA expertise to build a surveillance system to identify invasive non-surgical procedures with preventable adverse events; these procedures are not subject to any VA surveillance activities. A second gap this work addresses is the lack of a nationally available dataset capturing procedural anesthesia use. We will use chart review and text-query data mining methods to obtain this information. The culmination of our IIR work will be a comprehensive database of adverse events and potentially modifiable contributing factors, including procedural anesthesia data, available to VA researchers. Specific Aims: 1) develop and validate surveillance models using FY17-20 data; 2) test the surveillance system (apply model coefficients, perform limited chart review on a monthly basis) from FY21-22, and refine the system using additional CDW variables; 3) test hypotheses related to modifiable processes including whether a trained anesthesia provider was involved or patients received inappropriate antibiotics. Methodology: Our sample includes non-surgical invasive procedures defined by expert clinician co- investigators. We will follow the methods outlined in Dr. Mull's CDA work to aggregate patient, procedure, provider and facility data from the CDW. Next, we will review cases to determine whether a preventable event occurred and use chart review data to estimate logistic regression models predicting the likelihood of an adverse event. Model coefficients will be applied on an ongoing fashion to identify cases likely to have an adverse event to target chart review. Surveillance data will be used to test study hypotheses. Next Steps/Implementation: Through this IIR, and in a subsequent partner-funded QUERI proposal, Dr. Mull and her team will establish an adverse event surveillance system designed for invasive non-surgical procedures that can be used to assess modifiable processes of care to prevent patient harm. By identifying risk factors for preventable adverse events, we can determine where we, with our operational partners, should focus QUERI-funded QI work to improve patient safety. Study results will provide much needed information to the research and clinical communities as they continue to measure and improve the quality of VA care.