Strategies for Engineering Reliable Value Sets (SERVS)

项目概要/摘要 大量证据表明，CDS，如果有效使用，可以提高医疗保健质量，安全性， 和有效性。然而，尽管其潜力，CDS可能会导致严重的不良事件，由于错误- 功能协调发展的在我们以前的工作中，我们发现CDS故障的一个重要来源与以下因素有关： 保持准确和一致的价值集的问题。值集是“代码和相应的列表”， 来自NLM托管的标准临床词汇表（如SNOMED CT®，RxNorm， LOINC®和其他），定义临床概念。它们通常用于临床决策 支持和临床质量措施（CQM）来定义复杂的概念。创建和维护 值集本身就具有挑战性，值集错误可能导致CDS和质量错误 测量.我们发现，这些错误是普遍存在的，有各种各样的原因。在 MALDIVES（机器学习驱动的交互式值集增强系统）项目，我们提出 第一次全面研究价值集的创造和维护，小说和创新的发展， 用于改进价值集的可变机器学习和基于本体的方法，以及 新的、开源的工具来帮助作者和用户设定价值。马尔代夫依赖于一个混合的质量 方法，开发新的本体和基于机器学习的方法，以提高价值 集，和新的开源工具，我们预计，它将产生新的理论，创新的临床应用， 机器学习的应用，以及价值集作者和用户的实用工具和流程。这些 这些进展将改善临床决策支持和质量测量，减少警报疲劳和欺诈， 对患者安全和医疗质量的改善表示敬意。

Project Summary/Abstract Significant evidence suggests that CDS, when used effectively, can improve health care quality, safety, and effectiveness. However, despite its potential, CDS can cause significant adverse events due to mal- functions. In our previous work, we identified that a significant source of CDS malfunctions is related to problems with maintaining accurate and consistent value sets. Value sets are “lists of codes and corre- sponding terms, from NLM-hosted standard clinical vocabularies (such as SNOMED CT®, RxNorm, LOINC® and others), that define clinical concepts.” They are commonly used in both clinical decision support and clinical quality measures (CQMs) to define complex concepts. Creating and maintaining value sets is inherently challenging, and value set errors can lead to errors in both CDS and quality measurement. We have found that these errors are widespread and have a variety of causes. In the MALDIVES (Machine Learning-Driven Interactive Value Set Enhancement System) project, we propose the first comprehensive study of value set creation and maintenance, the development of novel and inno- vative machine learning and ontology-based approaches for improving value sets, and the creation of new, open-source tools to help value set authors and users. MALDIVES relies on a mix of qualitative methods, development of novel ontology and machine learning-based methods for improvement of value sets, and new open-source tools, and we anticipate that it will yield new theory, innovations in clinical ap- plications of machine learning, and practical tools and processes for value set authors and users. These advances will improve clinical decision support and quality measurement, reduce alert fatigue and con- tribute to improvements in patient safety and healthcare quality.