Enhance Arthroplasty Research through Electronic Health Records and Nlp-Enabled Informatics

摘要 全关节置换术(TJA)是最常见、发展最快的外科手术。 国家。尽管程序量很大，但TJA程序和 相关干预措施是有限的。这主要是由于缺乏高质量的数据源和 与手动提取TJA信息相关的后勤困难 电子健康记录(EHR)的非结构化文本。与此同时，电子病历的快速采用 医疗信息技术的进步提供了转变非结构化的潜力 将电子病历记录转换为结构化、编码化的格式，然后可进行分析并与本地和 国家关节置换术登记处和其他机构。 因此，我们建议利用独特的数据资源和自然语言处理 (NLP)技术，以构建用于自动电子病历数据提取和 分析。我们将(1)开发高性能、外部验证和以用户为中心的NLP- 使能的算法，用于从结构化的和 EHR的非结构化文本，(2)在多个EHR平台上外部验证算法，并 医院环境，以及(3)开展以假体预测为重点的示范项目 使用启用NLP的算法收集的数据元素进行联合感染。我们最重要的是 目标是开发有效、开源和可移植的、支持NLP的数据收集和风险 预测工具，并将其广泛传播给区域和国家参与的医院 TJA登记处。 这项研究具有重要意义，因为它利用强大的数据资源和专业知识来解决 TJA迫切需要高质量的数据和准确的预测模型。自动化数据 收集和处理能力将导致EHR的二次使用热潮 增进有关TJA风险因素、医疗质量和患者预后的科学知识。 准确预测人工关节感染的高危患者将指导预防和治疗 治疗决策给TJA患者带来了显著的健康益处。这项研究是 创新是因为TJA特定的生物信息学技术将改变TJA的研究 动力不足的单中心研究到基于多中心注册的大型观察性研究 和临床试验。我们的交付成果有可能对以下方面产生持续的下游影响 未来TJA的研究、实践和政策。

ABSTRACT Total joint arthroplasty (TJA) is the most common and fastest growing surgical procedure in the nation. Despite the high procedure volume, the evidence base for TJA procedures and associated interventions are limited. This is mainly due to lack of high quality data sources and the logistical difficulties associated with manually extracting TJA information from the unstructured text of the Electronic Health Records (EHR). Meanwhile, the rapid adoption of EHR and the advances in health information technology offer the potential to transform unstructured EHR notes into structured, codified format that can then be analyzed and shared with local and national arthroplasty registries and other agencies. We therefore propose to leverage unique data resources and natural language processing (NLP) technologies to build an informatics infrastructure for automated EHR data extraction and analysis. We will (1) develop a high performance, externally validated and user centric NLP- enabled algorithm for extraction of complex TJA-specific data elements from the structured and unstructured text of the EHR, (2) validate the algorithm externally in multiple EHR platforms and hospital settings, and (3) conduct a demonstration project focused on prediction of prosthetic joint infections using data elements collected by the NLP-enabled algorithm. Our overarching goal is to develop valid, open source and portable NLP-enabled data collection and risk prediction tools and disseminate them widely to hospitals participating in regional and national TJA registries. This research is significant as it leverages strong data resources and expertise to tackle the pressing need for high quality data and accurate prediction models in TJA. Automated data collection and processing capabilities will lead to an upsurge in secondary use of EHR to advance scientific knowledge on TJA risk factors, healthcare quality and patient outcomes. Accurate prediction of high risk patients for prosthetic joint infections will guide prevention and treatment decisions resulting in significant health benefits to TJA patients. The research is innovative because TJA-specific bioinformatics technology will shift TJA research from current under-powered, single-center studies to large, multi-center registry-based observational studies and clinical trials. Our deliverables have the potential to exert a sustained downstream effect on future TJA research, practice and policy.