Identifying and reducing stigmatizing language in home healthcare: the ENGAGE study

项目总结/摘要 护士是美国最大的医疗服务提供者。最近的研究发现 护士中普遍存在偏见，最常见的偏见是种族和民族方面。护士 偏见影响治疗决定，从而影响患者的结果。护士的偏见尤其重要， 护士是医疗保健服务主要提供者的环境，例如家庭医疗保健（HHC）， 在美国，护士每年都要到500多万病人家中出诊。种族偏见反映在 医疗文件;在医院环境中，关于黑人患者的临床记录有高达50%的几率更高， 包含侮辱性语言（即，语言的负面特征的患者）比白色患者的 notes.在HHC环境中，我们还发现黑人和西班牙裔患者的临床记录中有高达20%的 包括侮辱性语言的几率高于白色和亚洲患者。重要的是，我们的研究发现， 临床记录中的侮辱性语言与临床医生的消极态度和低质量的 病人护理 自然语言处理（NLP）是一种很有前途的技术，它有可能帮助人们发现 在数百万份HHC护理记录中使用侮辱性语言。与两个最大的供应商合作， 美国的HHC服务（路易斯安那州卫生保健集团和VNS健康，有超过10万名患者 结合每日人口普查），这项研究汇集了一个跨学科的专家团队，在HHC护理，NLP， 和临床决策支持，以建立设计以护士为中心的NLP为基础的系统的第一步， stigmatiziNG language（"engage"）通过以下具体目标。目标1：扩展和完善 适用于HHC的侮辱性语言。目标2：确定最佳的NLP方法， 准确识别地理上分散的HHC机构的临床记录中的侮辱性语言。目的 3：按种族和族裔比较侮辱性语言的流行程度。目标4：开发NLP驱动的 ENGAGE系统减少HHC临床记录中的污名化语言。 实现这些目标将导致参与-一种技术驱动的行为改变干预 这将有助于识别和消除HHC护士之间的种族偏见。

Project Summary/Abstract Nurses are the largest sector of health providers in the United States (US). Recent studies have found widespread biases among nurses, with the most common bias being in the area of race and ethnicity. Nurses' biases affect treatment decisions, thereby affecting patient outcomes. Nurses' biases are especially critical in settings where nurses are the main provider of healthcare services, such as home healthcare (HHC) where nurses visit more than 5 million patients in their homes across the US every year. Racial biases are reflected in medical documentation; in hospital settings, clinical notes about Black patients have up to 50% higher odds of containing stigmatizing language (i.e., language that negatively characterizes patients) than White patients' notes. In the HHC setting, we also found that clinical notes of Black and Hispanic patients had up to 20% higher odds of including stigmatizing language than White and Asian patients. Critically, our studies found that stigmatizing language in the clinical notes is associated with negative clinicians' attitudes and lower quality of patient care. One promising technology—natural language processing (NLP)—has the potential to help uncover stigmatizing language in millions of HHC nursing notes. In collaboration with two of the largest providers of HHC services in the US (Louisiana Health Care Group and VNS Health, with more than 100,000 patients on the combined daily census), this study assembles an interdisciplinary team of experts in HHC nursing, NLP, and clinical decision support to build the first step in designing a nurse-centered NLP-based system to rEduce stigmatiziNG languAGE ("ENGAGE") via the following specific aims. Aim 1: Expand and refine the ontology of stigmatizing language applicable to HHC. Aim 2: Determine the optimal NLP approach to automatically and accurately identify stigmatizing language in the clinical notes of geographically dispersed HHC agencies. Aim 3: Compare the prevalence of stigmatizing language by race and ethnicity. Aim 4: Develop an NLP-driven ENGAGE system to reduce stigmatizing language in HHC clinical notes. Accomplishing these aims will result in ENGAGE- a technology-driven behavior change intervention that will help to identify and eliminate racial biases among HHC nurses.