Predictive Analytics in Hemodialysis: Enabling Precision Care for Patient with ESKD

血液透析中的预测分析：为 ESKD 患者提供精准护理

• 批准号: 10598693
• 负责人: Benjamin Alan Goldstein
• 金额: $ 32.2万
• 依托单位: DUKE UNIVERSITY
• 依托单位国家: 美国
• 财政年份: 2020
• 资助国家: 美国
• 起止时间: 2020-07-01 至 2024-04-30
• 项目状态: 已结题
• 来源: https://reporter.nih.gov/project-details/10598693
• 关键词: Address; Advance Care Planning; Caregivers; Clinical; Communication; Data Scientist; Decision Making; Development; Dialysis procedure; Ensure; Ethics; Focus Groups; Future; Goals; Health; Healthcare; Hemodialysis; Hospitalization; Individual; Infection; Internet; Learning; Life Expectancy; Long-Term Care; Machine Learning; Methods; Modeling; Outcome; Output; Palliative Care; Parents; Patient Care; Patient risk; Patients; Performance; Predictive Analytics; Process; Proliferating; Provider; Research; Risk; Risk Factors; Scientist; Specialist; Time; Translating; Transplantation; Trust; Uncertainty; Update; Work; algorithmic bias; base; clinical practice; design; experience; improved; insight; machine learning prediction; model building; model design; mortality; mortality risk; personalized care; predictive modeling; predictive tools; response; secondary outcome; shared decision making; tool

Machine learning (ML) based clinical prediction models (CPMs) have proliferated over the past few years, becoming a central component of healthcare. These tools show great promise in informing both providers and patients of impending health outcomes, ultimately allowing for greater personalization of patient care. In our parent R01 we are developing a ML based mortality prediction model for patients undergoing hemodialysis. The goal of this tool is to predict both short and long term mortality risk in order to promote shared decision making between patients and providers. Importantly, there are a number of ethical concerns inherent in ML based CPMs. These include consideration of ethical performance (i.e., algorithmic bias), ethical usage (i.e., ensuring outputs are properly interpreted) and ethical implementation (i.e., the tool is properly integrated into the clinical workflow. For this current NOSI we propose to explore questions of ethical usage of ML based CPMs. To do this, we will conduct focus group of patients & their caregivers, providers and data scientists. We will address questions of trust in ML based CPMs, understanding of risk and optimal risk communication, interaction with CPMs and AI/ML tools and usage of CPMs to empower patients and promote shared decision-making. We will compare responses across constituencies. We will use our findings to develop practice oriented guidances that target both develops and users of ML based CPMs.

基于机器学习 (ML) 的临床预测模型 (CPM) 在过去激增 几年来，成为医疗保健的核心组成部分。这些工具显示出巨大的前景 告知提供者和患者即将发生的健康结果，最终允许 患者护理更加个性化。在我们的母体 R01 中，我们正在开发基于 ML 的死亡率 血液透析患者的预测模型。该工具的目标是预测短期 和长期死亡风险，以促进患者和患者之间的共同决策 提供商。重要的是，基于机器学习的 CPM 存在许多固有的道德问题。这些 包括考虑道德表现（即算法偏差）、道德使用（即确保 输出得到正确解释）和道德实施（即该工具被正确集成 进入临床工作流程。对于当前的 NOSI，我们建议探讨道德使用问题 基于 ML 的 CPM。为此，我们将对患者及其护理人员进行焦点小组， 提供商和数据科学家。我们将解决基于 ML 的 CPM 的信任问题， 了解风险和最佳风险沟通、与 CPM 和 AI/ML 工具的交互以及 使用 CPM 赋予患者权力并促进共同决策。我们会比较 跨选区的回应。我们将利用我们的发现来制定以实践为导向的指南 面向基于 ML 的 CPM 的开发人员和用户。