FW-HTF-RL: Learning systems that enable healthcare workers to perfect safety-critical hospital work

该项目开发了帮助医护人员的新技术，从而使医疗保健更安全、更具成本效益。医疗保健占国内生产总值的18%，但失误每年导致多达25万人死亡和超过5000亿美元的成本。该项目正在重症监护病房和外科病房创建虚拟助理，指导护士和其他卫生保健工作者，使用工作人员的投入和实验结果来指导技术开发。这些虚拟助手在复杂的任务中提供循序渐进的指导和反馈，并监控日常活动，以确保没有遗漏任何步骤。好处包括帮助学习新的活动，减轻压力，检测可能的问题，以及自动化一些提供者现在所做的工作来记录他们正在进行的活动。研究团队正在为医疗保健环境中的虚拟助手开发原型、实验结果和有效性测量，称为“环境智能监测器”(AIMS)。该团队包括斯坦福大学医院、山间医院和斯坦福大学工程、商业和医学院的计算机科学家、电气工程师、临床医生、行为科学家和医疗保健经济学家。研究目标包括为AIMS开发传感器、计算和软件，发现临床工作者如何更喜欢与AIMS进行沟通，以及在工作环境中进行实验，以最大化他们对客户、家庭和临床医生的价值。该项目包括三个周期的研究，涵盖手部卫生、重症监护病房护理包和腹腔镜手术。每个周期包括分析、设计、开发、在正常医院护理中的实施和评估。这项研究沿着四条平行的轨道进行。One Track旨在开发硬件，用于在相机、深度传感器和麦克风等边缘设备上捕获受隐私保护的数据。第二个途径是开发算法，以识别在制定预期的临床床边活动时的成败。第三个途径是获取医护人员的意见，以开发旨在与人互动的首选方式，例如提供口头提示、确认传感器所做的观察，以及自动更新医疗记录。第四个轨道是运行实验，以衡量AIM的临床、行为、操作和经济影响。通过将工程研究与真实世界的部署和临床研究相结合，这项研究将创造在纯粹技术驱动或纯粹观察性研究中无法获得的新知识。该奖项反映了NSF的法定使命，并通过使用基金会的智力优势和更广泛的影响审查标准进行评估，被认为值得支持。

This project develops new technology to assist healthcare workers, thereby making healthcaresafer and more cost-effective. Healthcare is 18% of the Gross Domestic Product, yet errors annuallylead to as many as 250,000 deaths and more than $500 billion in costs. The project is creatingvirtual assistants in intensive care and surgical units, which guide nurses and other healthcareworkers, using worker input and experimental results to guide technology development. Thesevirtual assistants give step-by-step guidance and feedback during complex tasks and monitorroutine activities to make sure no steps are missed. Benefits include assistance in learning newactivities, reducing stress, detecting possible problems, and automating some of the work providersdo now to document their ongoing activities.The research team is developing prototypes, experimental results, and effectivenessmeasurements in healthcare settings for virtual assistants, called “ambient intelligent monitors”(AIMs). The team includes computer scientists, electrical engineers, clinicians, behavioralscientists, and healthcare economists at Stanford hospitals, Intermountain hospitals, andStanford’s Schools of Engineering, Business, and Medicine. Research goals include developingsensors, computing, and software for AIMs, discovering how clinical workers prefer tocommunicate with AIMs, and experimenting with work environments to maximize their value topatients, families, and clinicians. The project includes three cycles of research, covering handhygiene, Intensive Care Unit care bundles, and laparoscopic surgery. Each cycle includes analysis,design, development, implementation in normal hospital care, and evaluation. The research isproceeding along four parallel tracks. One track is designed to develop hardware for capturing privacy-protecteddata on edge devices such as cameras, depth sensors, and microphones. A secondtrack is for developing algorithms that discern success and failure in enacting intended clinicianbedside activity. A third track is for obtaining healthcare worker input to develop preferred waysfor AIMs to interact with people, such as providing verbal prompts, confirming observationsmade by sensors, and automatically updating medical records. A fourth track is for runningexperiments to measure the clinical, behavioral, operational, and economic impacts of the AIMs.By coupling the engineering research with real-world deployment and clinical studies, this researchwill create new knowledge that is unattainable in a purely technology-driven or purelyobservational investigation.This award reflects NSF's statutory mission and has been deemed worthy of support through evaluation using the Foundation's intellectual merit and broader impacts review criteria.