与技术一起发展：机器学习是手术室护士改善外科护理的机会

Full text

1 INTRODUCTION

Despite the critical role of the operating room (OR) Registered Nurse (RN) in the patient journey and being one of the first documented specializations in nursing (Hamlin, 2020), the role of the OR RN is poorly defined (McGarry et al., 2018). Most articles overemphasize the technical aspects of the role, such as building a sterile field, gowning surgeons and passing instruments (Mathenge, 2020). Operating room RN are not the only professionals capable of performing these technical skills; surgical technologists and registered practical nurses are cited as being acceptable replacements (Shewchuk, 2007). Other articles attempt to define the OR RN role through ill-defined or ambiguous duties all interdisciplinary surgical team members should practice, such as effective documentation and the promotion of patient safety (von Vogelsang et al., 2020).

Where OR RN appear to have successfully found their niche is in what McGarry et al. (2018) refer to as “maintaining momentum”, described as anticipatory work that ensures surgical procedures run smoothly with minimal interruptions. Maintaining momentum is performed by managing intrinsic and extrinsic factors that can increase surgical time or cause delays (McGarry et al., 2018) that can compound leading to cancelled surgeries, increased costs, and poor patient outcomes like increased wait times and more exposure to anaesthetics (Bellini et al., 2019; Koch et al., 2020). Intrinsic factors are problems that can arise within OR theatres, such as equipment not being ready. Extrinsic factors are problems that can arise outside the OR theatre, such as patients not arriving on time. Managing intrinsic and extrinsic factors is a challenging task as OR nurses report a lack of time and resources being barriers to providing effective care for their patients (Eriksson et al., 2020). Artificial intelligence (AI) technologies and the application of machine learning that can imitate the way humans learn are potential new resources for OR RN that can not only perform technical skills, but also assist in maintaining momentum (Bellini et al., 2019). This commentary will focus on the ways machine learning can be used to augment the anticipatory work of OR RN and then describe future opportunities for surgical nursing to utilize their skills and knowledge to co-create and implement technologies in the OR.

2 MACHINES MAINTAINING MOMENTUM

Machine learning is a form of AI that uses algorithms to analyse massive quantities of data, learn from its mistakes through an iterative process, and create schemes and predictive models (Obermeyer & Emanuel, 2016). Machine learning has already been applied across multiple medical disciplines to analyse medical imaging, speech signals, and to predict myocardial infarctions (Shamout et al., 2021). Relevance to OR settings stems from recent attempts to utilize machine learning in research to optimize OR efficiency and reduce costs associated with human errors (Bellini et al., 2019). An example of machine learning is computer vision, a programme that functions very similar to the watchful circulating nurse, studying surgeons' eye movements, hand patterns and instrumentation use in videos or images (Hashimoto et al., 2019). Computer vision algorithms decode the information in these videos or images to predict the surgeons' actions and list instrumentation needed for each procedural step (Chadebecq et al., 2020; Egert et al., 2020). The rapid proliferation of machine learning has spurred the possibility of creating “Intelligent ORs” that use deep learning to monitor surgeries, detect abnormalities, and provide context-specific real-time support (Tanzi et al., 2020).

Although still in its infancy in OR settings, machine learning has the potential to aid nurses in anticipating the needs of their surgical team and assist in maintaining momentum (Kennedy-Metz et al., 2021). Machine learning could assist in managing the intrinsic factors that cause the most disruption to surgical workflows: managing equipment and OR theatre layouts (Neyens et al., 2019). To someone unfamiliar with the OR environment, layouts and locating equipment may seem trivial; however, something as seemingly innocuous as the OR table facing the wrong direction can lead to a nearly two times increased chance of having surgical flow disruptions (Joseph et al., 2021). These disruptions are important to minimize because they prolong surgical procedures, which means higher costs for hospitals, more exposure to anaesthetics for patients and a higher chance of developing a surgical site infection (Cheng et al., 2017; Koch et al., 2020). From an operational cost perspective, delays need to be minimized because each minute an OR runs costs an average of between $22 and $133, and unexpected cancellations can cost between $215 and $619 per case (Bellini et al., 2019). Intelligent ORs can theoretically address these issues by surveying the room, identifying the current surgical phase, and providing real-time feedback of what equipment will be imminently needed and where it should ideally be placed to maintain momentum. Machine learning is a powerful tool to supplement OR RNs who have human limitations such as having a wide range of experience, educational preparation and, most importantly, not being able to work with the same surgical team daily, making it challenging to anticipate surgical needs like instrumentation requirements and room layouts.

In addition, machine learning could effectively manage the extrinsic factors OR RN have difficulties predicting, such as when to call a patient to come for surgery from home or their unit (McGarry et al., 2018). For many institutions, current practice involves OR managers reading charts, calling into OR theatres and using their own clinical instincts to determine when patients need to be present for their procedure (Jiao et al., 2022). The human variability in these tasks causes a multitude of issues like patients becoming dissatisfied by waiting for extended periods of time and OR managers being unsure how many staff need to stay late on a given day (Jiao et al., 2022). Machine learning programmes are currently being developed that can analyse multiple streams of data including the type of surgical procedure being conducted, the patient's intraoperative vital signs, and intraoperative medications to predict with high accuracy when a procedure will finish in real time (Jiao et al., 2022). This algorithm can be combined with existing machine learning programmes that predict how long a patient will spend in recovery to outline when and where patients will be in a hospital's perioperative unit (Fairley et al., 2019). Removing the individual variability and skill of managing extrinsic factors will allow the OR leadership team to not only improve patient satisfaction, but also potentially reduce staff burnout and consequentially improve retention (Shah et al., 2021).

3 A NEED FOR SURGICAL NURSING TO EVOLVE WITH TECHNOLOGY

ORs are historically areas with specialized equipment and technology. More recently, ORs have seen a dramatic rise in the creation of digital devices, like sensors and videos (Birkhoff et al., 2021), that force RN to constantly adapt to new technologies. Most AI development is led by technical experts (e.g., computer scientists and engineers) that lack insight about the clinical context and nursing care, resulting in medical technology that is not optimized to streamline patient care or improve nursing workflows (Dykes & Chu, 2021). Machine learning compounds this problem because algorithms need to be trained using precisely labelled quantitative or qualitative data until it can accurately predict outcomes with high sensitivity, specificity and accuracy (Sidey-Gibbons & Sidey-Gibbons, 2019). The labelling of surgical data requires expert surgical knowledge and is a “bottleneck” to producing effective machine learning programmes (Bodenstedt et al., 2020). Surgeons have been referenced as being the “key” in training machine learning programmes (Ward et al., 2021), whereas nurses on the other hand appear to be absent from literature. This is a potential reason why there are machine learning programmes that ignore the parts of a patient's perioperative experience that nurses have the expert knowledge to advise on, such as the required number of nurses to safely care for a patient (Fairley et al., 2019). Smart glasses have even been designed for OR RNs that display the current step in a procedure and instruct the nurse which equipment the surgeon will need next, without the apparent inclusion of nurses' input (Nishio et al., 2018). These smart glasses were able to recognize where the operative field was; however, they showed poor performance in recognizing which step of the surgery the team was on. The machine learning programme was not trained to recognize that similar equipment can be needed at multiple phases of the same surgery, knowledge ingrained in all OR RNs.

4 NEW ROLES FOR OR RN IN CO-CREATION

As intraoperative surgical technologies continue to advance and innovate patient care in hospitals, it will become increasingly important to leverage nursing knowledge to improve surgical workflows and patient outcomes. Nurses are well positioned to participate in the creation and implementation of AI systems, and this presents exciting future opportunities for OR RN. First, AI and the use of state-of-the-art technologies in ORs is not new for RN. From the early 2000s, when the da Vinci Surgical System was introduced into ORs and then widely adopted across hospitals in North America, many RN were required to develop new skills to troubleshoot the robot's console (Kang et al., 2016). Second, OR RN can help determine what data should be included in predictive models based on their clinical knowledge and nursing workflow. Their insights about what constitutes relevant and representative data, such as how best to set up a theatre, how many nurses are required for a procedure, and what equipment is needed, are imperative for creating effective machine learning programmes. Finally, OR RN have the clinical knowledge to select appropriate outcomes that would enable the successful implementation of machine learning for all clinical and administrative end users, for example, OR key performance indicator she  that nurse managers are monitoring like annual rates of surgical site infections and associated hospital and surgical costs (Bellini et al., 2019). In the context of this commentary, for nurses to realize these emerging opportunities in AI and technology design, there is a need to support the advancement of informatics and AI in nursing education internationally (Ronquillo et al., 2021). Educational opportunities such as interdisciplinary training alongside other disciplines would also bolster nursing knowledge and prepare nursing for the future (Chu et al., 2022).

5 CONCLUSION

AI and machine learning are playing a key role in advancing surgical nursing, and they present an exciting opportunity for OR RN to embrace and contribute to medical technologies. Machine learning has the potential to be a powerful tool that can augment the anticipatory work of OR RNs and improve clinical outcomes, but more tailoring of these technologies to the nursing workflow, setting, and specificities of patients' surgical experiences can better optimize this novel technology. OR RNs are well suited to support the co-creation and actively engage in AI implementation research that can support the continued professional development and progression of surgical nursing as a profession.

6 IMPLICATIONS FOR NURSING MANAGEMENT

Operating room nurses and nurse managers have the knowledge, skill and judgment required to co-create and implement machine learning programmes that would be effective for the entire surgical team. Nurse managers will be exposed to machine learning in the near future and are encouraged to advocate for the incorporation of programmes in their operating rooms that will not only improve patient outcomes but also potentially reduce staff mental stress and costs.

全文翻译（仅供参考）

1 简介

尽管手术室 (OR) 注册护士 (RN) 在患者旅程中发挥着关键作用，并且是最早记录在案的护理专业之一（Hamlin，  2020 年），但 OR RN 的角色定义不明确（McGarry 等人。,  2018 年）。大多数文章都过分强调了该角色的技术方面，例如建立无菌区域、给外科医生穿衣服和传递器械（Mathenge，  2020 年）。手术室注册护士并不是唯一能够执行这些技术技能的专业人员。外科技术专家和注册实习护士被认为是可接受的替代品（Shewchuk，  2007）。其他文章试图通过定义不明确或模棱两可的职责来定义 OR RN 的角色，所有跨学科外科团队成员都应实践，例如有效的文件记录和促进患者安全（von Vogelsang 等人，  2020 年）。

OR RN 似乎已成功找到自己的利基市场的地方是 McGarry 等人。（2018 年）称为“保持势头”，被描述为确保外科手术顺利进行且中断最少的预期工作。通过管理可能增加手术时间或导致延误（McGarry 等人，  2018 年）的内在和外在因素来保持势头，这些因素可能会导致手术取消、成本增加和患者结果不佳（如等待时间增加和更多地接触麻醉剂） （贝里尼等人，  2019 年；科赫等人，  2020 年）。内在因素是手术室内可能出现的问题，例如设备未准备好。外在因素是手术室以外可能出现的问题，例如患者未按时到达。管理内在和外在因素是一项具有挑战性的任务，因为 OR 护士报告缺乏时间和资源是为患者提供有效护理的障碍（Eriksson 等人，2020 年）。人工智能 (AI) 技术和可以模仿人类学习方式的机器学习应用是 OR RN 的潜在新资源，它们不仅可以执行技术技能，还可以帮助保持势头（Bellini 等人，  2019 年））。本评论将重点介绍机器学习可用于增强 OR RN 的预期工作的方式，然后描述外科护理未来利用他们的技能和知识在 OR 中共同创造和实施技术的机会。

2 机器保持势头

机器学习是一种人工智能形式，它使用算法分析大量数据，通过迭代过程从错误中学习，并创建方案和预测模型（Obermeyer & Emanuel，  2016 年）。机器学习已经应用于多个医学学科，以分析医学成像、语音信号和预测心肌梗塞（Shamout 等人，  2021 年）。与 OR 设置的相关性源于最近尝试在研究中利用机器学习来优化 OR 效率并降低与人为错误相关的成本（Bellini 等人，  2019）。机器学习的一个例子是计算机视觉，该程序的功能非常类似于警惕的循环护士，研究外科医生在视频或图像中的眼球运动、手部模式和仪器使用（Hashimoto 等人，  2019 年）。计算机视觉算法对这些视频或图像中的信息进行解码，以预测外科医生的动作并列出每个程序步骤所需的仪器（Chadebecq 等人，  2020 年；  Egert 等人， 2020 年）。机器学习的迅速普及激发了创建“智能手术室”的可能性，该手术室使用深度学习来监控手术、检测异常情况并提供特定于上下文的实时支持（Tanzi 等人，  2020 年）。

尽管在手术室环境中仍处于起步阶段，但机器学习有可能帮助护士预测其手术团队的需求并帮助保持势头（Kennedy-Metz 等人，  2021 年）。机器学习可以帮助管理对手术工作流程造成最大破坏的内在因素：管理设备和手术室布局（Neyens 等人，2019 年）。对于不熟悉手术室环境的人来说，布局和定位设备可能看起来微不足道；然而，像手术台面朝错误方向这样看似无害的事情可能会导致手术流量中断的机会增加近两倍（Joseph 等人，  2021 年））。尽量减少这些干扰很重要，因为它们会延长手术时间，这意味着医院的成本更高，患者接触麻醉剂的机会更多，并且发生手术部位感染的机会更高（Cheng 等人，  2017 年；Koch 等人，  2020 年） . 从运营成本的角度来看，延迟需要最小化，因为 OR 运行每分钟的平均成本在 22 美元到 133 美元之间，而意外取消的成本可能在每例 215 美元到 619 美元之间（Bellini 等人，  2019 年））。从理论上讲，智能手术室可以通过检查房间、确定当前的手术阶段并提供实时反馈来解决这些问题，即急需哪些设备以及理想的放置位置以保持动力。机器学习是一种强大的工具，可以补充具有人类局限性的 OR RN，例如拥有广泛的经验、教育准备，最重要的是，不能每天与同一个手术团队合作，这使得预测手术需求变得具有挑战性，例如仪器要求和房间布局。

此外，机器学习可以有效地管理外部因素或 RN 难以预测，例如何时呼叫患者从家里或他们的单位来进行手术（McGarry 等人，  2018 年）。对于许多机构来说，目前的做法包括手术室经理阅读图表、打电话到手术室并使用他们自己的临床直觉来确定患者何时需要在他们的手术中出现（Jiao 等人，  2022 年）。这些任务中的人为可变性会导致许多问题，例如患者因等待较长时间而变得不满意，以及手术室管理人员不确定某一天需要多少工作人员迟到（Jiao 等人，  2022 年））。目前正在开发机器学习程序，它可以分析多个数据流，包括正在进行的手术类型、患者的术中生命体征和术中药物，以高精度预测手术何时会实时完成（Jiao et al. ,  2022 )。该算法可以与现有的机器学习程序相结合，预测患者将花费多长时间恢复，以概述患者将在医院围手术期的时间和地点（Fairley 等人，  2019）。消除个体差异和管理外在因素的技能将使 OR 领导团队不仅可以提高患者满意度，还可以潜在地减少员工倦怠，从而提高保留率（Shah 等人，  2021 年）。

3 外科护理需要随着技术的发展而发展

手术室在历史上是拥有专业设备和技术的领域。最近，OR 的数字设备（例如传感器和视频）的创建出现了显着增长（Birkhoff 等人，  2021 年），这迫使 RN 不断适应新技术。大多数 AI 开发由技术专家（例如计算机科学家和工程师）领导，他们缺乏对临床环境和护理的洞察力，导致医疗技术无法优化以简化患者护理或改善护理工作流程（Dykes & Chu，2021）。机器学习使这个问题更加复杂，因为算法需要使用精确标记的定量或定性数据进行训练，直到它能够以高灵敏度、特异性和准确性准确预测结果（Sidey-Gibbons & Sidey-Gibbons，  2019）。手术数据的标记需要专业的手术知识，是生产有效机器学习程序的“瓶颈”（Bodenstedt 等人，  2020 年）。外科医生被认为是训练机器学习计划的“关键”（Ward 等人，  2021)，而另一方面，护士似乎没有出现在文献中。这就是为什么有些机器学习程序忽略了护士拥有专家知识可以提供建议的患者围手术期经验的一个潜在原因，例如安全护理患者所需的护士人数（Fairley 等人，  2019 年） ）。甚至已经为手术室注册护士设计了智能眼镜，它可以显示手术中的当前步骤并指导护士外科医生下一步需要哪些设备，而没有明显包含护士的意见（Nishio 等人，  2018 年））。这些智能眼镜能够识别手术区域的位置；然而，他们在识别团队正在进行的手术步骤方面表现不佳。机器学习程序没有经过培训，无法识别在同一手术的多个阶段可能需要类似的设备，所有 OR RN 中都根深蒂固的知识。

OR RN 在共同创造中的 4 个新角色

随着术中手术技术不断推进和创新医院的患者护理，利用护理知识来改善手术工作流程和患者结果将变得越来越重要。护士能够很好地参与人工智能系统的创建和实施，这为 OR RN 提供了令人兴奋的未来机会。首先，人工智能和在手术室中使用最先进的技术对 RN 来说并不新鲜。从 2000 年代初期开始，当 da Vinci 手术系统被引入手术室并随后在北美的医院中广泛采用时，许多 RN 需要开发新技能来排除机器人控制台的故障（Kang 等人，  2016）。其次，OR RN 可以根据其临床知识和护理工作流程，帮助确定预测模型中应包含哪些数据。他们对什么构成相关和代表性数据的见解，例如如何最好地建立一个剧院，一个程序需要多少护士，以及需要什么设备，对于创建有效的机器学习程序是必不可少的。最后，OR RN 具有选择适当结果的临床知识，可以为所有临床和行政最终用户成功实施机器学习，例如，护士管理人员正在监控的关键绩效指标，如手术部位感染的年率和相关的医院和手术费用（Bellini 等人，  2019）。在本评论的背景下，护士要实现人工智能和技术设计中的这些新兴机会，需要支持信息学和人工智能在国际护理教育中的进步（Ronquillo 等人，  2021 年）。教育机会，例如与其他学科一起进行的跨学科培训，也将增强护理知识并为未来的护理做好准备（Chu et al.,  2022）。

5 结论

人工智能和机器学习在推进外科护理方面发挥着关键作用，它们为 OR RN 提供了一个令人兴奋的机会来拥抱和贡献医疗技术。机器学习有可能成为一种强大的工具，可以增强 OR RN 的预期工作并改善临床结果，但将这些技术更多地调整到护理工作流程、环境和患者手术经历的特殊性可以更好地优化这项新技术. OR RN 非常适合支持共同创造并积极参与人工智能实施研究，以支持外科护理作为一个职业的持续专业发展和进步。

6 对护理管理的启示

手术室护士和护士经理拥有共同创建和实施对整个手术团队有效的机器学习程序所需的知识、技能和判断力。护士管理人员将在不久的将来接触机器学习，并鼓励他们提倡在他们的手术室中加入程序，这不仅可以改善患者的治疗效果，还可以减少员工的精神压力和成本。

THE

END

打赏