Generative Al in health care:
From proof of concept to transformational impact 医疗保健中的生成性 Al：从概念验证到变革性影响
Bill Fera, MD, principal, Deloitte Consulting LLP 德勤咨询公司负责人、医学博士 Bill Fera
Michael Giannopoulos, director of Federal Healthcare/Global Healthcare, CTO & CISO, Dell Technologies Michael Giannopoulos，戴尔技术公司联邦医疗保健/全球医疗保健总监、首席技术官兼首席信息安全官
Vega Shah, PhD, Healthcare & Life Sciences PMM lead, NVIDIA Vega Shah 博士，英伟达™（NVIDIA®）医疗保健与生命科学 PMM 负责人
A recent report from the NEJM Catalyst Insights Council indicates a widespread consensus that generative artificial intelligence (GenAI) will play an inevitable role in health care, with the potential to reshape the field in the coming years. NEJM Catalyst Insights 委员会最近发布的一份报告显示，人们普遍认为，生成式人工智能（GenAI）将在医疗保健领域发挥不可避免的作用，并有可能在未来几年重塑该领域的格局。
According to the survey, 67%67 \% of respondents anticipate a moderate or significant impact from the technology within two years, a figure that rises to nearly 90%90 \% beyond that time frame. The survey, cosponsored by Deloitte, gathered insights from 683 health care executives, clinical leaders, and clinicians worldwide. 调查显示， 67%67 \% 的受访者预计该技术将在两年内产生中等或重大影响，两年后这一数字将上升到近 90%90 \% 。这项调查由德勤公司联合发起，收集了全球 683 名医疗保健高管、临床领导者和临床医生的意见。
The health care sector has traditionally adopted new technologies at a different pace compared to other industries, and GenAl is no exception. While many health care organizations are exploring use cases, the level of GenAl maturity varies, and widespread implementation remains limited. However, this is expected to evolve as organizations gain experience and enhance their technical capabilities in artificial intelligence. Technologies including natural language processing and machine learning have already been integrated into many health care organizations, and GenAl 与其他行业相比，医疗行业采用新技术的速度历来不同，GenAl 也不例外。虽然许多医疗机构都在探索使用案例，但 GenAl 的成熟度参差不齐，广泛实施仍然有限。不过，随着各机构在人工智能方面积累经验并提高技术能力，这种情况有望得到改善。包括自然语言处理和机器学习在内的技术已经被许多医疗机构所采用，而 GenAl
will likely go beyond those capabilities for chart summarizations, real-time language translation, insight surfacing, automated message replies, reasoning, triaging, and managing unstructured or unlabeled data. Some health systems are adopting GenAl tools reactively and opportunistically, but many lack a clear, comprehensive strategy. Only 10% of NEJM Catalyst survey respondents consider their organization’s GenAl capabilities to be advanced. About 40%40 \% of respondents said their organizations have at least a few use cases in place, and 6% said they have not explored GenAl and have no plans to do so over the next year. 在图表摘要、实时语言翻译、洞察力浮现、自动信息回复、推理、分流和管理非结构化或无标记数据等方面，GenAl 工具的功能可能会超出这些功能。一些医疗系统正在被动地、随机地采用 GenAl 工具，但许多医疗系统缺乏明确、全面的战略。在 NEJM Catalyst 调查的受访者中，只有 10% 认为其组织的 GenAl 能力是先进的。约 40%40 \% 的受访者表示，他们的机构至少有几个用例，6%的受访者表示他们还没有探索过 GenAl，也没有计划在未来一年内这样做。
The survey results also indicate an interesting difference in perception of GenAl’s impact on patient experience and diagnostics in the United States compared to organizations outside of the United States-81% of non-US respondents expect it will lead to more personalized care and enhanced diagnostic processes, compared to about 60% of US respondents. 调查结果还表明，美国与美国以外的组织在看待 GenAl 对患者体验和诊断的影响方面存在有趣的差异--81% 的非美国受访者预计 GenAl 将带来更个性化的护理和更先进的诊断流程，而美国受访者的这一比例约为 60%。
Another trend to follow will be the balance of application-based and/or vendor-based tools and the development of internally built models. Cost of commercial large language models versus open-source models that can be trained and customized for specific use cases, as well as fit-for-purpose medium language models and small language models, will be an important influencer of this trend as will the ability to develop necessary skills and capabilities for internal build and deployment. 另一个趋势是基于应用程序和/或供应商的工具与内部构建模型的开发之间的平衡。商用大型语言模型的成本与可针对特定用例进行培训和定制的开源模型的成本，以及适用的中型语言模型和小型语言模型的成本，将是这一趋势的重要影响因素，而为内部构建和部署开发必要技能和能力的能力也将是这一趋势的重要影响因素。
Figure 1. Benefits and challenges of organization’s GenAl use 图 1.组织使用 GenAl 的好处和挑战
To what extent do you agree or disagree with the following statements about the use of GenAl in your organization? 您在多大程度上同意或不同意以下关于贵组织使用 GenAl 的说法？
Gen Al is essential for the future competitiveness of our health care organization Gen Al 对我们医疗机构未来的竞争力至关重要
Gen Al will improve the overall efficiency of our health care operations Gen Al 将提高我们医疗保健业务的整体效率
GenAl can assist in personalizing patient care more efficiently than current methods 与目前的方法相比，GenAl 可以更有效地帮助实现对患者的个性化护理
Governance is the foundation 治理是基础
Many health systems have been able to train an Al model and develop use cases. The challenge is often in bringing that model and operating paradigm to scale. Without the appropriate infrastructure, health care organizations might be unable to integrate a successful use case into a workflow. As 许多医疗系统已经能够培训 Al 模型并开发使用案例。挑战往往在于如何将这种模式和操作范例推广开来。如果没有适当的基础设施，医疗机构可能无法将成功的使用案例整合到工作流程中。因为
a result, many use cases fail to progress beyond the proof-of-concept stage. To ensure the success of GenAl applications, organizations should focus on data management, cybersecurity, infrastructure support, governance, and bias mitigation. A clear governance framework is essential 因此，许多用例未能超越概念验证阶段。为确保 GenAl 应用的成功，企业应重点关注数据管理、网络安全、基础设施支持、治理和减少偏差。明确的管理框架至关重要
for health systems to scale Al across their operations. However, the NEJM Catalyst survey shows that many hospitals are still in the early stages of this process, with only 18%18 \% of NEJM Catalyst survey respondents reporting the adoption of GenAl governance standards. 医疗系统在其运营中推广 Al。然而，NEJM Catalyst 调查显示，许多医院仍处于这一进程的早期阶段，只有 18%18 \% 的 NEJM Catalyst 调查受访者报告采用了 GenAl 治理标准。
Figure 2. Organization’s adoption of governance standards for GenAI 图 2.组织采用 GenAI 治理标准的情况
Has your organization adopted governance standards for GenAl? 贵组织是否采用了 GenAl 治理标准？
Global (%)全球 (%)
US only (%)仅美国（%）
Outside US (%)美国以外 (%)
是 否，但正在进行
Yes
No, but in process
Yes
No, but in process| Yes |
| :--- |
| No, but in process |
14
18
9
35
34
37
No, and no plans to do so 没有，也不打算这样做
25
19
33
Don't know不知道
25
29
21
Global (%) US only (%) Outside US (%)
"Yes
No, but in process" 14 18 9
35 34 37
No, and no plans to do so 25 19 33
Don't know 25 29 21| Global (%) | | | US only (%) | | Outside US (%) |
| :--- | :--- | :--- | :--- | :--- | :--- |
| Yes <br> No, but in process | 14 | | 18 | | 9 |
| | 35 | | 34 | | 37 |
| No, and no plans to do so | 25 | | 19 | | 33 |
| Don't know | 25 | | 29 | | 21 |
GenAl governance considerations include interdisciplinary collaboration, fostering innovation, ensuring ethical adoption, and helping with the management of vendor solutions and roadmaps. This framework can be used to set up guardrails that help guide an organization’s responsible use of the technology through methodologies, processes, and accountability (see “Navigating the emergence of Generative Al in health care”). GenAl governance should combine traditional governance constructs (e.g., policy and accountability) with new capabilities such as ethics review, bias testing, and surveillance. The adoption and integration of GenAl can be thought of as GenAl 治理的考虑因素包括跨学科合作、促进创新、确保符合道德规范的采用，以及帮助管理供应商解决方案和路线图。这一框架可用于设置防护栏，通过方法、流程和问责制，帮助指导组织负责任地使用技术（见 "引领医疗保健领域创生 Al 的出现"）。GenAl 治理应将传统的治理结构（如政策和问责制）与伦理审查、偏见测试和监控等新能力相结合。可将 GenAl 的采用和整合视为
a pyramid. Governance is the foundation of the pyramid with the operating model, technology considerations, and strategies relying on that foundation. 一个金字塔。治理是金字塔的基础，而运营模式、技术考虑因素和战略则依托于这一基础。
A sample of important questions that should be addressed through governance: 应通过治理解决的重要问题示例：
Are trustworthy/responsible AI frameworks in place? 是否建立了值得信赖/负责任的人工智能框架？
Are the models unbiased and secure? 模型是否公正、安全？
Is access restricted to appropriate staff only? 是否仅限适当的工作人员访问？
Are the models doing what we want them to do? 模型是否在做我们希望它们做的事情？
What are the clinical and operational validation proof points for model graduation from proof of concept to production? 模型从概念验证到生产的临床和操作验证证明点是什么？
Are we appropriately addressing and ensuring compliance? 我们是否妥善处理并确保合规？
Health systems should assess their AI maturity 医疗系统应评估其人工智能成熟度
Hospital and health system executives may be motivated to adopt GenAl but could face challenges in determining where to begin, managing access, and maintaining progress 医院和医疗系统的高管可能会积极采用 GenAl，但在确定从何处开始、管理访问和保持进展方面可能面临挑战
through the change-management process. Deloitte’s GenAl Maturity Model and other publicly available tools can be used to help organizations take stock of their current 通过变革管理过程。德勤公司的 GenAl 成熟度模型和其他可公开获取的工具可用于帮助组织评估其当前的变革管理进程。
capabilities and help determine where to focus their efforts and investments over the coming years. 能力，并帮助确定未来几年的工作重点和投资方向。
Figure 3. Deloitte’s GenAI Maturity Model 图 3.德勤 GenAI 成熟度模型
We have identified four levels of GenAl maturity based on an organization’s capabilities, as well as where NEJM Catalyst survey respondents place their organizations: 我们根据组织的能力以及 NEJM Catalyst 调查受访者对其组织的定位，确定了 GenAl 成熟度的四个级别：
Foundational (48%): At this stage, governance structures are nonexistent or minimal. GenAl applications are just starting to be explored. 基础阶段（48%）：在这一阶段，管理结构不存在或处于最低水平。GenAl 应用刚刚开始探索。
Evolving (35%): GenAl is in pockets. Use cases are generally limited, siloed, and typically confined to nonclinical areas. At this stage, a governance structure has been established along with processes for evaluating potential risks associated with data and Al investments. 不断进化（35%）：GenAl 处于零散状态。用例一般有限，各自为政，通常局限于非临床领域。在这一阶段，已经建立了治理结构，并制定了评估与数据和 Al 投资相关的潜在风险的流程。
Advanced (10%): Organizations at this stage of maturity are supporting multiple use cases. Advanced organizations are also able to use analytics for metric tracking, insights, and risk identification. A governance structure is well established, widely socialized, and adopted. Proactive measures are in place to continuously monitor GenAl tools that are being deployed across clinical and administrative functions. 高级（10%）：处于这一成熟阶段的组织支持多种用例。高级组织还能够利用分析进行指标跟踪、洞察和风险识别。治理结构已经建立、广泛社会化并被采用。采取积极措施，持续监控在临床和行政职能部门部署的 GenAl 工具。
Innovative (8%): At this stage, GenAl has been seamlessly integrated across the enterprise and is applicable across various use cases. Use of the technology is continuously improved and is being 创新（8%）：在这一阶段，GenAl 已在整个企业实现无缝集成，并适用于各种用例。该技术的使用得到不断改进，并正在
refreshed by new data and functionality. Data and Al governance policies are externally recognized and include dynamic strategies for risk mitigation with advanced monitoring tools widely deployed throughout the organization. 新的数据和功能不断刷新。数据和 Al 治理政策得到外部认可，包括动态的风险缓解战略，并在整个组织内广泛部署先进的监控工具。
How is GenAl being used by hospitals and health systems? 医院和医疗系统如何使用 GenAl？
Deloitte’s 2024 Life Sciences and Health Care GenAl Outlook Survey found that 75%75 \% of responding US health systems have launched GenAl pilots or are actively scaling across the enterprise. Here is a look at how several health systems are using GenAl. 德勤的 2024 年生命科学和医疗保健 GenAl 展望调查发现， 75%75 \% 做出回应的美国医疗系统已经启动了 GenAl 试点或正在企业内部积极推广。以下是几个医疗系统使用 GenAl 的情况。
Patient communications: About 60% of US NEJM Catalyst survey respondents-and 80%80 \% of respondents outside of the United States-agree that GenAl will be useful in personalizing patient care. In terms of current state, 20%20 \% of respondents said their organizations are using chatbots or virtual assistants. 患者沟通：约 60% 的美国 NEJM Catalyst 调查受访者-- 80%80 \% 美国以外的受访者--同意 GenAl 将有助于个性化患者护理。就现状而言， 20%20 \% 的受访者表示他们的机构正在使用聊天机器人或虚拟助手。
The Ottawa Hospital in Canada is testing an Al-powered digital assistant that can provide humanlike interactions with staff, patients, and visitors. ^(1){ }^{1} The Digital Teammate is being piloted to answer preoperative questions to help prepare patients for their surgery and ease anxiety about their procedure. There is also an avatar that can answer questions about the hospital’s new campus (in French or English). In the future, the Digital Teammate could be used to answer additional questions from patients, provide directions, and share potential additional resources. ^(2){ }^{2} 加拿大渥太华医院正在测试一种由 Al 驱动的数字助理，它可以与工作人员、病人和来访者进行类似人类的互动。 ^(1){ }^{1} "数字队友 "正在试用中，它可以回答术前问题，帮助患者为手术做好准备，缓解对手术的焦虑。此外，还有一个头像可以回答有关医院新园区的问题（法语或英语）。将来，"数字队友 "还可以用来回答病人提出的其他问题，提供路线指引，并分享潜在的其他资源。 ^(2){ }^{2}
- NYU Langone Health's Grossman - 纽约大学朗贡医疗中心的格罗斯曼
School of Medicine asked 16 primary care physicians to rate masked Al and human responses to patient queries. The physicians determined the Al-generated messages were comparable to messages sent by clinicians and gave them higher ratings for empathy and communication style. However, the use of complex language could cause challenges for patients who are not completely literate in English, according to the study. ^(3){ }^{3} 医学院要求 16 名初级保健医生对掩盖的 Al 和人类对病人询问的回复进行评分。医生们认为，人工智能生成的信息可与临床医生发送的信息相媲美，并在同理心和沟通方式方面给予了更高的评价。不过，根据这项研究，使用复杂的语言可能会给那些不完全懂英语的患者带来挑战。 ^(3){ }^{3}
Providence, a multi-state health system based in Renton, Washington, has developed a GenAl tool to prioritize incoming messages and support medical assistants who lead the responses. Providence recently deployed the tool to manage the messages for all of its primary care, family medicine, and internal medicine clinics across its seven-state footprint. ^(4){ }^{4} 普罗维登斯是一家位于华盛顿州伦顿的多州医疗系统，它开发了一种 GenAl 工具，用于对收到的信息进行优先排序，并为负责回复的医疗助理提供支持。普罗维登斯最近部署了该工具，用于管理其分布在七个州的所有初级保健、家庭医疗和内科诊所的信息。 ^(4){ }^{4}
Care coordination: Prior to recommending a certain surgical procedure, Al could be used to evaluate a patient’s health data and demographics to predict likely outcomes and to highlight possible complications or adverse events. Some health systems are already using the technology to estimate a patient’s length of stay, rank their risk of developing future medical conditions, and determine the likelihood of readmission or the need for hospice care. ^(5){ }^{5} 护理协调：在推荐某种外科手术之前，Al 可以用来评估病人的健康数据和人口统计数据，以预测可能出现的结果，并突出可能出现的并发症或不良事件。一些医疗系统已经在使用该技术来估算患者的住院时间、对其未来患病风险进行排序，并确定再次入院的可能性或对临终关怀的需求。 ^(5){ }^{5}
Cleveland Clinic is using and developing Al to accelerate the pace of medical research, enhance patient care, and address caregiver and organizational challenges. Current projects include using Al for bed management, medical research, screening for various conditions, and sepsis-risk prediction. ^(6){ }^{6} 克利夫兰诊所正在使用和开发 Al，以加快医学研究的步伐，加强病人护理，解决护理人员和组织面临的挑战。目前的项目包括将 Al 用于病床管理、医学研究、各种疾病筛查和败血症风险预测。 ^(6){ }^{6}
Northwell Health is using GenAl to improve both clinical and nonclinical workflows across its organization, and to enhance care quality and patient experiences. The technology supports efficient allocation of care resources by integrating Al algorithms into customizable clinical workflows and activating clinicians across service lines to deliver timely care interventions and next-best actions. Through an agreement with its software vendor, the health system has access to 13 FDA-cleared AI algorithms for triage, quantification, and care coordination of acute medical cases. In addition, the health system’s internal chat-Al capabilities act as information assistance on a multitude of topics. An enterprise Al governance and review process has also been established to ensure the safe, responsible, and optimized use of AI. ^(7){ }^{7} 诺斯韦尔医疗集团正在使用 GenAl 改善整个组织的临床和非临床工作流程，并提高护理质量和患者体验。该技术通过将 Al 算法集成到可定制的临床工作流程中，并激活各服务线的临床医生及时提供护理干预和下一步最佳行动，从而支持护理资源的有效分配。通过与软件供应商签订协议，该医疗系统可使用 13 种经 FDA 批准的人工智能算法，对急诊病例进行分诊、量化和护理协调。此外，该医疗系统的内部人工智能聊天功能还可就多个主题提供信息帮助。此外，还建立了企业 Al 治理和审查流程，以确保安全、负责和优化地使用人工智能。 ^(7){ }^{7}
Mayo Clinic has deployed 63 internally developed machine-learning algorithms in its practice and has more than 220 others in development. The organization is using these Al tools to identify cancers at early stages, to improve outcomes, and to predict atrial fibrillation before it can cause a stroke. It is also building GenAl tools to unlock information in imaging, genomics, and pathology. Mayo Clinic is also using Al tools that help manage paperwork from referring physicians (on average, more than 31 million unique documents each year). Over the past four years, 梅奥诊所已在实践中部署了 63 种内部开发的机器学习算法，另有 220 多种算法正在开发中。该机构正在使用这些 Al 工具来早期识别癌症，改善治疗效果，并在心房颤动导致中风之前对其进行预测。它还在开发 GenAl 工具，以释放成像、基因组学和病理学方面的信息。梅奥诊所还在使用 Al 工具，帮助管理转诊医生的文书工作（平均每年超过 3100 万份独特的文件）。在过去的四年中
Mayo Clinic’s research teams and clinical investigators have published more than 1,300 peer-reviewed articles on AI. ^(8){ }^{8} 梅奥诊所的研究团队和临床研究人员已发表了 1,300 多篇有关人工智能的同行评审文章。 ^(8){ }^{8}
Clinical documentation: About 20% of NEJM Catalyst survey respondents are using GenAl for clinical documentation. Ambient listening can be used to translate clinicians’ conversations into documented notes in real time. It could help reduce clinician burnout by allowing them to spend more time with patients and less time on documentation, which also improves patient satisfaction. 临床文档：约 20% 的 NEJM Catalyst 调查受访者正在使用 GenAl 进行临床文档记录。环境监听可用于将临床医生的谈话实时转化为文档记录。这有助于减少临床医生的职业倦怠，让他们有更多的时间与患者在一起，减少记录的时间，从而提高患者的满意度。
The Ottawa Hospital is the first Canadian hospital to pilot a solution that uses ambient, conversational GenAI to securely record physician-patient conversations. GenAl is used to convert those conversations into medical notes, which are reviewed and finalized by the physician and then entered into the electronic health record system. By saving physicians time and effort in preparing patient charts, the hospital expects to expand care access for patients and reduce physician burnout. ^(9){ }^{9} 渥太华医院是加拿大第一家试点使用环境对话式 GenAI 安全记录医患对话的医院。GenAl 用于将这些对话转换成医疗笔记，由医生审阅和定稿，然后输入电子健康记录系统。通过节省医生准备病历的时间和精力，医院希望扩大患者的就医范围，减少医生的职业倦怠。 ^(9){ }^{9}
Kaiser Permanente has made a GenAlpowered medical note-taking app available to its physicians, nurses, and other clinicians at 40 hospitals and more than 600 medical offices across its system. The system uses ambient listening technology to improve the speed and accuracy of medical note-taking. ^(10){ }^{10} 凯撒医疗集团（Kaiser Permanente）向其系统内 40 家医院和 600 多个医疗诊所的医生、护士和其他临床医生提供了 GenAlpowered 医疗笔记应用程序。该系统使用环境监听技术来提高医疗笔记的速度和准确性。 ^(10){ }^{10}
Trustworthy GenAl frameworks 值得信赖的 GenAl 框架
The adoption of GenAl could stall if consumers don’t trust it. Deloitte’s 2024 如果消费者不信任 GenAl，其应用可能会停滞不前。德勤的 2024
Consumer Health Care survey found that 30%30 \% of consumers do not trust the information provided by GenAl, and 80% want to know when the technology is used by their doctor to make decisions. Some consumers may be using free, publicly available GenAl tools to gather information or to make health care decisions. Those tools, however, might rely on inaccurate or incomplete information.11 This means it’s important that health care organizations ensure that their patient-facing Al-generated results are both accurate and reliable. 消费者医疗保健调查发现， 30%30 \% 消费者不信任 GenAl 提供的信息，80% 的消费者希望了解医生何时使用该技术做出决定。一些消费者可能会使用免费、公开的 GenAl 工具来收集信息或做出医疗决策。11 这意味着医疗机构必须确保其面向患者的 GenAl 生成的结果既准确又可靠。
GenAl also has the potential to exacerbate mistrust and introduce new skepticism among consumers and other health care stakeholders. For example, if the data used to train Al models is biased or not balanced, the information being generated might not reliably reflect the population being served. In addition, the technology has been shown to hallucinate and generate false information if it hasn’t been trained on an appropriate data set or tuned for context to generate accurate information. ^(12){ }^{12} Such blind spots are important to consider (see “Overcoming. GenAl implementation blind spots in health care”) when developing a GenAl strategy. GenAl 还有可能加剧消费者和其他医疗保健利益相关者之间的不信任，并带来新的怀疑。例如，如果用于训练 Al 模型的数据有偏差或不平衡，生成的信息可能无法可靠地反映所服务的人群。此外，如果该技术未在适当的数据集上进行训练，或未根据上下文进行调整以生成准确的信息，那么它就会产生幻觉并生成错误信息。 ^(12){ }^{12} 在制定 GenAl 战略时，必须考虑到这些盲点（请参阅 "克服医疗保健中的 GenAl 实施盲点"）。
Conclusion结论
While the integration of GenAl in health care still appears to be in its nascent stages, the technology has the potential to democratize knowledge, eliminate administrative burden, increase interoperability, accelerate discovery, and personalize care. In the future, health care organizations could develop novel care models by applying GenAl to genomic data bases, tissue banks, advanced medical devices, and other emerging data sets. 虽然 GenAl 与医疗保健的整合似乎仍处于初级阶段，但该技术具有使知识民主化、消除行政负担、提高互操作性、加速发现和个性化护理的潜力。未来，医疗机构可将 GenAl 应用于基因组数据库、组织库、先进医疗设备和其他新兴数据集，从而开发出新颖的医疗模式。
The NEJM Catalyst Insights Council report shows that many health care organizations are in the foundational or evolving stages of GenAl maturity, but they appear to be seeking a trajectory toward broader adoption and advanced implementation. To help them realize the benefits of GenAl, health systems should prioritize robust governance frameworks to build trust among consumers and stakeholders and accelerate their GenAl journey. NEJM Catalyst Insights Council 的报告显示，许多医疗机构正处于 GenAl 成熟度的基础或发展阶段，但他们似乎正在寻求更广泛的采用和更先进的实施。为了帮助他们实现 GenAl 的优势，医疗系统应优先考虑建立健全的管理框架，以在消费者和利益相关者之间建立信任，并加快 GenAl 的进程。
Special thanks to Steve Davis, Tom Hittinger, Ted Jerse, and Lauren Druz for their efforts on this piece. 特别感谢 Steve Davis、Tom Hittinger、Ted Jerse 和 Lauren Druz 为本稿件付出的努力。
Endnotes尾注
Renae Yao, “NVIDIA works with Deloitte to deploy digital Al agents for healthcare,” NVIDIA blog, October 21, 2024. Renae Yao，"英伟达与德勤合作，为医疗保健部署数字 Al 代理"，英伟达博客，2024 年 10 月 21 日。
The Ottawa Hospital, “The Ottawa Hospital introduces Al-powered Digital Teammate to share information with staff, patients and visitors,” press release, June 13, 2024. 渥太华医院，"渥太华医院引入 Al-powered Digital Teammate，与员工、患者和访客共享信息"，新闻稿，2024 年 6 月 13 日。
NYU Langone Health, “Al tool successfully responds to patient questions in electronic health record,” news release, July 16, 2024. 纽约大学朗贡医疗中心，"Al tool successfully responds to patient questions in electronic health record，"新闻稿，2024 年 7 月 16 日。
Providence, “Generative Al: The next frontier of health care,” September 27, 2023. 普罗维登斯，"Generative Al：医疗保健的下一个前沿"，2023 年 9 月 27 日。
Rachel Curry, "A.I. shows promise of reducing length of hospital stay-a growing healthcare concern, Observer, July 2, 2024. Rachel Curry，"A.I. 显示出缩短住院时间的希望--医疗保健领域日益严重的问题"，《观察家》，2024 年 7 月 2 日。
Naomi Diaz, “Cleveland Clinic leans into Al to stay atop care needs,” Becker’s Health IT, April 2, 2024. Naomi Diaz，"克利夫兰诊所利用 Al 技术满足医疗需求"，《Becker's Health IT》，2024 年 4 月 2 日。
Aidoc Medical LTD, “Northwell Health integrates Aidoc’s enterprise Al platform,” press release, July 18, 2023. Aidoc Medical LTD，"Northwell Health 整合 Aidoc 的企业 Al 平台"，新闻稿，2023 年 7 月 18 日。
Susan Murphy, “Mayo Clinic’s data-driven quest to advance individualized medicine,” Mayo Clinic, October 30, 2024; Mayo Communications staff. 苏珊-墨菲，《梅奥诊所以数据为驱动，探索推进个性化医疗》，梅奥诊所，2024 年 10 月 30 日；梅奥通讯社员工。
Ontario Hospital Association, “Using Al to increase access to care and reduce physician burnout,” accessed February 2024. 安大略省医院协会，"利用 Al 增加医疗服务的可及性，减少医生的职业倦怠"，2024 年 2 月访问。
Heather Landi, “Kaiser Permanente rolls out Abridge’s gen Al clinical tech across 40-hospital system,” Fierce Healthcare, August 14, 2024. Heather Landi，《Kaiser Permanente 在 40 家医院系统中推广 Abridge 的 gen Al 临床技术》，《Fierce Healthcare》，2024 年 8 月 14 日。
Bill Fera et al., “Building and maintaining health care consumers’ trust in generative Al,” Deloitte Insights, June 6, 2024. Bill Fera 等人，"Building and maintaining health care consumers' trust in generative Al"，Deloitte Insights，2024 年 6 月 6 日。
Gyana Swain, “Patients may suffer from hallucinations of Al medical transcription tools,” ClO, October 28, 2024. Gyana Swain，"患者可能会对 Al 医疗转录工具产生幻觉"，ClO，2024 年 10 月 28 日。
Deloitte.德勤
This publication contains general information only, and Deloitte is not, by means of this publication, rendering accounting, business, financial, investment, legal, tax, or other professional advice or services. This publication is not a substitute for such professional advice or services, nor should it be used as a basis for any decision or action that may affect your business. Before making any decision or taking any action that may affect your business, you should consult a qualified professional advisor. 本出版物仅包含一般信息，德勤并不通过本出版物提供会计、商业、金融、投资、法律、税务或其他专业建议或服务。本出版物不能替代此类专业建议或服务，也不应作为可能影响您业务的任何决策或行动的依据。在做出任何可能影响您业务的决定或采取任何可能影响您业务的行动之前，您应咨询合格的专业顾问。
Deloitte shall not be responsible for any loss sustained by any person who relies on this publication. 德勤对任何人因依赖本出版物而遭受的任何损失概不负责。
About Deloitte关于德勤
Deloitte refers to one or more of Deloitte Touche Tohmatsu Limited, a UK private company limited by guarantee (“DTTL”), its network of member firms, and their related entities. DTTL and each of its member firms are legally separate and independent entities. DTTL (also referred to as “Deloitte Global”) does not provide services to clients. In the United States, Deloitte refers to one or more of the US member firms of DTTL, their related entities that operate using the “Deloitte” name in the United States, and their respective affiliates. Certain services may not be available to attest clients under the rules and regulations of public accounting. Please see http://www.deloitte.com/about to learn more about our global network of member firms. 德勤是指英国私营担保有限公司德勤有限公司（"DTTL"）、其成员所网络及其相关实体中的一家或多家。DTTL 及其每个成员公司都是法律上独立的实体。DTTL（又称 "德勤全球"）不向客户提供服务。在美国，德勤是指 DTTL 的一家或多家美国成员所、在美国使用 "德勤 "名称运营的相关实体及其各自的关联公司。根据公共会计的规则和规定，某些服务可能不向验证客户提供。请参阅 http://www.deloitte.com/about，了解有关我们全球成员所网络的更多信息。
