Bhatt Ami B, Bae Jennifer
Chief Innovation Officer, American College of Cardiology, Associate Professor, Harvard Medical School, Washington, USA.
Executive Director of Global Innovation, American College of Cardiology, Washington, USA.
NPJ Digit Med. 2023 Sep 25;6(1):177. doi: 10.1038/s41746-023-00920-w.
Collaborative intelligence reflects the promise and limits of leveraging artificial intelligence (AI) technologies in clinical care. It involves the use of advanced analytics and computing power with an understanding that humans bear responsibility for the accuracy, completeness and any inherent bias found in the training data. Clinicians benefit from using this technology to address increased complexity and information overload, support continuous care and optimized resource allocation, and to enact efforts to eradicate disparities in health care access and quality. This requires active clinician engagement with the technology, a general understanding of how the machine produced its insight, the limitations of the algorithms, and the need to screen datasets for bias. Importantly, by interacting, the clinician and the analytics will create trust based on the clinician's critical thinking skills leveraged to discern value of machine outputs within clinical context. Utilization of collaborative intelligence should be staged with the level of understanding and evidence. It is particularly well suited to low-complexity non-urgent care and to identifying individuals at rising risk within a population. Clinician involvement in algorithm development and the amassing of evidence to support safety and efficacy will propel adoption. Utilization of collaborative intelligence represents the natural progression of health care innovation, and if thoughtfully constructed and equitably deployed, holds the promise to decrease clinician burden and improve access to care.
协作智能体现了在临床护理中利用人工智能（AI）技术的前景与局限。它涉及运用先进的分析方法和计算能力，同时要明白人类对训练数据中的准确性、完整性以及任何内在偏差负有责任。临床医生可借助这项技术应对日益增加的复杂性和信息过载问题，支持持续护理和优化资源分配，并努力消除医疗保健获取和质量方面的差距。这需要临床医生积极参与这项技术，全面了解机器如何得出其见解、算法的局限性以及筛选数据集以发现偏差的必要性。重要的是，通过互动，临床医生和分析工具将基于临床医生运用批判性思维技能来辨别临床背景下机器输出的价值而建立信任。协作智能的应用应根据理解程度和证据水平分阶段进行。它特别适用于低复杂性的非紧急护理以及识别群体中风险上升的个体。临床医生参与算法开发并积累支持安全性和有效性的证据将推动其应用。协作智能的应用代表了医疗保健创新的自然发展趋势，如果精心构建并公平部署，有望减轻临床医生的负担并改善医疗服务的可及性。
