Artificial intelligence in hospital medicine

Abstract

Artificial intelligence (AI) is potentially the most transformative innovation in society—and medicine—in decades. In the past year, large language models (LLMs) like OpenAI's GPT-4 or Google's BARD have shown themselves capable of tasks that, only a few years ago, seemed firmly in the realm of science fiction. For instance, exhibitions of AI have shown that it can excel at the United States Medical Licensing Examination and the law school graduate bar exam, create original music and art, and generate sophisticated computer code, among numerous other tasks.1, 2 AI has the potential to dramatically augment, if not disrupt, how we care for patients and operate our hospitals. While the healthcare industry has often lagged behind others in adopting and integrating certain technologies, many in medicine seem to be rapidly exploring incorporating this technology into the clinical setting. As hospitalists, we are responsible for exploring and understanding this revolutionary technology. Leaving AI considerations entirely to non-clinicians without us comprehending it or knowing how to properly integrate it into the hospital risks foregoing important benefits and, more importantly, potential harm to our patients, colleagues, and selves. Further, we have an opportunity, if not an obligation, to help decide what activities and outcomes we aim to improve using AI. If we do not have a seat at the table around which decisions about AI in healthcare are made, there is a risk that this technology will be used to drive further value misalignments between healthcare systems, hospitalists, and patients. For this reason, we introduce a special issue of the Journal of Hospital Medicine to explore the impact of AI on hospital medicine. We are honored by these thoughtful, scholarly, and practical deep dives by leaders, innovators, and experts in this emerging space. We begin with a primer by Arora et al. on AI and LLMs to help our readers digest and understand AI technology.3 By explaining terms like Deep Learning, LLMs, and Hallucinations, the authors help demystify the “black box” aspect of this emerging technology. Gleber et al. discuss AI and diagnostic reasoning, shedding light on how AIs “think” about diagnosis and proposing ways in which we can use this information to improve our own diagnostic abilities.4 Rodman and Kanjee explore the potential uses and pitfalls for daily inpatient direct care practice through use-case scenarios.5 Those interested in a pragmatic demonstration of AI and clinical reasoning may want to read the previously published “Conversations on reasoning: LLMs in diagnosis” in which a clinician goes “head-to-head” with GPT-4 to diagnose a patient presenting with low blood pressure and tachycardia.6 Tse and Khan discuss how AI can augment patient communication in the hospital, while Schaye et al. explore the role of AI in medical education.7, 8 Like many new technologies, the use of AI in clinical medicine raises a plethora of ethical quandaries. Prochaska and Alfanre investigate this challenge, particularly around the issues of patient ownership of their data and informed consent in the context of the sheer volume of data needed for AI applications.9 Finally, Ma et al. discuss how AI can be used in quality improvement, patient safety, and hospital medicine research.10 We hope our journal readership finds these perspectives engaging and educational and that this issue stimulates the vital discussions needed as we rapidly and increasingly integrate AI into clinical medicine and our professional lives. In this spirit, we invited ChatGPT to close out the editorial and kick off the series. (For what it's worth, we think GPT-4 did a pretty good job.) “You are a journal editor for the Journal of Hospital Medicine, co-writing an editorial introducing a series on AI for hospitalists. Based on this text, written by your co-authors, please write a succinct and powerful concluding single paragraph that addresses the potential enormous impact of AI and LLMs in hospital medicine, returning to the idea of AI as a ‘revolutionary innovation’ for the field. [INSERT TEXT]” As we conclude this editorial, it is evident that the advent of AI and LLMs heralds a pivotal transformation in hospital medicine, akin to the most disruptive innovations in modern history. These technologies promise to reimagine and reshape our practice, offering unprecedented advancements in clinical reasoning, patient communication, and medical education. As we venture into this uncharted territory, it is imperative for us, as hospitalists, to lead the way—embracing, understanding, and guiding AI's integration into healthcare. This series in the Journal of Hospital Medicine is just the beginning of a profound journey, marking our commitment to harnessing AI as a revolutionary force for the betterment of patient care and the evolution of our field. Zahir Kanjee receives support from the Gordon and Betty Moore Foundation for research on AI and clinical reasoning, was a paid advisory board member for Wolters Kluwer for non-AI-related medical education products, and receives royalties from Wolters Kluwer for books published and from Oakstone Publishing for continuing medical education delivered. The authors declare no conflict of interest.

Ähnliche Arbeiten