The editors of the prestigious New England Journal of Medicine have decided to wade into the deep waters of the policy and pragmatic issues around artificial intelligence (AI) and machine learning. The publication’s editors signed their March 30 editorial, entitled simply, “Artificial Intelligence in Medicine,” and in it, while announcing their creation of a journal dedicated to AI, they look at two challenges around how to cover developments in AI—the huge scope of AI innovations and work, and the complexity around commercial sponsorship of or involvement in, such innovative work.
The editors—Andrew L. Beam, Ph.D., Jeffrey M. Drazen, M.D., Isaac S. Kohane, M.D, Ph.D., Tze-Yun Leong, Ph.D., Arjun K. Manrai, Ph.D., and Eric Rubin, M.D., Ph.D., note that AI “has gained recent public prominence with the release of deep-learning models that can generate anything from art to term papers with minimal human intervention.” With regard to medicine specifically, they announced the publication in the March 30 issue of the first articles in their new series, “AI in Medicine,” and further announce that, “Moreover, to further our commitment to this area, we are also announcing the 2024 launch of a new journal, NEJM AI, which aims to provide a forum for high-quality evidence and resource sharing for medical AI along with informed discussions of its potential and limitations.”
The editors note that they face “two new publishing challenges” in creating their new publication around AI. “The first,” they write, “is the breadth of potential AI applications. There is virtually no area in medicine and care delivery that is not already being touched by AI. For example, AI-driven applications are available to capture the dictation of medical notes; many such applications are attempting to synthesize patient interviews and laboratory test results to write notes directly, without clinician intervention. AI is playing an increasing role in health insurance coverage, assisting caregivers in making claims and payors in adjudicating them. We have already seen many published reports that use AI to interpret images — radiographs, histology, and optic fundi. Tools that utilize AI have come into increasing use in analyzing and interpreting large research databases containing information ranging from laboratory findings to clinical data. All these tools offer the potential for increased efficiency and may, perhaps, render insights that are difficult to attain with more traditional data-analysis methods.” That said, they emphasize that AI methods “are not necessarily a panacea; they can be brittle, they may work only in a narrow domain, and they can have built-in biases that disproportionally affect marginalized groups.”
The second issue that they take on in the editorial is that “[E]xpertise in the field of AI and machine learning is closely linked to commercial applications. The underlying technology is rapidly changing and, in many cases, is being produced by companies and academic investigators with financial interests in their products. For a growing class of large-scale AI models, companies that have the necessary resources may be the only ones able to push the frontier of AI systems. Since many such models are not widely available yet, hands-on experience and a detailed understanding of a model’s operating characteristics often rest with only a small handful of model developers,” they note. Yet, they emphasize, “Despite the potential for financial incentives that could create conflicts of interest, a deep understanding of AI and machine learning and their uses in medicine requires the participation of people involved in their development. Thus, in the series of AI articles we are publishing in the Journal and in NEJM AI, we will not restrict authorship and editorial control to persons without relevant financial ties but will follow a policy of transparency and disclosure.”
All of these aspects of AI in medicine make writing about and curating articles on AI challenging; yet at the same time, the NEJM editors conclude, there is no thoughtful option not to cover the subject. That said, they insist, “Because of concerns about both utility and safety, new applications will generally have to adhere to the same standards applied to other medical technologies. This will require a level of rigor in testing similar to that used in other areas of medicine, but it also can present challenges, such as the “dataset shift” that can result when there is a mismatch between the data set with which an AI system was developed and the data on which it is being deployed. This summer, we hope to begin evaluating research studies for NEJM AI that bring careful methodology to understanding how to use AI and machine learning approaches in medicine. And as always, we welcome such studies at the Journal. We are excited to use our resources to encourage high-quality work in AI and to disseminate it with the same standards that we apply to everything we publish.”