Will NEJM AI Be the Change American Medicine Needs?

"This journal is trying to be one voice of quality and sanity in this huge gold rush that’s going to be pretty overwhelming for the next few years. And I really do believe [artificial intelligence] is going to change medicine—it could be for the bad, but I want to make sure I do my best make sure it’s for the good, it’s solid."

That’s how Isaac Kohane, MD, PhD, Chair of Biomedical Informatics at Harvard Medical School, describes the latest publishing venture of the venerable NEJM Group—NEJM AI. Kohane is Editor-in-Chief of the new journal.

NEJM AI joins a family of publications that includes NEJM Catalyst and NEJM Evidence, all of which were developed under the imprimatur of the flagship publication—The New England Journal of Medicine and all published by the Massachusetts Medical Society, which does not publicly report revenue generated by these publications.

The NEJM Group announced the new journal launch with an email blast April 7, 2023, a little more than a week after Elon Musk and Steve Wozniak joined a few hundred who issued a call to pause AI over risks to humanity, and closely tracking OpenAI’s boast about the growing list of exams its GPT-4 chatbot was passing.

Asked if the timing of the NEJM AI, which coincided with a very hot AI news cycle, was simply luck of the draw or a planned build off the AI news momentum, NEJM Editor-in-Chief Eric J. Rubin, MD, PhD, who took over as editor-in-chief in September 2019 told BreakingMED that it was coincidental. He said "planning for this started before I took this job, actually. It takes years of planning to get a new journal up, so we certainly weren’t trying to time it to chase a news cycle."

Rubin explained that he was not "an early adopter of the internet, nor an early adopter of email. I was not impressed with either of those things, which shows you a bit about the amount of insight I have into these things … but this feels really different to me. Right away, you can go on and ask it a question, and it can give you an answer. … it is so much easier than looking something up. … I’m speaking as someone who every time I see a patient, I look up things in UpToDate, and it is not an efficient way getting information, and doctors are really busy. Even just from an efficiency standpoint, I think AI is going to change things and it [AI] may be kind of lousy now, but it will be different in a month, and again a month after that. So, my own guess—and I’m a poor guesser, is that AI will really change things for doctors."

BreakingMED pointed out that to be a success, AI will have to overcome an number of obstacles. For example, as a group, physicians have traditionally not been early adapters of technology—of this there is no better example than the continued push back to Electronic Health Records. But Rubin pointed out that AI is already "being used in medicine right now behind the scenes, even if patients aren’t aware of it. The diagnostic tests that are out there are being interpreted increasingly by AI. The billing systems are being used and the reimbursement systems are heavily dependent on AI right now. It’s creeping into everything that we do. So I think it’s actually an important time for clinicians to start learning about it."

Kohane added, "… hundreds of companies have developed and are developing AI widgets artifacts, with various claims being made on how this can improve patient care, improve doctor experience, improve outcomes, help patients, and accelerate research. At the same time, the FDA has approved over 500 AI augmented devices. And yet these FDA approvals say basically what [this widget] purports to do on this patient population. It doesn’t say how well it will do it, it doesn’t say if it will generalize results from one hospital to another hospital, from one country to another country —unlike drugs, which themselves have some problems generalizing, but nonetheless human beings are pretty much more standardized more than health care systems. So you have a widget, which is going to affect doctor behavior, and recommend or assist in decision making, you’re going to have a lot of variation depending on who the patients are, and what the healthcare systems are like. And so in that huge frothy scenario that we’re having is a lot of hype and hope around these AI widgets. But how do we know what’s good? What’s clinical grade?"

That is the gap, Kohane said, that he thinks NEJM AI will fill, likening it to Consumer Reports. "It will tell you which one has the best rating based on standardized mechanisms, saying these claims are based on trials and supported by the data. That’s the primary mission of the journal: to be able to stand-up and say based on trials, based on experiences, what’s good and what’s clinically worthwhile. What’s good for the patient and good for the doctor."

But while a Consumer Reports rating can be useful when shopping for a dishwasher, how can it be applied to clinical practice? Or healthcare systems? Or healthcare policy?

One doesn’t have to look far to find examples of a Roman candle effect in healthcare innovations—on April 7, the very day of the NEJM AI launch announcement—Forbes reported the bankruptcy filing of Pear Therapeutics, which had received FDA approval prescription apps to help treat substance use disorder and insomnia, apps the company said doctors and patients liked, but insurers refused to cover.

Asked about the payment barrier, Kohane said that right now, payment for AI is "being made at the C suite level, not at the bedside," but he believes that NEJM AI—with a focus on data-driven measures—will give "doctors what they need to take to CFOs and say ’this is what you should be buying, because it is superior or non-inferior and it is clinically better for patients."

The road to success for NEJM AI —just like the road to success for AI itself—will be paved with consensus, and Kohane says his job is to build that consensus by building audience for his new journal.

The official announcement introducing NEJM AI described it this way:

"NEJM AI will be an interdisciplinary journal facilitating dialogue among stakeholders invested in using AI to transform medicine. NEJM AI will intentionally pair ’pre-clinical’ and clinical articles to deliver critical context to both clinicians and non-clinician researchers. The journal will bridge the fast-moving developments in AI, informatics, and technology in medicine with the application of these advancements to clinical practice. NEJM AI will cover the application of AI methodologies and data science to biomedical informatics, connected health, telemedicine, medical images and imaging, personalized medicine, policy and regulation, and the ethical and medicolegal implications of AI."

Kohane said he envisions an audience that will be a mix of physicians in clinical practice, clinical researchers, and—importantly—AI developers.

BreakingMED asked about the subscription cost of the journal, and was told "subscription prices will be in keeping with other NEJM Group products." The current digital-only subscription for NEJM and NEJM products is $166 a year, which is the likely price for the new journal, since it will be published online only.