The evidence on 2024 is already in: this year has been an important one for the forward evolution of artificial intelligence (AI). Throughout this year, we’ve reported on important breakthroughs in the application of AI based on the testing of algorithms (sometimes called “traditional” AI, though for obvious reasons, that tag doesn’t really fit), and generative AI, across U.S. healthcare. What’s clear is that AI is advancing in every area, from non-clinical applications to clinical decision support and process management, to actual diagnostics.
Here are just a few of the developments we’ve covered this year:
Ø As Senior Contributing Editor David Raths reported in November, “The Mount Sinai Health System has opened a center that will combine artificial intelligence with data science and genomics in a location at the center of the campus of the Mount Sinai Hospital in Manhattan. The health system said the Hamilton and Amabel James Center for Artificial Intelligence and Human Health is dedicated to enhancing healthcare delivery through the research, development, and application of innovative AI tools and technologies. The 12-story, 65,000-square-foot facility will initially house approximately 40 principal investigators, alongside 250 graduate students, postdoctoral fellows, computer scientists, and support staff.” And he quoted Eric Nestler, M.D., Ph.D., director of the Friedman Brain Institute and chief scientific officer at Mount Sinai, as stating that, “By integrating AI technology across genomics, imaging, pathology, electronic health records, and beyond, Mount Sinai is revolutionizing doctors’ capacity to diagnose and treat patients, reshaping the future of health care. Mount Sinai has been at the forefront of AI research and development in health care, and now we stand as one of the first medical schools to establish a dedicated AI research center.”
Ø Also in November, Raths reported that “Washington University School of Medicine and BJC Health System, both located in St. Louis, have launched a joint Center for Health AI. A major focus of the center will be using AI to streamline workflows and administrative tasks, making healthcare more efficient. The center is the first major initiative to evolve from the new, long-term affiliation between WashU Medicine and BJC that was finalized earlier this year.” And he quoted David H. Perlmutter, M.D., executive vice chancellor for medical affairs, as stating that “WashU Medicine and BJC are committed to pushing the boundaries of health care innovation to ensure that our caregivers, our patients and the communities we serve benefit from AI technologies.”
Ø And there is every indication that AI investment will only continue to advance rapidly. As Associate Editor Pietje Kobus noted, “SNS Insider reported on November 4 that the healthcare predictive analytics market size is expected to reach $126.15 billion by 2032. This, SNS reported, is driven by an increasing demand for Artificial Intelligence (AI)-powered patient outcomes. Research by SNS indicates that advancements in AI and machine learning (ML) fuel the growth.” According to SNS, the healthcare predictive analytics market size was valued at $14.02 Billion in 2023. “To date, 66 percent of US healthcare organizations currently utilize predictive analytics.”
There have been many, many more developments; but what I found most fascinating was to attend RSNA24 a few weeks ago—the annual meeting of the Oak Brook, Ill.-based Radiological Society of North America, held every year the week after Thanksgiving at Chicago’s McCormick Place. The range of innovations being discussed at this year’s RSNA Conference was impressive and thought-provoking.
Among the numerous speakers who noted where things are headed was Tessa S. Cook, M.D., Ph.D., of the University of Pennsylvania, who spoke on the topic of “Clinical Implementation of LLMs,” said that, “As a cardiovascular radiologist, I spend a lot of my time looking at aortas; and every time I open up a case, I spend ten minutes looking for who the ordering physician is, what they were looking for, etc. Generative AI could really help a lot” in that regard, she told her audience of radiology professionals, noting that a host of small tasks could be automated in order to make radiologists’ workdays more efficient and effective, including categorizing incidental findings, and automatically processing a study, given a particular clinical content.
Cook went on to share with the audience her “wish list” for the use of LLMs and generative AI:
Ø Patient engagement: patients can ask questions about their health and radiology care and instantly get lay-language answers.
Ø Decision support: LLMs can provide guidance to ordering clinicians so they can choose the examination most likely to answer the clinical question.
Ø Intelligent imaging: LLMs can facilitate automatic scheduling and protocoling so patients can get the right exam performed in the correct way at the appropriate site.
Ø EMR summarization: LLMs can provide intelligent search and summarization of a patient’s chart and prior workup.
Ø Custom reporting: LLMs can convert the radiologist’s report into a lay-language version for patients and customized versions for generalists and other-specialized specialists.”
And another speaker in the same session, Dania Daye, M.D., Ph.D., associate professor of radiology at Harvard Medical School and director of the Precision Interventional and Medical Imaging lab in the Division of Vascular and Interventional Radiology at Mass General Brigham, told the audience that the entire process around the ordering and execution of a diagnostic imaging study could be vastly improved through the leveraging of LLMs in the process. “Usually,” she said, “the imaging-care process begins with someone in the clinic entering an order. There is a decision, then a radiology requisition, a radiologist protocol, and then the patient will be prepared, the imaging is performed, the radiologist will prepare and issue a report, and the report is then accessed. LLMs can be performed at every step of this journey.”
In that regard, Daye referenced an article in Radiology entitled “A Context-based Chatbot Surpasses Radiologists and Generic ChatGPT in Following the ACR Appropriateness Guidelines,” in which a study found that Chatbot provided substantial time and cost savings. She cited several other studies in the recent literature, including one that appeared in the October 5, 2023 edition of JAMA Network Open, entitled “Generative Artificial Intelligence for Chest Radiograph Interpretation in the Emergency Department,” in which the GPT-generated reports were found to be equivalent to radiologists in the ED and better than teleradiologists.
Meanwhile, per that, more and more clinical journal studies are exploring what is possible, including one entitled “FDA-Authorized AI/ML Tool for Sepsis Prediction: Development and Validation,” published in NEJM AI in November. That study discusses the question of how accurate the sepsis models created through large-language models actually are. That particular study didn’t examine implementation, and looked at scores initiated after blood cultures are ordered; so it had its limitations. But what’s clear is that the leaders in the development of AI models designed to predict sepsis are advancing the models themselves, with great potential.
What lies ahead?
So it seems clear that advancement in all these areas is accelerating now. Among the advances that we can expect to see in 2025 should include:
Ø Extensive support to physicians and nurses through the creation of “starter” notes and documentation, both in the electronic health record, and for the purposes of communication with patients.
Ø Support for the improvement of clinical workflows across entire departments in patient care organizations.
Ø Increasing sensitivity and accuracy in LLM-based algorithms used to predict the onset of sepsis in inpatients, an absolutely vital area in inpatient care.
Ø Advances in diagnostic imaging care delivery processes, from improvements in clinical decision support supporting ordering physicians, to improved “set-up” information for radiologists, as they prepare to initiate diagnostic imaging studies, to improved communication back and forth between radiologists and ordering physicians.
Ø Related to the above, improved electronic health record summarization for radiologists as they initiate radiological care.
Ø Major progress in the leveraging of LLMs to help radiologists convert the texts of radiology reports into lay language, for patients who have undergone diagnostic imaging studies.
Ø Intensive and extensive work to broadly streamline physician and nurse workflow across many specialties in the inpatient and outpatient care delivery settings.
Ø Improved clinical decision support across all medical specialties.
Ø Improved diagnostics support across many medical specialties.
Interviewing patient care leaders about this area, it is clear to me that 2025 will usher in a whole new level of AI development, one that will leave healthcare meaningfully better off at the end of 2025 than it is now at the end of 2024. The world’s AI’s oyster—and the healthcare field has the right combination of intelligence and expertise to crack open that oyster for the benefit of clinicians, non-clinician administrators, entire patient care enterprises, and patients, families and communities. If there’s one area that looks like all promise in healthcare right now, this is it.
