DermWorld Academy Insider: Augmented and artificial intelligence in dermatology in the age of ChatGPT
Apr. 11, 2023
This spring, Dialogues in Dermatology is publishing a series of three podcasts about augmented and artificial intelligence and its potential role in dermatology, which are available to all members. With the recent advent of ChatGPT, Google Bard, and other AI-enabled chatbots, many dermatologists are wondering how these developments may affect them and the care they’re able to offer to patients.
In this exclusive DermWorld Academy Insider interview, Ivy Lee, MD, FAAD, and Veronica Rotemberg, MD, PhD, FAAD, chair and deputy chair of the AAD’s Augmented Intelligence Committee, offer their insights on the development of AI and augmented intelligence (AuI) in dermatology.
DermWorld: The podcast series addresses the need to identify the best uses of AI and then optimize them—rather than simply focusing on areas where there’s plenty of data to train a machine, like image recognition. What areas of dermatology practice do you see as ripe for helpful intervention by AI as an assistive technology?
Dr. Rotemberg: The AAD’s AuI Committee advocates for human-centered design, development, and deployment of AI that address unmet needs or gaps in care to improve patient care; these technologies also have the potential to improve clinical workflow by helping us triage patients, optimize telehealth, and increase our efficiency with administrative tasks like scheduling, documentation, pre-approvals, and billing. I suspect we’ll see AI really change how practices are managed..
Dr. Lee: I also think these technologies will help us be more efficient at tailoring patient care. A lot of us are feeling pressure to see more patients and the requirements of documentation have really escalated over the past two decades; by automating some of these tasks artificial intelligence may help us focus on the aspects we love about being physicians — caring for our patients. On this front, I think we’ll see major advances in things like care coordination, remote patient monitoring, and virtual check-ins. Further down the road, these technologies may help us customize patient education and coaching materials that are specific to individual patients — materials that are culturally and contextually attuned to a patient’s unique needs and life situation.
DermWorld: What’s needed for AI and AuI to really make a difference in dermatology?
Dr. Lee: For meaningful impact and true digital transformation (and not just digitization), innovations in these technologies must be led by dermatologists. We understand the needs of our patients and the complexity of their diagnoses. The stakes for our patients are incredibly high and mistakes can have incredible impact on a patient’s quality of life. Unlike an AI algorithm that is reasonably capable of suggesting the next cat video, medicine can’t move fast and break things since those “things” are our patients and their trust in us. To answer your question more directly, to get these technologies in a place to really make a difference will require us, as dermatologists, to actively collaborate with stakeholders including patients, developers, regulators, payers, malpractice carriers, and a host of other potential stakeholders who may be affected by these technologies.
Dr. Rotemberg: Getting dermatologists involved while this technology is still in its infancy is critical so we can match the intended use of these technologies to their true clinical need. We need to shape the trajectory of AI development so that these tools can augment our capacity to more efficiently, effectively, ethically, and equitably care for patients with skin diseases.
DermWorld: The podcast series discusses concerns about the training data used to develop new AI-supported technologies inadvertently entrenching current health care disparities. How can we ensure that AI doesn’t exacerbate current health disparities?
Dr. Lee: Minimizing bias and optimizing equity and fairness are essential for creating AI that is trustworthy, generalizable, and usable. Our dermatology colleagues on this podcast series are trailblazers in elevating the concern of representational bias in the data and allocative bias in outcomes to the level of national awareness. Because of their call to action, there has been a collective intent and focus to gather and create diverse datasets. This “equity by design” approach applies an anti-bias and inclusivity lens throughout the AI product life cycle (from data collection to validation and real-world monitoring).
Dr. Rotemberg: I think we also need to be cognizant that addressing health disparities isn’t a once-and-done exercise, we in health care need to be aware of how these biases can creep into these technologies and shift over time. There is significant potential for dermatologists to lead in this area by acknowledging how biases influence care delivery and data available to train AI models, and it’s an active area for advocacy.
DermWorld: How can AI help dermatologists spend more time on the parts of their jobs that they love and that add value to the care they provide — and reduce the time spent on administrative burdens?
Dr. Rotemberg: AI technologies, including generative AI tools that are currently available like ChatGPT, BioGPT, and DocGPT, can augment our clinical efficiency by reducing the administrative burden many of us feel. I think it goes without saying, that all the other stuff we as physicians are expected to do beyond patient care, really contributes to physician burnout; this includes managing our inboxes, documentation, patient and caregiver support, prior authorizations, care coordination, managing our clinics … the list goes on and on. AI’s most immediate impact may be helping us recapture some of that time so we can focus on caring for our patients.
Dr. Lee: I also think it’s important that we don’t confuse what AI can do now, with tasks these technologies may be able to do in the future; I think this is where the marketing hype about these technologies really needs some grounding. We have seen a lot of research recently that makes some impressive claims regarding the diagnostic accuracy of these tools in controlled settings; but I think it’s safe to say that these results look pretty fragile in real world settings. For example, these tools can make wildly illogical errors a trained dermatologist would never make; these tools have to be closely supervised for accuracy and applicability. So yes, AI may be an important asset for administrative tasks, but I’m not sure I would make an appointment with one of these tools just yet.
DermWorld: What studies are needed before dermatologists can trust new AI platforms?
Dr. Lee: Dr. Rotemberg and Dr. Daneshjou’s team have thoughtfully created CLEAR Derm Consensus Guidelines that provide an actionable checklist for assessing AI algorithm reporting. The next research and practice gap to address is the need for benchmarks in data governance, performance including algorithmic and clinical validation, and a codification of ethics standards and practices. Dr. Rotemberg’s team is doing this foundational work right now and we’re hoping these efforts will have wide ranging effects by setting minimum standards for the quality of research that’s being reported in academic journals. This work should also help curate the flow of reporting we’re seeing in popular press that often misinterprets or misrepresents research that requires a trained eye to accurately interpret.
Dr. Rotemberg: Let me also extend an invitation to dermatologists who are interested in AI to join our community! One of our main objectives is to create opportunities for dermatologists and residents who are interested in shaping the development of these technologies.
DermWorld: What should dermatologists who are interested in this topic do to learn more? How can they get involved to help the technology develop in the right way?
Dr. Rotemberg: AAD’s AuI Committee is committed to engaging all dermatologists to help create educational and practice management resources for our members. In addition to the Dialogues in Dermatology podcast series, we are updating our Position Statement on Artificial Intelligence, applying for an AI Bootcamp Symposium for the 2024 Annual Meeting in San Diego, and supplementing the virtual Basic Dermatology Curriculum with AI modules. We are also exploring research and mentorship opportunities sponsored by our AuI Committee. Lastly, we co-host a fun networking event for all dermatologists interested in digital health technologies.
Dr. Lee: I would just close by inviting our colleagues to join us on this exciting journey. We hope to shape the evolution of AI in dermatology but to do that, we need input from across the Academy. One of the critical challenges we as dermatologists face is balancing the hype around AI, particularly in regard to consumer facing apps, with the reality of treating complex medical conditions in real live people. Dermatology is on the frontline of this revolution, and while that might be unsettling for some, it also gives us a unique opportunity to lead the development of these potentially transformative technologies.
