BY ERIKA MCCARTHY, MMS SENIOR EDITOR AND WRITER
Dr. Ashwini Nadkarni
Q&A with Ashwini Nadkarni, MD, Assistant Professor of Psychiatry, Harvard Medical School; Vice Chair of Faculty Enrichment, MGB AMC Department of Psychiatry; Associate Medical Director, Brigham Psychiatric Specialties
What AI system do you use?
I have experimented with a couple of different systems, including Dax Copilot and Abridge. Both offer the opportunity to shift attention from documentation during the visit to patient care, which is, of course, any physician's biggest priority!
How long have you been using AI, and what were the biggest challenges you encountered when you started?
I first became intrigued by AI in 2023. As an Associate Medical Director for our outpatient general psychiatry practice at Brigham and Women's Hospital and now also Vice Chair for Faculty Enrichment at Mass General Brigham Psychiatry, I've had a lot of
interest in understanding how AI can be used to enhance professional fulfillment for physicians. When it comes to AI that relieves documentation burden, the biggest challenge I've experienced is that there's a learning curve in integrating its use
with one's typical clinical workflow.
What are your top three tips for physicians who want to implement AI in their practice but aren’t sure how/where to start?
First, it's helpful to clarify what AI is. Technically, AI has been around for some time. As an example, recommendation systems like Netflix’s are powered primarily by machine learning and predictive algorithms that are classified as AI. The foundation
for large language models such as ChatGPT (e.g., research form) has also been around for several years. Of course, we're seeing widespread adoption only in the last couple of years.
Second, it's important to have a clear sense of what sort of problem you're trying to solve in your clinical practice and pinpoint your pain point. For instance, with respect to tools for ambient AI documentation, the primary challenge is related to time
and cognitive load from documentation burden. But some physicians may find that their bigger problem is workload due to patient portal messages. So, defining your problem and piloting a specific solution for that pain point helps to set the stage
for determining feasibility, acceptability, and impact.
Finally, understanding the risks and benefits of any solutions your practice adopts is critical.
Have you seen benefits, and if so, what are they? What are pitfalls/concerns physicians should look for?
Research on the use of ambient AI to relieve documentation burden shows that physicians can experience some early increase in workload related to reviewing, editing, and correcting AI-generated notes. There's also the risk of overly long notes or filler
content. Patients can also have differing levels of comfort with being recorded. The benefit, however, is the long-term ability to reduce documentation burden.
Any other tips/tools that would be valuable for physicians trying this for the first time to know?
Starting with low-stakes tasks (e.g., inbox replies) before moving to more complex clinical applications is one consideration. Another is joining a user community — recognizing that sharing best practices can help to reduce any workload associated with
adapting to new tools!
For those looking for more information on implementing AI in your practice, enjoy free access to the New England Journal of Medicine’s “A Novel Playbook for Pragmatic Trial Operations to Monitor and Evaluate Ambient Artificial Intelligence in Clinical Practice” — a study introducing a framework and protocols to design, monitor, and deploy ambient AI within routine care.
If you are an MMS member and wish to subscribe to NEJM AI, please enjoy a 50% discount.
