EPtalk by Dr. Jayne 3/13/25

It’s rare for me to see patients without having at least some discussion about the cost of care. Patients usually want to know if the medication I’ve prescribed is a generic, or if they’re in a high-deductible health plan, whether it’s going to be cheaper if they use GoodRx or another discount program.

I trained in a place where generic prescribing was not only encouraged but expected, so many of these conversations are fairly straightforward unless I’m having to recommend a second- or third-line medication after others haven’t worked, or if I’m stuck prescribing one that I know tends to have coverage issues. Patients are becoming more financially savvy when it comes to healthcare costs, which is good considering that patients in the US borrowed $74 billion to cover healthcare costs last year. The US leads the world in medical bankruptcies, and according to recent data, nearly a third of patients are “very concerned” about the possibility of medical debt.

EHRs have become better at embedding data about drug pricing, sometimes putting it at the point of prescribing. This can be cool when implemented well, but when implemented poorly, it’s annoying. For example, if most of the medications that I prescribe cost less than $15 because they’re generic, I really don’t want to see warnings telling me that the drug is going to cost $6 or $8 or whatever it is. EHR-embedded data is also less than helpful when it doesn’t take into account things like deductibles or coinsurances or patients who have both primary and secondary insurance coverage. Maybe we can put some AI resources to work making that information more actionable and also more accessible.

The University of Michigan is planning to reach rural patients through the use of AI-powered mobile clinics. The aspirational goal is one where “general practitioners with AI help could make diagnoses, run and interpret tests, and perform procedures like specialists.” The project is in partnership with the Advanced Research Projects Agency for Health (ARPA-H) and would use vehicles “equipped somewhere between a doctor’s office and hospital.” Even without consideration of the AI element, the devil is in the details for something like this. What kind of lab services will be performed? Is it considered a moderate complexity lab? What is the location of service? How will regulators think about a facility that is constantly on the move?

Those interviewed in the article note that AI agents would “coach” physician assistants and nurses to performed more advanced procedures. I’d like to offer a wild solution to help increase the numbers of advanced procedures that are done across the US. How about we allow family physicians (MDs and DOs) to perform the procedures for which they were trained?

I trained in a high-acuity residency program and developed the skills to perform a variety of outpatient surgical procedures, endoscopies, and even C-sections. But there’s not a hospital in a 100-mile radius that would allow me to have privileges to perform any of those unless I’m a member of a residency program’s faculty. It seems that if we could leverage the skills that highly trained physicians are actually mastering during their training, we could help more patients without the expense of developing and implementing AI. Just a thought. But of course, AI is a lot sexier than allowing family physicians to operate at the top of their licensure, so there you have it.

I was excited to learn that progress has been made in having state licensing boards and hospital credentialing offices remove certain questions from their applications that raise the risk that professionals won’t seek mental health services when they are needed. The Dr. Lorna Breen Heroes’ Foundation has been instrumental in promoting this effort to reduce the stigma of mental health issues among licensed clinicians. The foundation honors the memory of Dr. Lorna Breen, a long time emergency physician who died by suicide during the spring of 2020 while serving on the front lines of COVID. Many of us were in very dark places during that time as well as during other points in our careers. Thank you to the foundation for working to make sure that physicians can get help when they need it rather than worrying whether they will lose their licenses.

I missed this article the first time through the inbox, but I was glad to have caught it later. An original investigation that was published in JAMA Network Open looked at “Clinician Experiences With Ambient Scribe Technology to Assist With Documentation Burden and Efficiency.” The study follows roughly four dozen clinicians at the University of Pennsylvania Health System during two months in 2024, when they were using the DAX Copilot AI tool. Interesting tidbits: while use of the tool correlated with improved efficiency, reduced cognitive burden, and improved patient engagement during appointments, there was “mixed feedback regarding the length and quality of ambient scribe-generated notes.” Some users noted high error rates, but it’s unclear if that has any association with clinician subspecialty since there were 17 specialties represented among the 46 participants.

Notes were about 20% longer, which isn’t always a good thing, although sometimes having more details can make the difference long term. Multiple clinicians felt that the time they spent editing the notes balanced out any time savings during the workday. One of the most interesting tidbits, at least for this data nerd, was that net promoter scores (NPS) were all over the map. Thirteen clinicians were promoters, 11 were passive, and 13 were detractors, leading to an overall NPS of zero. It’s certainly not the overwhelming victory that those who are paying for it were likely hoping for.

I’d love to see this kind of project reproduced in single-specialty cohorts, matching for patient complexity and other factors. I’d also like to see it done at more than one academic institution. Another study limitation was the fact that all participants opted in, which certainly doesn’t reflect how I see physicians embracing (or not embracing) new technologies in the real world.

Have you personally used ambient documentation tools, and what did you think of them? Leave a comment or email me.

Email Dr. Jayne.