Curbside Consult with Dr. Jayne 10/13/25

Nearly every physician I meet wants to talk about AI scribe solutions, and whether I think they will truly help put the joy back in medicine.

The first thing I think about when confronted with those kinds of questions is how we define joy in medicine. For some of the older physicians I know, that would be represented by the years prior to 1992, when the Evaluation & Management codes came into being. Before that time, medical coding was much vaguer than it is now, using phrases such as “brief, limited, and extended” to describe how physicians should code a visit. That level of detail evolved later in the 1990s, when physicians had to start using rules that felt more like a mathematical exercise in choosing various numbers of elements from different categories than actually caring for patients.

For those who entered practice later, such coding exercises were the norm until the widespread implementation of electronic health records (EHRs), which were in part designed to help free us from those coding burdens. Instead, they brought other burdens, many of which individually might serve to extinguish joy, but that in aggregate, became downright soul-sucking.

Organizations initially implemented EHRs because they thought it was the right thing to do. Along the way, they had to build consensus and achieve buy-in. I think those clinicians had more joy than those who implemented EHRs later, when it was a mandate and there were fewer focused efforts to ensure that systems worked well, or to ensure that users understood what their implementation was supposed to accomplish.

Moving beyond coding concerns, many of us feel that the rise of consumerism in medicine is also somewhat responsible for sending the joy of medicine on the run. I’m not talking about patient engagement and patient empowerment, which are good things. I’m talking about a focus on consumption and an attitude that the customer is always right.

When looking at excessive consumption as a factor, we saw it increase with the rise of insurance premiums, and also with the rise of high-tech medical interventions. I started to hear comments like, “I’m paying a lot for my insurance and I want an MRI to know for sure” even when an MRI was the most expensive test with the least likelihood of actually improving a patient’s outcome. Economic factors aside, there was a point where technology seemed to become a proxy for good care, and where clinicians’ skills, especially those in the realm of physical diagnosis, started waning to some degree.

Patients didn’t want to have their heart murmurs diagnosed by a physician listening to and interpreting a pattern of sounds, which had been the way prior to the invention of ultrasound. Instead, they wanted an echocardiogram so they could know for sure. Parents who previously would have been content with their child’s physician telling them a murmur was “innocent” and would not cause issues instead wanted tests that in turn drove up the cost of care. Clinicians began to over order certain kinds of studies, which resulted in the creation of clinical decision support rules to help them know when tests were indicated and when they weren’t.

A great example is the Ottawa Ankle Rule, which helps rule out clinically significant foot and ankle fractures and avoids unnecessary X-ray studies. Even after explaining it, however, patients still demand films, even though the risk of those films telling me something that I don’t already know is low. And if you are an employed clinician and don’t order the study, you’re likely to generate a patient complaint, which is going to be a problem. You get in the habit of ordering the study “just to be sure” which is not only clinically questionable, but drives up the cost of care.

These things have taken the joy out of medicine, and they are are unlikely to be impacted by AI scribes. I don’t disagree that spending hours documenting makes your job more difficult, and that people don’t like it. But in speaking with physicians who are using AI scribes, I am hearing more stories of late where they’re replacing that documentation time with other clinical tasks rather than truly taking their day back.

One of my colleagues told me last week that he’s still working from home in the evenings, but now he’s using that time to prep charts for the next day and to begin the documentation process for those visits. He wasn’t sure whether that time was showing up in organizational metrics about time spent in the system outside of work since he might not be actively documenting while doing that work. It’s an important point for CMIOs, physician wellness leaders, and other quality folks to look at as they look at how they are reporting on physician behavior before and after implementation of new documentation technologies.

A recent study in JAMA Network Open looked at EHR documentation and improved efficiency for AI scribe users. It found that although there were “reductions in the time spent in the EHR system and time in notes (per appointment),” there were no changes in “after-hours time spent documenting per appointment, mean time to close encounter, mean appointment length, or monthly number of completed office visits.” The study was relatively small and was conducted at a single site over a three-month period in 2024, so it would be interesting to see how it plays out across diverse sites of care or over a longer period following implementation of an AI scribe solution.

We also need a deeper dive into the factors that didn’t change, such as the after-hours work and the time needed to close encounters. Many physicians complain about so-called pajama time when they’re documenting at home in the evening, but if after- hours work didn’t change, do the physicians still perceive that pajama time improved? I would be interested to see some qualitative research overlaid on the quantitative elements to see how those correlate. Are clinicians really satisfied with working the same number of hours from home, or does it just seem different because they’re doing activities other than writing notes?

The authors did note that some subjects “may exhibit an ‘early adopter’ phenotype,” which may have differed from the control group. They also found that measurements of work in the EHR could not differentiate between active work and times when the EHR was open but unused. They also didn’t account for patient-level factors that can influence documentation burden and noted that the study was done at an institution that already had voice-to-text documentation that might have had an influence. I would be interested to hear from others doing similar work if trends show where and how the work shifts when AI scribes enter the room.

Do you think AI scribes are living up to the hype, and will they will truly help put the joy back in medicine? Or are they just the shiniest thing in the room with us now? Leave a comment or email me.

Email Dr. Jayne.