EPtalk by Dr. Jayne 8/14/25

Perplexity made an unsolicited offer to buy the Chrome browser from Google for $34.5 billion. Several people I spoke with agree with the Axios statement that it’s a great marketing play, but unlikely to actually be accepted by Google.

I’ve seen friends and colleagues move away from Google in the months since it added its AI overview feature. I’ve been back and forth with it. I had three significant hallucinations in the same day recently, and all were related to simple fact-based searches that shouldn’t have been problematic. Perplexity claims to have financing in place for the deal, but we’ll likely never know who agreed to back it.

JAMA Network Open has become one of my go-to journals for relevant research that addresses hot topics in healthcare information technology, but at a level that is accessible to more frontline clinicians than might be found in a journal that was targeted towards clinical informaticists. An article this week addressed a great question: “Can a patient portal message with either a physician-created video or an infographic with a physician photograph increase end-of-season influenza vaccination rates?” The study was done at UCLA Health with 22,000 patients from 21 practices. Neither approach raised overall vaccination rates, but both methods increased immunization rates for children and the video message option scored slightly higher.

There’s a lot of vaccine hesitancy in the US, and the Health and Human Service secretary’s recent approval of influenza recommendations received little press coverage. Here’s to hoping that messages from trusted physicians can help drive the needle.

Another feature in the same issue looked at whether physicians made more edits to hospital course summary documents that were generated by large language models (LLM) compared to those generated by physicians. The study was small, looking at only 100 inpatient admissions to the general medical service. The authors found that the percentage of LLM-generated summaries that required edits was smaller than the percentage of physician-generated summaries. The studies were evaluated against a quality standard, with the authors concluding that since the LLM-generated documents needed fewer edits, they were of higher quality than those created by physicians.

I found the study design particularly interesting on this one. The hospital course summaries were randomly assigned to one of 10 internal medicine residents. They had three minutes to review each pair of summaries and edit them for quality purposes. The output of those editing steps was then reviewed and scored for quality by an attending hospitalist physician.

The authors controlled for document length by using a “percentage edited” score and also looked at how much the meaning of the original summary was altered. The authors noted that while the LLM-generated summaries required less editing and may have been “comparably or more complete, concise, and cohesive” they also “contained more confabulations.” They noted that the artificial time constraints may have influenced the result. The study overall supports the idea that using LLMs to help complete this task could be of value.

OpenAI has been trumpeting the release of its GPT-5 model, saying it does a better job with medical questions than its predecessor, but users have been clamoring for an option to return to the previous model. The majority of complaints are around system speed and increased errors. Others took issue with the fact that the new model was rolled out without notice, leading CEO Sam Altman to admit that “suddenly deprecating old models that users depended on in their workflows was a mistake.”

Those of us who have been in the healthcare IT trenches for years understand the value of adequate change management and communication strategies, so I was surprised to learn that the company thought it would be no big deal to just hot-swap the models. If they’re looking for a change management sensei, I might know a girl. Another great quote from Altman: “the autoswitcher broke and was out of commission for a chunk of the day, and the result was GPT-5 seemed way dumber.” Something to ponder for all the folks relying on these technologies. Sounds like they may need a testing advisor as well.

One of my favorite colleagues from residency was in town the other day, doing college visits with one of her children. Her family is going through additional challenges in the college hunt as they evaluate the medical and support resources available to help students manage chronic health conditions in their first few months away from their families. My friend is a brilliant physician who has worked in environments from academic to military to rural health, so she has seen it all.

One of her concerns was the sheer number of communications she receives from her child’s care team: “Seriously, I think I got 15 reminders and a survey, I don’t want to have this kind of a relationship. I already replied, so why are we still having this conversation?” She’s worried that when her child is on her own and receiving all those reminders and messages that they will cause anxiety, which is certainly valid.

Props to health organizations who allow patients to customize reminders and communications. I personally just need one reminder three days out and that’s all. My dentist sends a reminder at 10 days, seven days, three days, one day, and then hourly until you arrive. They claim they can’t adjust it. I’m not sure I’m buying that, but I’m not well versed in dental platforms.

Dr. Nick van Terheyden reached out to let me know that the Lown Institute is accepting nominations for its annual Shkreli Awards, named after notorious “pharma bro” Martin Shkreli. The awards are given “to perpetrators of the ten most egregious examples of profiteering and dysfunction in healthcare.” Previous winners have done such things as: selling the body parts of the deceased without notifying the next of kin, defrauding Medicare by submitting claims on behalf of patients who never received services, and bankrupting community hospitals while living a lavish lifestyle.

What’s the most egregious thing you’ve seen lately in healthcare, regardless of whether it’s worthy of a Shkreli award? Leave a comment or email me.

Email Dr. Jayne.