EPtalk by Dr. Jayne 1/22/26

The American Board of Preventive Medicine is notifying candidates that they have successfully passed the Clinical Informatics board certification exam. The certifications are retroactive to January 1, 2026. Congratulations to all the new Diplomates, and welcome to yet another continuing certification process that will have you asking yourself why you decided to become double-boarded.

From Straight A Student: “Re: online registration form for a training course that I completed recently. Prompts were in a ‘are there any’ format that asked about mobility restrictions or food allergies. A dropdown choice list appeared to be pre-populated with ‘none.’ My answer was ‘none’ for all of them, so I tried to just submit the form, which popped me back to the top with no feedback. The course vendor responded to my help desk ticket to say that the dropdown requires choosing ‘none’ and people miss that all the time.”

These sorts of Process Improvement 101 issues drive end users batty. The time wasted by users and the help desk adds up.

I wonder if user acceptance testing was done, since it should have been caught. Sometimes teams give the users detailed testing instructions outside of the application, such as “click here, then choose that,” which makes it impossible to determine how they will interact with the workflow. I also wonder if they are analyzing call volume to to identify ongoing issues. Then, has the help desk team reported the issue to development and asked for an update?

It feels like it would be more efficient to change the default to “please select from the list” or “choose a response.” Or, to add a page instruction telling users what to do.

I have been in countless conversations about the safety of healthcare AI solutions. I’m always interested in how the risks and benefits are portrayed to patients and other non-clinical, non-tech individuals. Mr. H mentioned a preliminary report by the VA Office of Inspector General that found that the Veterans Health Administration had some gaps in AI chatbot oversight. The story was also picked up by military-focused Task & Purpose, which ran its own version. 

Risks that were highlighted for the general audience included “producing misinformation, privacy violations, and bias, and that the systems had been put in place without review by the VA’s own patient safety experts.” I didn’t see mention of concerns that were noted by other publications, such as whether lags exist in providing current information for the LLMs to use.

An article commenter shared their physician assistant’s thoughts that “the AI is egregiously wrong 90% of the time, so he doesn’t bother with it.” Based on my own experiences with clinical-focused and consumer-focused AI solutions, that’s probably a significant exaggeration. I wonder if the user would benefit from additional education on prompt construction or effective use of AI tools.

The VA providers who I’ve talked to locally are happy with the AI solutions that are available to them. They are looking forward to continued expansion of their capabilities, such as helping craft more readable medical information for patients. If you’re a VA user, feel free to chime in. We can keep your comments anonymous.

I’m still in my New Year’s inbox cleanup extravaganza, and found an article about Hackensack Meridian Health’s canine-powered cancer detection program. The health system partnered with startup SpotitEarly for a clinical trial that examines the ability of trained dogs to detect cancer via patients’ breath samples. The goal is to validate the technique as a noninvasive cancer detection approach that might be more attractive to patients who are unwilling or unable to complete traditional screening recommendations.

The test is conducted by having patients breathe into a mask-like device for several minutes, followed by the dogs sniffing the devices. The dogs are trained to recognize odor signatures in the exhaled volatile organic compounds that can be associated with cancer. The dogs indicate detection by sitting next to a sample.

We know AI has to be involved somehow, and indeed it is. The company is using AI tools to document and analyze the behaviors of the dogs based on behavioral and physiological data.

SpotitEarly has been in the US market since May 2025, although it was founded in 2020. Previous studies of the technique found that the test was 94% accurate for detecting lung, colorectal, breast, and prostate cancers. If any readers are involved in the study, I’d love to hear about the “best boys” and “good girls” that are doing the sniffing and whether they prefer belly scratches or having their ears rubbed. My medical school had some public-facing research animals and they were the most amazing companions when they retired, resulting in a years-long waitlist for adoption opportunities.

Based on some of the other email traffic in my inbox, quite a few physicians made a New Year’s resolution to look for different employment. Several of them seem to think that informatics is something that you can just jump into because you are “techy” without any formal training or experience.

Some startups will hire clinicians in this situation, but I always encourage people to consider formal coursework to better understand the informatics landscape. I’m a big fan of the courses offered by the American Medical Informatics Association. The virtual courses are convenient, and the in-person ones are great for networking with colleagues working in the field.

A number of highly qualified clinical informatics physicians have recently been displaced from EHR vendors and health systems, so it seems that as long as mergers continue, the job market will remain challenging.

Are you looking to make a career change in 2026, and if so, how are you approaching it? Leave a comment or email me.

Email Dr. Jayne.