Sometimes You Just Need to Get Input from a Real Human Being

If you’re not using AI in clinical medicine yet, allow me to strongly recommend you start — soon. Like now. It’s a phenomenal tool for looking things up quickly. I’d call it the future of information retrieval, but that future has already arrived.

OpenEvidence is currently leading the pack, at least around here. Doximity is close behind, with its slick suite of patient communication tools — HIPAA-compliant Dialer, video visits, even free faxing (still sometimes necessary, unbelievably). Doximity’s note-generating feature is increasingly popular as well. Even UpToDate is entering the arena; their AI is in beta, and what I’ve seen so far is genuinely impressive.

And of course, that doesn’t even touch the broader AI ecosystem — ChatGPT, Claude, Gemini — which our patients are already using in droves.

But as I’ve written before, AI has its limits. It is exceptionally good at summarizing consensus guidelines, but less good at navigating the messy gray zones where guidelines don’t quite apply. And as anyone caring for real patients in real hospitals and clinics knows, those gray zones are often where the hardest decisions live.

That’s where input from real people becomes essential. Below are a few existing and emerging strategies we ID doctors use when the question isn’t just “What do the guidelines say?” but rather, “What would you actually do?”

I’m deliberately setting aside the site formerly known as Twitter — and the somewhat less entertaining, differently annoying Bluesky — since those well-worn platforms have already been extensively discussed elsewhere.

A Listserv

Let’s start with the oldest technology of the bunch — a listserv. If that word makes you nostalgic, you know what I’m talking about. If not, think “reply all,” but with thousands of people with similar interests.

The Infectious Diseases Society of America (IDSA) maintains a members-only listserv, and nearly every ID doctor in the country seems to be on it. (It’s not just ID docs; most professional societies have one.) When someone posts a clinical question, responses roll in — bundled into a single daily email rather than flooding your inbox one by one (mercifully).

Posting a truly difficult case there can be quite helpful. With a patient’s permission — and full transparency that the “audience” would be other ID physicians and pharmacists — I once shared the details of his extraordinarily difficult-to-treat infection. Within a day, I had multiple thoughtful responses from clinicians who had faced almost the exact same scenario. The cumulative wisdom of that group was impressive.

The drawbacks are predictable. The interface is clunky. Good luck finding archived discussions on the IDSA website. And not every thread is about patient care. Longtime members will recall the spirited debate over whether our journals should arrive wrapped in plastic — a topic that dominated the listserv for weeks.

And of course, participation is voluntary, which means extroverts dominate the discussions. Which, come to think of it, is true of most human discourse.

TheMednet

This online forum of tricky clinical questions has been around for years. It was started in 2014 by Dr. Nadine Housri, a radiation oncologist, and her brother Samir when they were looking for answers about their father’s diagnosis of prostate cancer. Originally focused on oncology, it now covers multiple specialties.

But it was completely unknown to me until recently, probably because the ID community didn’t launch until 2024. (Thanks to their ID editor, Dr. Patrick Passarelli, for this inside information.) The questions they post, and the responses they receive are both relevant and at times quite helpful, if only to show the diversity of clinical practice when guidelines won’t cut it. Confess it’s surprised me how genuine the questions are to day-to-day practice.

Here are a few examples (login required):

In light of recent measles outbreaks in the US, would you recommend an MMR booster for immunocompetent patients born before 1957?

Do you recommend a prolonged duration of antibiotics and/or suppression for patients without pre-existing hardware who have placement of new hardware after decompression/washout of Staph aureus epidural abscess?

What is your preferred third antimicrobial agent for a patient with treatment-naive pulmonary MAC without cavitary disease and strict contraindications to utilization of rifampin or rifabutin?

Great stuff, right? You can tell these questions are generated by actual ID doctors in practice, which makes them so real. Often there is no “right” answer, but that’s the point! If these emails are getting screened out by your spam filter, you might want to give them permission to enter your inbox for a while and see what you think. And for you AI gurus out there, these are terrific queries to challenge your large language models.

Roon

If the IDSA Listserv represents the early internet and theMednet feels like a curated collection of tough clinical questions, then Roon is something else entirely — a new, physicians-only social network that’s trying to make medical Twitter good again. Only clinicians with verified NPI numbers can join, which immediately changes the tone. No anonymous nasties. No bots. No performative outrage. Just doctors talking to other doctors.

Full disclosure: I know one of the physicians behind Roon (a cardiologist), and have been following the platform with interest since he reached out to me. It’s very much still early days, with a slowly growing community and only a partial launch in most specialties. In other words, it hasn’t fully found its rhythm yet. But just the fact that Dr. Anthony Breu signed up is reason alone to give it a try. He’s the author of some of the most fascinating medical Twitter posts in history, threads so successful that he even landed a piece in the august pages of the New England Journal of Medicine! 

If the Roon people want my advice, here it is: broaden your reach to include clinicians worldwide, expand beyond physicians (pharmacists, nurses, PAs), and look carefully at the graphic capabilities of existing platforms. Seeing images, figures, and videos from clinical studies is what made ID Twitter live, and made it fun.

I’m sure there are other platforms offering similar clinician-to-clinician advice; if so, let me know what you’re using in the comments. AI is already remarkably good at telling us what the guidelines say. But when the guidelines don’t quite apply, I suspect the next step — ironically — may be something rather old-fashioned.

Asking a real human being.