A study published in the journal Nature Medicine revealed that general artificial intelligence models like Claude and Gemini were better than dedicated medical tools.

Doximity, Inc. reported positive research results for its AI tool, Doximity Ask, which will now integrate with UpToDate medical content. The developer also expanded its Clinical AI Suite deployment to more than 165 health systems, covering half of the top 20 U.S. hospitals.

絶飲食や飲水制限のある重症患者さんの口渇ケアに悩む看護師向けです。冷水スプレー・口腔清拭・リップクリームの組み合わせは口渇の強さが2.3点低下し、通常ケアの0.6点より大きな低下でした。2025年のメタ分析では氷冷メントール水スプレーが第1位でした。

Hi Erin, Thank you, this is really helpful. I’ve seen UpToDate and it’s truly an amazing tool. I understand why you rely on it several times a day. One gap I’m thinking about is access: UpToDate isn’t free unless you pay out of pocket or have an institutional affiliation, which leaves out many students, independent researchers and patients. My apologies if my earlier examples made it seem like a chatbot for clinical questions. It’s meant more for biomedical research questions: what the published literature says about a topic, how strong it is, and exactly where each claim comes from. It isn’t meant to replace physician-curated guidance like UpToDate at the point of care. On surveying end users, we did some early surveying, but not of physicians. It covered a mixed audience of people whose jobs rely on this research. The goal is to democratize this knowledge, so people without a subscription or institutional access can still see what the evidence says. Your point about understanding each group’s standards of evidence is well taken, and I’d like to go deeper there. Mosisa

Hi Moisa The reason I ask about your intended audience is because exactly this type of resource has already existed for many years for clinicians- it’s called Uptodate. If you know a physician with a subscription, I’d suggest you ask them if you can explore the site. It’s well respected and heavily used. As a family physician, I use it several times during each clinic day. This might be a generational issue (as I’m in my last decade of practice), but I don’t rely on AI as my “go-to” source for clinical questions. When I want an answer to an open-ended medical question of the following type: “What is the evidence for X,” I want to know that _human_ _physicians_ have assessed/summarized the strength of ALL the evidence while also considering a spectrum of possible clinical constraints, therapeutic options, and patient preferences. I usually won’t rely on AI output in this context. I do use AI occasionally through a clinic day, but only for very closed-end/“lower stakes” questions (e.g., What is the monthly cost of drug X in my province?), rather than for the purpose of evidence synthesis or Continuing Medical Education. I suspect that younger physicians might be more gung-ho about AI and willing to allow it to supplant other evidence sources like Uptodate, but I’m not there yet. And I worry _a lot_ about the potential for bandwagon-jumping in younger generations of physicians and how this will affect their clinical acumen and problem-solving ability. As for patients, they already have multiple AI options available to them. Another recent thread in this forum highlighted both positive and negative ways in which patients are currently using AI. I think you’re questioning the factual medical accuracy of existing AI options (?) This is a very reasonable concern, but I suspect that it would be a Herculean feat to garner users for a “bespoke” version of AI (honed for medical accuracy) in today’s hyper-competitive tech climate. In short, I respect your underlying intention with this tool. I just question its chances of actually being _used._ This is a recurring issue in medical research. It’s much easier to _design_ something (e.g., a clinical prediction tool, observational study) than it is to get people to actually use the product. And this is why it’s so vital, before embarking on any time-consuming/potentially costly project/study, to survey potential “end-users” of your work. Without a detailed understanding of how stakeholders make decisions (specifically, a deep appreciation of the _standards of evidence_ required for decision-making by various stakeholders), you won’t be able to anticipate whether your product will have a chance of actually being used by your intended audience. Not sure if any of this is helpful…

Neurodiversity-Affirming Support To Help Autistic People Thrive And Grow. Autistic Advocate, Author & Consultant