Wolters Kluwer Health launched a version of UpToDate supercharged with artificial intelligence as competition in the generative AI healthcare market heats up.
UpToDate Expert AI is a gen-AI-powered version of the clinical decision support solution that's been on the market for 30 years. Wolters Kluwer says it's the next evolution of UpToDate, which has more than 3 million clinician users working at thousands of hospitals. The company aims to meet clinicians' need for the speed and power of gen AI with reliable medical knowledge.
Similar to a chatbot, UpToDate Expert AI provides fast access to gen-AI-powered answers to clinical questions. What makes Wolters Kluwer's solution distinct is that its answers are only drawn from UpToDate’s expert-authored and peer-reviewed content, executives said. Unlike gen AI tools like ChatGPT, UpToDate Expert AI is closed off to the broader web, which, according to the company, eliminates the chance of hallucination.
UpToDate's medical knowledge database is reviewed and developed by 7,600 expert clinicians. UpToDate also has more than 50 in-house physician deputy editors who are part of the team of medical experts responsible for producing and updating the resource.
"We grounded the model exclusively in UpToDate, which meant that we weren't letting internet information creep into our responses," Peter Bonis, M.D., chief medical officer at Wolters Kluwer Health, told Fierce Healthcare.
Wolters Kluwer will have UpToDate Expert AI active for about 250,000 active users this month and will then continue to expand the rollout. Thousands of users participated in the company's beta program over the last few months.
Several tech companies beat Wolters Kluwer to the punch in launching AI-powered clinical decision support tools, and these companies have been raising a lot of money as they rapidly grow. OpenEvidence, founded in 2022, developed an AI-powered medical search engine and gen AI chatbot exclusively for doctors that summarizes and simplifies evidence-based medical information. The startup has raised more than $300 million since its founding.
Pathway, a seven-year-old company, built an AI-powered clinical reference tool that lets doctors ask questions about care guidelines, approved drugs, drug interactions and clinical trials, among other topics. The company was recently acquired by Doximity to bulk up its healthcare AI capabilities.
While it may not have been first to market with an AI tool, UpToDate, as a 30-year-old brand with millions of users, does have a distribution advantage in the healthcare market and is already embedded in clinicians' workflow, along with its recognized, trusted name.
Two years ago, Wolters Kluwer began testing a gen-AI-powered version of UpToDate in an AI lab environment with a select group of customers.
Bonis said the company wanted to be deliberate and cautious in its approach to developing a gen-AI-powered medical resource at the point of care.
Healthcare is a higher-stakes environment, he noted, and AI hallucinations pose significant risks to patients through potential misdiagnoses, incorrect treatments and medication errors due to false or misleading outputs.
"What the literature is telling us is that hallucinations with the foundational models and all sort of generative instances, even the ones that are based on medical content, hallucinations persist, and that in itself is a problem. But the real problem is that clinicians and others don't recognize hallucinations always," Bonis said.
"Another concern is the matter of automaticity, which is that they're so efficient that they can cause clinicians to act without fully considering the context of the recommendation. And then related to that is the 'deskilling' that can occur and there are studies showing that dependency on LLMs for a variety of tests leads study subjects to understand that material less well and not apply the material as well and accurately," he noted.
And there are business model considerations, he added. "If you're selling data, selling ads to pharmaceutical companies, so if your business model is one that is potentially vulnerable to biases from the sponsor, then that's another consideration that's not fully yet understood."
Wolters Kluwer's researchers conducted internal testing to evaluate other AI-based clinical decision support solutions and found they were capable of serious "misfires."
Some AI-based medical reference tools provided clinicians with the following guidance: recommended surgery on the carotid artery for a patient who did not need it; suggested a specific antidepressant known to have a withdrawal reaction could be safely withdrawn cold turkey without tapering off; recommended the wrong starting treatment for a patient with rheumatoid arthritis; recommended drugs known to cause fetal harm without inquiring as to whether the patient might be pregnant; and recommended that an appropriate drug for blood pressure be avoided in a patient who actually needed that particular drug.
"Knowing that you can have such profound misfires, and that at least some of those misfires won't be recognized and end up being implemented in real patients, we thought that we needed to really take our time and to do the very best that we could to deliver a solution that was taking advantage of what the LLMs could offer and the generative AI could offer without having all of these pitfalls," Bonis said.
Wolters Kluwer built explainability and transparency into its AI outputs for trusted and safe adoption, executives said. And the AI-powered UpToDate was designed to promote clinical reasoning and to answer complex, highly specific questions with nuance and clarity, Bonis noted.
Users receive context-rich responses for each question with single-click access to the assumptions driving the answers, the source of the answer in UpToDate and the step-by-step rationale used to generate it.
Including the context and assumptions that mirror an experienced clinician's thought process required "an enormous amount of engineering," Bonis noted. "That was an enormous amount of work between our product managers, our physicians, engineers, UX people—we have to get that piece of it really right. It's an exciting user interface because it's subtle, but so impactful," he said.
Judith Wolfe, M.D., chief medical officer of University Hospitals’ St. John Medical Center and an emergency medicine physician, has been an early user and tester of UpToDate Expert AI.
Wolfe, who considers herself a "non-professional AI enthusiast," said she prefers UpToDate's LLM platform because it's built on a trusted, reliable medical resource.
"With this particular product, you know that it's constrained, the only information you're going to get are information pieces that have been validated. They are the things that I've trusted for years. UpToDate has been my go-to for probably a decade. Knowing that it's valid is critically important," Wolfe told Fierce Healthcare.
The tool, which is integrated with Epic, gives quick, reliable clinical answers, Wolfe noted, and the AI's speed is beneficial in a fast-paced emergency department work environment.
"In the emergency department, we have lots of distractions. There are ambulances coming in. There's noises going off in the corner. There's somebody screaming in the background, 'Hey doc, can you take a look at this EKG?' So there's not a lot of time to really process and get the info that you need. So that's the value," Wolfe said.
Charles Sonday, associate chief medical information officer at St. Luke’s University Health Network, said the AI-powered UpToDate is not just a tech upgrade but "a smarter, faster way to get trusted answers when they’re needed most."
"We’re rolling this out because it aligns with how we deliver care: quick, reliable and grounded in evidence. It’s another step forward in making sure our teams have the right tools at the point of care," Sonday said.