AI is accelerating the way healthcare professionals seek and apply medical knowledge. "Instead of searching through thousands of articles on PubMed, EvidenceHunt aggregates different medical sources on one platform," said Olivier Magnin, co-founder of EvidenceHunt, during the masterclass "Data and AI in Healthcare," organized by Dutch Health Hub andHealth~Holland.
EvidenceHunt is an online platform, driven by AI, where numerous medical resources come together and users can quickly arrive at an answer. Founded by medical professionals and scientific experts, the platform is run by Philippe Habets and Olivier Magnin. Their mission is to change the way medical information and research are sought, analyzed and processed.
Reliable sources give fewer source hallucinations
Like ChatGPT and Perplexity put EvidenceHunt from numerous sources to arrive at an answer. Generic Large Language Models (LLMs), however, use virtually the entire Internet, making source hallucinations likely. According to Magnin, "Analyzing the entire Web can provide answers that are nicer than reality or scientifically incorrect at all. Nevertheless, many healthcare professionals use ChatGPT, and that's worrisome."
EvidenceHunt searches only a reliable set of medical sources that have been scientifically validated. Which publications, guidelines or protocols are good and which are not so good? "This way, you never run the risk that, for example, an assertion in a blog that has not been scientifically validated will be incorporated into the output," Magnin states, "What your platform does has to be right and you have to be able to prove that. We ensure that the healthcare professional can quickly and easily check the generated answers for accuracy."
In addition, the models do not train themselves based on users' data and input. "We use a self-developed medical ranking engine that determines the most relevant sources to answer a question. So it's not a generative AI which builds a response on its own."
Artificial intelligence gives healthcare professionals time savings
LLMs can quickly process, analyze and summarize large amounts of text. This saves healthcare professionals valuable time. "Instead of searching thousands of articles on PubMed, or manually sifting through guidelines, protocols and even internal, unpublished research, EvidenceHunt aggregates various medical evidence on one platform. This makes it more quickly accessible to the user," Magnin said.
Get the latest with the click of a button
Science and the market never stand still. "The literature is different tomorrow. As a healthcare professional, it's almost impossible to stay up-to-date," Magnin says. EvidenceHunt makes this easier. "Every day PubMed is re-indexed. So the latest is always included on the platform."
Healthcare organization opportunity
EvidenceHunt has some 40,000 active users worldwide, most of them in the Netherlands. Almost all Dutch hospitals and mental health institutions have professionals using the platform in their daily practice. This also presents opportunities for healthcare organizations. "We are now making the step from individual licenses to team licenses and implementation processes for organizations. An organization can then, for example, do a cluster analysis of the kinds of questions users ask. As an organization, you can capitalize on that by investing in staff training, for example."
What are the prerequisites for a successful implementation?
Many healthcare innovations are coming not off the ground. Magnin: "People or companies want to make a quick pilot and show what is possible, but when they are successful they then run into obstacles. Only then, for example, the lawyer suddenly thinks something of it or there is no budget. It's important to sort that out in advance and think about that strategically and contractually."
Trust from the healthcare professional is also essential. "You have to take the healthcare professional with you in the development. AI is a big change in the way we work. The user has to trust the product and want to work with it." Ownership is important in this regard. At EvidenceHunt, the healthcare professional makes the choices himself. What is relevant to the question I have? "We do not interpret sources in the output of our program. We bring the relevant information together and pour it into the right form so professionals can make choices quickly."
On Wednesday, October 1, the next masterclass 'Data and AI for healthcare' at Jaarbeurs Utrecht. Curious about the experiences and stories of Wouter Kroese of Pacmed and Patrick Leenen of Agentic Alliance? Then sign up by sending an email to Ruud Koolen at ruud.koolen@jaarbeurs.nl, stating your name, position, organization and phone number.
