Just-in-time answers to clinical questions as an opportunity for effective professional development

Background: Clinicians raise about one question for every two patients they see. Research has shown that these questions frequently remain unanswered (1). Seeking and finding accurate answers is associated with improved quality of care. The difficulty of keeping knowledge up to date is expected to increase due to the rapid growth of medical information. Furthermore, delivering just-in-time answers to clinical questions offers a valuable chance to facilitate effective professional development through informal learning.  Generative AI could play a role in helping clinicians answer clinical questions by rapidly processing vast amounts of medical literature and providing concise, relevant and just-in-time personalized information. To improve accuracy of answers, AI solutions could be integrated with curated medical knowledge repositories that are regularly updated and verified. Summary of Work:  RedEMC (meaning Latin American CME Network for its acronym in Spanish, https://redemc.net ) provides online continuing education to healthcare professionals across Latin America, Spain and Portugal. It has a track record of innovation in continuing education with the use of technology (2,3). As a proof of concept, the construction of a domain-specific knowledge base called Galeno, using a generative AI vector database with a retrieval-augmented generation system (RAG), was decided. It was done with support from a team from the Center for Innovation and Entrepreneurship at ORT University, Uruguay. The database was created with the study materials of a series of courses about antimicrobial resistance for infectious disease specialists and microbiologists, with over 1,500 professionals participating in each edition. This narrow scope allows for an easier implementation, while the large potential audience allows for better testing of the pilot.  Summary of Results:  Initial informal assessment of Galeno by domain experts showed promise in accurately responding to queries posed by clinicians, and some adjustments to the system were made based on their feedback.  It was then assessed with two metrics: 1) use of Galeno over time, and 2) perceptions by users of its strengths and limitations, and barriers for its continued use.  200 professionals were randomly selected from the paid registrants to the 2025 edition to the course, thus covering the different professions (mainly infectious disease specialists and microbiologists) and countries (from across Spanish-speaking Latin America and Spain).  One invitation to use and test Galeno was sent to them in early April via email and via whatsapp. 40 professionals (20%) logged into the system at least once after that invitation. An initial peak but no persistent use was seen, neither a request for peers to be added (just one person asked for a colleague to be given access). A second peak was seen after the survey was sent to users and non users in early May, with a total of 51 participants using the system and 245 queries submitted during the two months. Still, no persistent use was seen after the second message. Assessment by users was good, with most of them envisioning using Galeno in the future, and opportunities for improvement were detected.  The primary reasons cited by non-users for not utilizing the system were as follows: They had forgotten they had access to the system. They lacked the time to use it. They did not clearly understand the benefits of the system.  Discussion and Conclusion: Generative AI, when combined with curated and domain-specific datasets, has the potential to create opportunities for informal learning, which accounts for most of the learning by clinicians. Continued engagement with the system needs reinforcement and could also be facilitated by the integration into the workflow of these busy practitioners. Further studies are needed to define its applicability for widespread use. Recommended readings: Del Fiol G, Workman TE, Gorman PN. Clinical questions raised by clinicians at the point of care: a systematic review. JAMA Intern Med. 2014 May;174(5):710-8. doi: 10.1001/jamainternmed.2014.368. PMID: 24663331. Margolis A, López-Arredondo A, García S et al. Social learning in large online audiences of health professionals: Improving dialogue with automated tools. MedEdPublish 2019, 8:55 ( https://doi.org/10.15694/mep.2019.000055.2 ) Margolis A, Balmer JT, Zimmerman A and López-Arredondo A. The Extended Congress: Reimagining scientific meetings after the COVID-19 pandemic. MedEdPublish 2020, 9:128 ( https://doi.org/10.15694/mep.2020.000128.1 )