Currently accepted at: JMIR Formative Research
Date Submitted: Dec 18, 2025
Open Peer Review Period: Dec 23, 2025 - Feb 17, 2026
Date Accepted: Feb 11, 2026
(closed for review but you can still tweet)
This paper has been accepted and is currently in production.
It will appear shortly on 10.2196/89939
The final accepted version (not copyedited yet) is in this tab.
Prospective Evaluation of Large Language Model Integration into a Classical Hematology Case Conference
ABSTRACT
Background:
Large-language models (LLMs) have emerging applications in clinical decision-making and medical education, but prospective evaluations in hematology are limited.
Objective:
We conducted a prospective feasibility study examining the integration of two LLM-based models into a weekly classical hematology case conference.
Methods:
Over 8 consecutive sessions, ChatGPT and Open Evidence AI were incorporated into real-time case discussions. Presenters used structured prompts to obtain differential diagnoses, diagnostic pathways, guideline-supported management options, and reference retrieval. AI outputs were displayed during the conference and discussed alongside clinical reasoning by hematology faculty. After the intervention, 25 attendees completed a structured survey assessing changes in familiarity and use of AI, perceived value, observed limitations, and preferred implementation strategies.
Results:
Participants included 16 attending hematologists (64%) and 7 trainees (28%). Familiarity with AI increased from 16% “very familiar” prior to the intervention to 36% “a lot of familiarity” afterward. Frequent or occasional AI use increased from 44% to 68%. Most respondents (84%) rated AI as “very” or “somewhat valuable.” AI was most often perceived as helpful for suggesting alternative diagnoses (80%) and retrieving relevant references (92%). Limitations included prompt dependency (60%), insufficient personalization (52%), and occasional irrelevant or incomplete recommendations (52%). Nearly all respondents (92%) favored an adjunctive rather than self-supervised role for AI.
Conclusions:
Prospective integration of LLM tools into a classical hematology challenging cases conference was feasible, increased clinician familiarity and interest, and was perceived as diagnostically and educationally valuable. Future investigations should evaluate accuracy, reliability, and optimal frameworks for structured, supervised AI use in hematology education. Clinical Trial: Not applicable
Citation
Request queued. Please wait while the file is being generated. It may take some time.
Copyright
© The authors. All rights reserved. This is a privileged document currently under peer-review/community review (or an accepted/rejected manuscript). Authors have provided JMIR Publications with an exclusive license to publish this preprint on it's website for review and ahead-of-print citation purposes only. While the final peer-reviewed paper may be licensed under a cc-by license on publication, at this stage authors and publisher expressively prohibit redistribution of this draft paper other than for review purposes.