#209: Augmenting LLMs with Knowledge Seeds for Clinical Question Generation

High-stakes clinical assessments require complex, context-rich items that test applied reasoning but are time-consuming to create. This study presents a learning-centered LLM pipeline that generates multiple-choice clinical items using knowledge seeds from validated item banks. The LLM-generated Multiple Choice (MC) items demonstrated comparable option-level ecological validity and distractor plausibility to expert-created items, with the pipeline facilitating item development and accelerating the refresh cycle to maintain item bank integrity.