Reimaginando la formación de resúmenes de alta a través de la inteligencia artificial

What is the educational challenge? Discharge summary (DS) writing is a core competency for junior physicians, yet persistent deficiencies in the quality, accuracy, and timeliness of these clinical documents are well-documented, with downstream repercussions in patient safety and continuity of care. Existing educational interventions rely heavily on faculty-intensive, small-group teaching models, which limits scalability and long-term sustainability. There is therefore a need to develop novel, more resource-efficient approaches to provide high-quality training in DS writing with individualised feedback.

What are the proposed solutions? We propose a new educational model that integrates artificial intelligence (AI)-generated feedback into a structured DS training programme. As a proof-of-concept, we conducted a small-scale evaluation comparing feedback quality from multiple AI platforms and a human trainer using a standardised rubric. Based on these findings, we designed an asynchronous Coursemology-based e-learning module incorporating customised generative-AI (cGen-AI) to generate draft feedback, with human moderation retained as a safety and quality assurance step. This model is currently in the pre-implementation phase.

What are the potential benefits to a wider global audience? This conceptual human-in-the-loop AI model has the potential to deliver scalable, consistent, and individualised feedback while substantially reducing faculty and logistical workload. By enabling asynchronous practice and standardised assessment, it directly addresses sustainability challenges faced by DS training programmes internationally.

What are the next steps? Full implementation and evaluation including reliability, learner acceptance, and educational impact of this model is being planned for an entire medical student cohort to replace the existing small-group, faculty-facilitated sessions. The success of such cGen-AI approach for DS training can also be extended to other similar domains of medical training in the future.