A Large Language Model-Based Simulation Protocol Using ChatGPT for Training Novice Critical Care Nurses in Family Crisis Communication and De-escalation.

High-stress conversations with family members in distress are a common part of the intensive care unit (ICU) nursing environment. Novice critical care nurses may have limited opportunities to practice these interactions before encountering them in clinical care. Traditional simulation methods can support communication training but often require faculty time, standardized patients, programming, and specialized resources. This technical report presents a structured simulation protocol using ChatGPT (OpenAI OpCo, LLC, San Francisco, CA, USA) to support formative practice in family crisis communication and verbal de-escalation. The protocol uses a fictional ICU scenario in which the trainee communicates with an AI-generated persona. This persona represents a mother in distress at the ICU entrance after visiting hours. The simulated family member is asked to respond dynamically to the learner's communication style, de-escalating when the learner uses empathy, active listening, and clear boundaries and escalating only when the learner uses defensive, dismissive, or overly technical language. At the end of the session, ChatGPT provides formative feedback in six areas of communication. The areas are empathy, active listening, clarity, boundary setting, avoidance of defensive language, and overall de-escalation. This protocol is not intended to replace supervised simulation, standardized patient encounters, clinical mentorship, or real-world communication experience. Rather, it offers a cost-effective, repeatable, and privacy-protected practice that can complement communication skills training for beginning critical care nurses.