Introduction: Effective communication is crucial in pediatrics, especially for delivering difficult news. Traditional standardized patient (SP) programs are resource-intensive and impractical for just-in-time training. To address this gap, we developed an AI-powered chatbot that simulates emotionally challenging conversations with on-demand availability. This innovation utilizes natural language processing (NLP) to enhance communication skills training for pediatric trainees. Methods: The AI chatbot engages learners in realistic patient and family interactions, allowing practice in critical conversations such as delivering bad news and discussing end-of-life care. Trainees can access the tool flexibly, enabling just-in-time, self-directed rehearsal of high-stakes communication. A pilot study is underway involving pediatric trainees randomized into two groups prior to a standardized patient encounter regarding an 8-year-old drowning victim. Group 1 receives traditional preparation via self-reflection and best practice didactic guidance in end-of-life communication. Group 2 uses the novel AI chatbot, prompted with the drowning scenario. A blinded evaluator scores the trainees using a validated communication evaluation tool. The primary outcome is feasibility of chatbot use. Secondary outcomes include performance scores and self-reported confidence and comfort. Results: Preliminary results demonstrate strong feasibility and suggest a trend toward improved learner confidence and comfort in navigating difficult conversations. Conclusions: This AI innovation offers flexible and feasible communication training, and while not a full replacement for SP encounters, it offers increased scalability and is more cost-effective. Ongoing development will enhance AI realism and scenario diversity. This innovation has significant implications for pediatric education, transforming communication training through AI augmentation. This tool could also serve care teams well by providing a just-in-time trainer for real-world clinical scenarios.
Havalad et al. (Sun,) studied this question.