Objectives/Purpose
This study aims to develop and evaluate an AI-assisted Truth-Telling Classification Model designed to enhance communication skills training (CST) for healthcare professionals.
Sample and Setting
A single-group pre- and post-test design was adopted, involving 99 nurse practitioners (NPs) from multiple hospitals across Taiwan who participated in 25 CST sessions.
Procedures
The SHARE Model was employed as the framework for CST. All truth-telling role-play sessions were videotaped and transcribed verbatim. Inter-rater reliability (r=0.85) was established prior to coding. To minimize bias, all coding was conducted exclusively by the principal investigator (PI).
Results
A total of 4,111 coded instances were identified based on the SHARE Model. The application rates of its components were as follows: S: Creating a supportive environment (16%), H: Delivering the bad news (24%), A: Providing additional information (44%), and RE: Offering reassurance and emotional support (15%).
These codes were used to train an AI-driven machine learning model for truth-telling classification. Additionally, comparisons between the PI’s SHARE classifications—considered the gold standard—and facilitators’ shorthand classifications documented on whiteboards revealed that while facilitators achieved an average classification accuracy of 82%, the integrity rate was only 41%. This underscores the need for an AI-assisted truth-telling classification system to improve CST precision and consistency.
Conclusion and Clinical Implications
CST sessions continue to collect extensive data to refine algorithmic modeling and machine learning. Moving forward, the development of "human-machine collaboration" in CST will be prioritized. This study represents a significant and innovative step in integrating AI with medical education, fostering enhanced communication training and generating long-term benefits for healthcare professionals