Speech recognition technology for assessing team debriefing communication and interaction patterns: An algorithmic...

Area of Science:

• Medical Education Technology
• Artificial Intelligence in Healthcare
• Simulation-Based Training

Background:

• Debriefings are crucial for learning in simulation-based medical education.
• Educators face challenges in conducting effective debriefings due to limited empirical data.
• Objective measurement of debriefing interactions is needed to improve processes.

Purpose of the Study:

• To explore the technical feasibility of audio-based speaker diarization for measuring debriefing interaction patterns.
• To investigate the automatic creation of statistical analyses and visualizations from audio recordings.
• To objectively assess interaction dynamics between debriefers and participants.

Main Methods:

• Audio recordings of ten debriefings from simulation-based team training were analyzed.
• Speaker diarization machine learning algorithms processed audio data.
• Manual validation assessed algorithm accuracy, and two sessions were compared for interaction pattern differences.

Main Results:

• Speaker diarization accurately visualized interaction patterns (sociograms) with 97.78% average accuracy.
• Quantified instructor-triggered feedback rates, showing variations between debriefing sessions (28% vs. 36%).
• Demonstrated technical feasibility of automated, objective analysis of debriefing communication.

Conclusions:

• Speaker diarization technology accurately and automatically visualizes debriefing interactions.
• This technology can aid in developing simulation faculty by providing objective feedback on debriefing facilitation.
• Visualizations support instructors in assessing sessions, potentially enhancing learning outcomes in healthcare education.