Capturing patient information at nursing shift changes: methodological evaluation of speech recognition and...

Area of Science:

• Medical Informatics
• Natural Language Processing
• Healthcare Technology

Background:

• Nursing handover is critical for patient safety.
• Manual transcription of handover documents is time-consuming and prone to errors.
• Automating this process can improve efficiency and data accuracy.

Purpose of the Study:

• To evaluate the feasibility of using speech recognition and information extraction for generating Australian nursing handover documents.
• To assess the correctness of speech recognition and information extraction methods.
• To understand clinician preferences for technology in handover documentation.

Main Methods:

• Evaluated 15 recorder-microphone combinations for speech recognition correctness and clinician preference.
• Assessed information extraction correctness using 260 documents and the CRF++ toolkit.
• Collected data through surveys and interviews with five participants.

Main Results:

• A noise-cancelling lapel microphone achieved 79% speech recognition correctness and was highly preferred.
• Information extraction effectively filtered irrelevant text (85% F1) and identified key classes (87% and 70% F1).
• Accented speech and female speakers presented greater recognition challenges.

Conclusions:

• Speech recognition and information extraction are feasible for automating nursing handover documents.
• This automation enhances text entry, data accessibility, and supports computerized decision-making.
• Benefits include instant access, reduced information loss, and fewer delays compared to manual processes.