Outpatient reception via collaboration between nurses and a large language model: a randomized controlled trial.

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare
• Patient Communication Systems

Background:

• Healthcare reception is vital but labor-intensive.
• Current communication relies heavily on human staff.
• Large language models (LLMs) offer a potential solution for improving efficiency.

Purpose of the Study:

• To develop and evaluate a site-specific prompt engineering chatbot (SSPEC) for medical reception.
• To assess the effectiveness of LLM-assisted communication compared to traditional methods.
• To investigate a collaborative model between nurses and the SSPEC.

Main Methods:

• Curated a dataset of 35,418 outpatient-receptionist conversations.
• Developed a site-specific prompt engineering chatbot (SSPEC).
• Conducted a randomized controlled trial with 2,164 participants comparing nurse-SSPEC collaboration to nurse-only sessions.

Main Results:

• SSPEC resolved 68.0% of queries in ≤2 rounds, outperforming nurses (50.5%).
• The nurse-SSPEC model achieved higher patient satisfaction (P < 0.001).
• Reduced repeated queries (3.2% vs 14.4%), negative emotions (2.4% vs 7.8%), and improved response quality.

Conclusions:

• LLM integration into hospital workflows is feasible.
• The nurse-SSPEC collaboration model enhances patient communication and satisfaction.
• This paradigm improves healthcare experiences for both patients and nurses.