Enhancing doctor-patient communication using large language models for pathology report interpretation.

Area of Science:

• Medical Informatics
• Artificial Intelligence in Healthcare

Background:

• Postoperative pathology reports contain complex data challenging for patients.
• This complexity can hinder doctor-patient communication and affect outcomes.

Purpose of the Study:

• To assess the utility of large language models (LLMs) in simplifying pathology reports.
• To evaluate the impact of AI-generated interpretive pathology reports (IPRs) on patient understanding and communication efficiency.

Main Methods:

• Analysis of 698 pathology reports from four hospitals (October-December 2023).
• Development of GPT-4 based templates for interpretive pathology reports (IPRs).
• Evaluation of patient understanding using a custom scale and measurement of communication time.

Main Results:

• LLM-generated IPRs significantly improved patient understanding scores (5.23 to 7.98, P < 0.001).
• Doctor-patient communication time decreased by over 70% (35 to 10 min, P < 0.001).
• High consistency (4.95/5) was observed between original and interpretive reports.

Conclusions:

• LLMs effectively enhance doctor-patient communication by translating complex medical information.
• Improved understanding and reduced communication time show potential for positive patient engagement.
• AI holds promise for bridging communication gaps in healthcare settings.