Enhancing readability and understandability of vascular surgery discharge summaries using artificial intelligence.

Area of Science:

• Medical Informatics
• Health Communication
• Artificial Intelligence in Healthcare

Background:

• Patient comprehension of medical information, particularly discharge summaries, is crucial for effective postoperative care and engagement.
• Traditional discharge summaries often present complex medical jargon, hindering patient understanding and adherence to treatment plans.
• Simplifying medical documentation is a key goal in improving patient-centered care and health literacy.

Purpose of the Study:

• To evaluate the efficacy of Large Language Models (LLMs) in simplifying vascular surgery discharge summaries.
• To assess if AI-generated summaries maintain accuracy and completeness while improving patient readability and engagement.
• To determine the impact of LLMs on patient understanding and actionability of postoperative care instructions.

Main Methods:

• A cross-sectional multicentric study involving 90 vascular surgery discharge summaries across three patient pathology groups.
• Discharge summaries were processed by LLMs (e.g., ChatGPT-4) to achieve a 6th-grade reading level.
• Readability (Flesch-Kincaid), understandability and actionability (PEMAT-P), accuracy, and completeness were quantitatively assessed.

Main Results:

• AI-generated summaries showed a 39.6% reduction in reading grade level and a 106.37% increase in readability ease.
• High scores for understandability (77.71) and accuracy (5.21) were achieved, though actionability scores were moderate (52.12).
• A small percentage of summaries contained omissions (10.8%) or hallucinations (7.5%), necessitating human review and correction.

Conclusions:

• LLMs can significantly enhance the readability and accessibility of vascular surgery discharge summaries, improving patient comprehension and engagement.
• AI-generated summaries demonstrate high accuracy, but vigilance is required to address and minimize errors like omissions and hallucinations.
• LLMs represent a promising tool for optimizing healthcare communication, though further research is needed to refine error reduction and enhance actionable content.