Can AI assist in reducing diagnostic error? A narrative review

Area of Science:

• Medical Informatics
• Clinical Decision Support
• Artificial Intelligence in Healthcare

Background:

• Diagnostic error impacts 5-10% of healthcare encounters, leading to patient harm and mortality.
• Most diagnostic errors stem from clinician reasoning flaws, highlighting a need for improved diagnostic processes.
• Artificial intelligence (AI), particularly large language models (LLMs), offers potential solutions for reducing diagnostic errors.

Purpose of the Study:

• To review the current capabilities of AI and LLMs in assisting diagnostic performance during clinician-patient encounters.
• To provide practicing clinicians with an understanding of AI/LLM adoption in diagnostics.
• To address key questions regarding the integration of AI/LLMs into clinical practice.

Main Methods:

• Narrative review of contemporary state-of-the-art research on AI and LLMs in diagnostics.
• Analysis of AI/LLM positioning in assisting diagnostic performance.
• Focused on clinician-patient encounters and diagnostic decision-making.

Main Results:

• AI tools have advanced to improve clinician diagnostic decision-making and institutional diagnostic safety.
• LLMs demonstrate evolving diagnostic capabilities requiring continuous monitoring.
• AI and LLMs are positioned to assist, not replace, clinician diagnostic reasoning.

Conclusions:

• AI and LLMs are mature enough to enhance diagnostic accuracy and safety in clinical settings.
• Continuous evaluation of rapidly developing LLMs is essential.
• A balanced approach integrating AI/LLMs to complement clinician expertise is crucial for optimal diagnostic outcomes.