Large Language Models in Spine Surgery : A Narrative Review of Performance Paradox and Clinical Integration...

Area of Science:

• Spine Surgery
• Artificial Intelligence
• Medical Informatics

Background:

• Large language models (LLMs) show promise in healthcare applications.
• A performance paradox exists between LLM technical capabilities and clinical utility in specialized fields like spine surgery.

Purpose of the Study:

• To synthesize the performance of LLMs in spine surgery applications.
• To examine the gap between LLM technical metrics and their real-world clinical value.

Main Methods:

• A narrative review of literature from January 2023 to February 2026.
• Searches were conducted across PubMed, EMBASE, and Google Scholar.
• Thematic analysis of 42 studies focusing on LLM performance in documentation, patient communication, and decision-making.

Main Results:

• LLMs demonstrated high accuracy in structured tasks like CPT coding (AUROC 0.87) and classification (91%).
• Patient communication showed high satisfaction but limited emotional intelligence; decision-making accuracy was lower than surgeons'.
• Image-based tasks (e.g., Cobb angle measurement) and procedure-level agreement remain significant challenges.

Conclusions:

• LLMs offer near-term utility for standardized, text-based tasks in spine surgery.
• Current evidence does not support autonomous LLM use in complex decision-making or image interpretation.
• Phased implementation with mandatory human oversight is crucial for safe LLM integration.