Related Experiment Video
Updated: Jul 17, 2026
09:10
Digital Hybrid Model Preparation for Virtual Planning of Reconstructive Dentoalveolar Surgical Procedures
Published on: August 5, 2021
Artificial Intelligence for Automated Current Procedural Terminology Coding in Plastic and Reconstructive Surgery
Daniel Oh1, Jeffrey Cripe2, James Baez-Silva2
1From the Department of Statistics and Data Science, University of California, Los Angeles, CA.
Plastic and Reconstructive Surgery. Global Open
|July 16, 2026
Summary
A specialized AI system, ProCode, significantly improves Current Procedural Terminology (CPT) coding accuracy in plastic surgery. This advanced large language model (LLM) outperforms human coders and standard LLMs, especially for complex procedures.
Area of Science:
- Medical Informatics
- Artificial Intelligence in Healthcare
- Surgical Coding Systems
Background:
- Current Procedural Terminology (CPT) coding in plastic and reconstructive surgery presents significant challenges due to complexity, time demands, and error proneness, especially with multiple codes.
- Existing large language models (LLMs) show variable performance in accurately coding surgical procedures.
Purpose of the Study:
- To compare the accuracy of a fine-tuned hybrid LLM system (ProCode) against baseline LLMs and professional auditors for CPT coding in plastic surgery.
- To evaluate coding accuracy across different levels of CPT code set complexity.
Main Methods:
- A retrospective study analyzed 120 operative reports, categorized by CPT code complexity (low, medium, high).
- Accuracy was assessed by comparing coding results from baseline LLMs (OpenAI, Gemini, Anthropic), professional auditors, and the ProCode system against expert consensus coding.
Main Results:
- ProCode achieved the highest accuracy across all complexity levels (low: 100%, medium: 80%, high: 80%), surpassing human auditors and baseline LLMs.
- Baseline LLM performance degraded with increasing procedural complexity.
- Statistical analysis confirmed significant differences in accuracy between methods across all complexity tiers (P < 0.05).
Conclusions:
- A domain-specific, fine-tuned LLM (ProCode) offers superior accuracy for automated CPT coding in plastic and reconstructive surgery compared to general LLMs and human auditors.
- Specialized AI systems hold promise for enhancing the accuracy and efficiency of surgical billing workflows.