Large language models for extraction of OPS-codes from operative reports in meningioma surgery - Takeaways - MDSpire

Large language models for extraction of OPS-codes from operative reports in meningioma surgery

Share

1

The study evaluates the ability of GPT models to extract OPS codes from surgical reports of meningioma resections.
2

Surgeons and professional coders achieved sufficient coding accuracy of 99-100%, while LLMs like GPT-4o scored 78-89%.
3

For optimal coding, professional coders performed best at 94%, while surgeons had a 69% error rate.
4

GPT CodeMedic outperformed GPT-4o in both sufficient and optimal coding categories.
5

Professional coders significantly surpassed LLMs in coding accuracy, highlighting the importance of human expertise.