Papers/2608.17051
🧪 Test?View on arXiv

Institution-Specific LLM Prompting Recovers PHI That De-identification Systems and Their Gold Standards Both Miss

Not specified in the provided content

de-identificationLLMhealthcareprompt engineering
2608.17051
Builder Relevance
80%
2h ago

Abstract

This study demonstrates that large language models (LLMs) can effectively recover institution-specific protected health information (PHI) that traditional de-identification systems overlook.

Reality Card

Core Claim

LLMs outperformed purpose-built de-identification systems in recovering institution-specific PHI while achieving a high recall rate.

Method / Result

LLMs achieved a recall of 0.981 with an F1 score of 0.907 after prompt adjustments.

Limitations

LLMs are more costly to run than traditional methods, which may limit their widespread adoption despite their effectiveness.

Paper to code

Verified implementation resources so builders can test the paper’s claims instead of stopping at the abstract.

No verified implementation link has been attached yet. AIBuzzHub will keep this panel separate from unverified search results.
← Back to all papers