Researchers have developed a novel two-stage framework for automated ICD coding, addressing limitations in current methods. This new approach integrates multimodal patient data, including structured electronic health records and unstructured clinical notes, through a gated fusion mechanism. The first stage retrieves potential ICD codes using a dual-encoder retrieval model, while the second stage employs an LLM-based re-ranker to provide refined, ranked codes with explanations. Experiments indicate improved Micro-F1 and Precision scores compared to existing multimodal classifiers, suggesting a practical alternative for automated ICD coding. AI
IMPACT This research offers a more accurate and transparent method for automated ICD coding, potentially improving clinical research and billing processes.
RANK_REASON This is a research paper detailing a new framework for automated ICD coding. [lever_c_demoted from research: ic=1 ai=1.0]
Read on arXiv cs.IR (Information Retrieval) →
- arXiv
- electronic health records
- ICD-9
- International Statistical Classification of Diseases and Related Health Problems
- Micro F1
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
