LLMs unify EHR data for improved medical diagnosis prediction

Researchers have developed a novel method for predicting primary ICD codes using large language models (LLMs) by creating a shared embedding space for multimodal data. This approach combines structured electronic health record (EHR) variables with clinical narratives, leveraging frozen medical LLM representations. The combined probe significantly outperformed single-modality probes and baseline methods on the MIMIC-IV dataset, demonstrating improved accuracy in diagnosis prediction and efficient cross-dataset transfer capabilities. AI

IMPACT This research could enhance the accuracy and efficiency of automated medical coding systems, improving healthcare administration and research.

RANK_REASON The cluster contains an academic paper detailing a new methodology for medical diagnosis prediction using LLMs. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• ICD codes
• LLM
• MedFound-Llama3-8B-finetuned
• MIMIC-III
• MIMIC-IV
• XGBoost

• paper
• model release
• infra

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →