Researchers propose longinus framework for Next-Visit diagnosis prediction
A new paper on arXiv introduces LONGINUS, a retrieval-memory multi-agent system designed to improve next-visit diagnosis prediction in healthcare. The authors argue that current methods fail to account for variability in diagnosis sets, temporal changes (persistent, new, or recurrent diagnoses), and varying patient history depth. Their evaluation framework compares rule-based heuristics, LLM…
Key points
- LONGINUS framework combines retrieval and memory for next-visit diagnosis prediction, outperforming LLM-only methods in transition-specific cases
- Researchers find simple repetition of last-visit diagnoses often surpasses LLM-based approaches despite limited predictive power
- Tests on MIMIC-III and MIMIC-IV datasets with four LLM backbones show LONGINUS leads in transition-specific performance
LONGINUS combines retrieved patient record evidence with structured memory of the patient’s trajectory, using cross-view agreement to adjudicate disagreements without exposing their sources. Tests on MIMIC-III and MIMIC-IV datasets with four LLM backbones show it achieves the highest transition-specific performance among LLM-based methods in nearly all scenarios while maintaining competitiveness on aggregate metrics. The work highlights the need for more nuanced evaluation frameworks in clinical prediction tasks.
The headline, key points and digest above were generated by Digest AI's editorial model from the linked sources. Automated summaries can contain errors: the sources are the record. Spotted a mistake? Tell us. Published by Martin K., who runs Digest AI and handles corrections.
More in Research
All →- Anthropic releases FlashProve open-source system for provable AI inference · 7 src
- Gemini and Claude debate AI personality, ending in a three-way tie · 1 src
- Tencent releases WeVisOne-VL, a 2B/4B open vision model for detection, OCR and matting · 1 src
- Lama-dt outperforms baselines on meta-world and mujoco · 1 src
- Researchers release Dory1.0-4B, a 4B-parameter looped model with 256K-token context · 1 src
Comments
via GitHub Discussions