The paper introduces Span-Centric Learning (SCL), a method that decouples coding knowledge acquisition from its application in full clinical documents. SCL first trains on short text spans to learn coding knowledge, then fine‑tunes on a small set of annotated documents, enabling an 8B LLM to outperform state‑of‑the‑art methods with only 248 full documents. The approach also yields explicit, human‑aligned supporting evidence for each predicted ICD code and can learn code knowledge from spans even when the codes are absent from the full documents.
arXiv:2608.20887v1 Announce Type: cross Abstract: Automatic Medical Coding (AMC), which assigns standardized International Classification of Diseases (ICD) codes to clinical notes, is essential for m...
Automatic medical coding assigns ICD codes to clinical notes, but it remains challenging due to long documents, imbalanced label distributions, and diverse terms. These challenges are especially sever...
CoLa-ICD is a knowledge‑enhanced framework designed to improve automatic medical coding of ICD codes in long, imbalanced clinical documents. It enriches ICD labels with external terms, models dependencies among related codes, and strengthens the alignment between label semantics and clinical evidence, particularly for rare codes. Experiments demonstrate that CoLa-ICD achieves state‑of‑the‑art performance in AUC, F1, and P@k, with larger gains in larger and sparser label spaces.
arXiv:2607. 11954v1 Announce Type: cross Abstract: Clinical code prediction maps unstructured discharge summaries to ICD-10-CM leaf codes in a large, sparse, and deeply hierarchical label space.
Generative approaches to automatic International Classification of Diseases (ICD) coding use large language models (LLMs) to predict ICD codes from clinical documents. However, original ICD code strin...
EHRAdapt is an adapter that maps electronic health record tuples directly into a frozen language model’s embedding space, using learned embeddings for modality, time gaps, and dedicated vectors for event codes. It represents each event vector as a sum of a frozen, ontology‑trained semantic prior and a low‑rank evidence residual, enabling robust learning even for rare clinical events. Experiments on four million patient records with three large language model backbones show that the full adapter outperforms all ablations in next‑event prediction and improves rare‑disease discrimination in classification tasks.
arXiv:2606. 15447v1 Announce Type: new Abstract: Electronic health record foundation models typically treat ICD diagnosis codes as flat tokens, overlooking the clinically meaningful hierarchical structure that captures disease families, subcategories, and fine-grained diagnostic detail.
arXiv:2608. 00935v1 Announce Type: new Abstract: Electronic Health Records (EHRs) are widely used for clinical risk prediction using machine learning.
REFINE is a framework that refines medical concept representations by creating patient‑specific temporal graphs from a global text‑attributed knowledge graph. It uses a reinforcement learning policy to allocate a personalized graph expansion budget for each observed code, then processes the resulting graph with a heterogeneous GNN and a frozen LLM that refines representations via graph‑aware soft prompts. Experiments on MIMIC‑III and MIMIC‑IV demonstrate that REFINE consistently improves various EHR prediction backbones, surpasses strong baselines, and shows robust gains across ablation studies, KG selection, and data insufficiency scenarios.
The paper introduces CAST, a concept-guided artifact suppression tuning framework that uses sparse autoencoders to identify and suppress note-specific artifacts in clinical language models. CAST labels latent features with an LLM-assisted pipeline and ICD‑10 constraints, then fine‑tunes the model while providing post‑hoc per‑concept attributions for auditability. In experiments on MIMIC‑IV discharge‑note mortality prediction, CAST outperforms standard fine‑tuned encoders and competes with strong LLM baselines while offering a feature‑level audit trail of clinical concepts and suppressed artifacts.
arXiv:2606. 29503v1 Announce Type: cross Abstract: The verbose context problem occurs when structured concepts have token-inefficient textual representations.
The paper introduces a unified pre‑training framework for medical representations that incorporates hierarchical sub‑token aggregation, partial masking, and cross‑reference mechanisms to better capture the structure of medical codes. The resulting model outperforms existing BERT‑based approaches on pre‑training tasks and downstream clinical predictions, such as dementia onset and hospitalization. An in‑silico drug repositioning study for Alzheimer’s disease demonstrates the framework’s ability to rediscover known drugs and prioritize new hypotheses without external literature, establishing a workflow for hypothesis generation and prioritization based on observational data.