Primary ICD Category Prediction using LLM-based Probing

• URL: http://arxiv.org/abs/2606.28798v1
• Date: Sat, 27 Jun 2026 08:10:05 GMT
• Title: Primary ICD Category Prediction using LLM-based Probing
• Abstract summary: We constructed a MIMIC-IV cohort of 13,645 admissions from the 10 most frequent primary ICD-10 codes.<n>We extracted hidden states from five transformer layers and trained linear probes for structured-only, unstructured-only, and combined inputs.<n>The combined probe performed best on MIMIC-IV (87.69% strict; 91.45% medical accuracy), exceeding both single-modality probes and baselines.
• Score: 7.948368617192842
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Objective: ICD codes are central to reimbursement, research, and population health surveillance, yet automated coding systems often struggle to integrate diagnostic signals from both clinical narratives and structured electronic health record (EHR) variables. We evaluated whether frozen medical large language model (LLM) representations can serve as a shared embedding space for multimodal primary diagnosis category prediction. Materials and Methods: We constructed a MIMIC-IV cohort of 13,645 admissions from the 10 most frequent primary ICD-10 codes, consolidated into seven categories. Structured variables were serialized into clinical narratives and combined with leakage-pruned discharge notes. Using a frozen MedFound-Llama3-8B-finetuned backbone, we extracted hidden states from five transformer layers and trained linear probes for structured-only, unstructured-only, and combined inputs, comparing against XGBoost and information-matched PLM-ICD baselines and evaluating MIMIC-III adaptation with a compact bottleneck adapter. Results: The combined probe performed best on MIMIC-IV (87.69% strict; 91.45% medical accuracy), exceeding both single-modality probes and baselines. The structured-only probe outperformed its standard baseline by 6.19 points in medical accuracy. Diagnostic information became increasingly linearly separable in deeper layers, and a 2M-parameter adapter restored cross-dataset transfer to MIMIC-III using only 5% of target labels. Discussion: LLM embeddings can unify structured and narrative EHR information for multimodal diagnosis prediction, supporting efficient reuse of clinical representations across modalities and datasets through a small representation-level module. Conclusion: Multimodal probing of frozen medical LLM representations provides a practical approach for studying EHR modalities and adapting clinical representations across datasets.

Related papers

• Concept-Grounded Reasoning with Prompt-Driven Localization for Interpretable Structured Report Generation [66.62662330558906]
  CORAL is a multimodal framework that integrates spatial grounding and concept-level supervision into a unified reasoning process.<n>CORAL employs a prompt-driven medical segmentation model to localize lesions and predicts multi-class clinical attributes through a Concept Bottleneck module.<n>Experiments on BUS-CoT and IU X-ray datasets demonstrate consistent improvements in diagnostic accuracy, concept consistency, and report quality over strong general-purpose and medical MLLMs.
  arXiv  Detail & Related papers  (2026-09-14T10:21:23Z)
• MTDiag: A Multi-Turn Diagnostic Dataset Towards Clinically Meaningful LLM Evaluation [39.83005700511833]
  We present MTDiag, a large multi-turn diagnostic dialogue dataset constructed from three heterogeneous sources.<n>We release the schema, a UserLM-8B-based utterance-generation pipeline, and the physician-validated dataset that converts structured clinical evidence into natural-language utterances.
  arXiv  Detail & Related papers  (2026-08-25T19:27:10Z)
• A Federated and Parameter-Efficient Framework for Large Language Model Training in Medicine [59.78991974851707]
  Large language models (LLMs) have demonstrated strong performance on medical benchmarks, including question answering and diagnosis.<n>Most medical LLMs are trained on data from a single institution, which faces limitations in generalizability and safety in heterogeneous systems.<n>We introduce the model-agnostic and parameter-efficient federated learning framework for adapting LLMs to medical applications.
  arXiv  Detail & Related papers  (2026-01-29T18:48:21Z)
• Revealing Interconnections between Diseases: from Statistical Methods to Large Language Models [0.15558822250482188]
  Identifying disease interconnections through manual analysis of large-scale clinical data is labor-intensive, subjective, and prone to expert disagreement.<n>We evaluate seven approaches for uncovering disease relationships based on two data sources: sequences of ICD-10 codes from MIMIC-IV EHRs and the full set of ICD-10 codes.<n>Our framework integrates the following: (i) a statistical co-occurrence analysis and a masked language modeling (MLM) approach using real clinical data; (ii) domain-specific BERT variants; and (iii) a general-purpose BERT and document retrieval.
  arXiv  Detail & Related papers  (2025-10-06T15:09:39Z)
• MedPatch: Confidence-Guided Multi-Stage Fusion for Multimodal Clinical Data [0.46040036610482665]
  Real-world medical data is heterogeneous in nature, limited in size, and sparse due to missing modalities.<n>Inspired by clinical prediction tasks, we introduce MedPatch, which seamlessly integrates multiple modalities via confidence-guided patching.<n>We evaluate MedPatch using real-world data consisting of clinical time-series data, chest X-ray images, radiology reports, and discharge notes extracted from the MIMIC-IV, MIMIC-CXR, and MIMIC-Notes datasets.
  arXiv  Detail & Related papers  (2025-08-07T12:46:26Z)
• Medchain: Bridging the Gap Between LLM Agents and Clinical Practice with Interactive Sequence [68.05876437208505]
  We present MedChain, a dataset of 12,163 clinical cases that covers five key stages of clinical workflow.<n>We also propose MedChain-Agent, an AI system that integrates a feedback mechanism and a MCase-RAG module to learn from previous cases and adapt its responses.
  arXiv  Detail & Related papers  (2024-12-02T15:25:02Z)
• MEDFuse: Multimodal EHR Data Fusion with Masked Lab-Test Modeling and Large Language Models [11.798375238713488]
  MEDFuse is a framework that integrates structured and unstructured medical data. It achieves over 90% F1 score in the 10-disease multi-label classification task.
  arXiv  Detail & Related papers  (2024-07-17T04:17:09Z)
• A Spectrum Evaluation Benchmark for Medical Multi-Modal Large Language Models [57.88111980149541]
  We introduce Asclepius, a novel Med-MLLM benchmark that assesses Med-MLLMs in terms of distinct medical specialties and different diagnostic capacities.<n>Grounded in 3 proposed core principles, Asclepius ensures a comprehensive evaluation by encompassing 15 medical specialties.<n>We also provide an in-depth analysis of 6 Med-MLLMs and compare them with 3 human specialists.
  arXiv  Detail & Related papers  (2024-02-17T08:04:23Z)
• XAI for In-hospital Mortality Prediction via Multimodal ICU Data [57.73357047856416]
  We propose an efficient, explainable AI solution for predicting in-hospital mortality via multimodal ICU data. We employ multimodal learning in our framework, which can receive heterogeneous inputs from clinical data and make decisions. Our framework can be easily transferred to other clinical tasks, which facilitates the discovery of crucial factors in healthcare research.
  arXiv  Detail & Related papers  (2023-12-29T14:28:04Z)
• Improving Multiple Sclerosis Lesion Segmentation Across Clinical Sites: A Federated Learning Approach with Noise-Resilient Training [75.40980802817349]
  Deep learning models have shown promise for automatically segmenting MS lesions, but the scarcity of accurately annotated data hinders progress in this area. We introduce a Decoupled Hard Label Correction (DHLC) strategy that considers the imbalanced distribution and fuzzy boundaries of MS lesions. We also introduce a Centrally Enhanced Label Correction (CELC) strategy, which leverages the aggregated central model as a correction teacher for all sites.
  arXiv  Detail & Related papers  (2023-08-31T00:36:10Z)
• Graph-Ensemble Learning Model for Multi-label Skin Lesion Classification using Dermoscopy and Clinical Images [7.159532626507458]
  This study introduces a Graph Convolution Network (GCN) to exploit prior co-occurrence between each category as a correlation matrix into the deep learning model for the multi-label classification. We propose a Graph-Ensemble Learning Model (GELN) that views the prediction from GCN as complementary information of the predictions from the fusion model.
  arXiv  Detail & Related papers  (2023-07-04T13:19:57Z)
• Self-supervised Answer Retrieval on Clinical Notes [68.87777592015402]
  We introduce CAPR, a rule-based self-supervision objective for training Transformer language models for domain-specific passage matching. We apply our objective in four Transformer-based architectures: Contextual Document Vectors, Bi-, Poly- and Cross-encoders. We report that CAPR outperforms strong baselines in the retrieval of domain-specific passages and effectively generalizes across rule-based and human-labeled passages.
  arXiv  Detail & Related papers  (2021-08-02T10:42:52Z)
• Inheritance-guided Hierarchical Assignment for Clinical Automatic Diagnosis [50.15205065710629]
  Clinical diagnosis, which aims to assign diagnosis codes for a patient based on the clinical note, plays an essential role in clinical decision-making. We propose a novel framework to combine the inheritance-guided hierarchical assignment and co-occurrence graph propagation for clinical automatic diagnosis.
  arXiv  Detail & Related papers  (2021-01-27T13:16:51Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.