ehr

CARing 框架通过组合式语义ID和覆盖感知的强化学习奖励，优化大语言模型在多标签下次就诊诊断预测中的推理轨迹， 在 MIMIC-III/IV 数据集上超越所有 EHR 训练基线。

This paper re-implements 12 historical and recent algorithms to study barriers in evaluating methodological progress in health AI for electronic health records, finding that aggregate algorithm comparisons transfer across tasks and datasets, but clinically meaningful tasks exhibit greater task-algorithm interaction.

This paper introduces ReTA, a reinforcement learning-based framework for dynamically augmenting electronic health record graphs with external knowledge graphs to improve prediction tasks like diagnosis and mortality.

Mr.Dec is a multimodal Transformer-based model that sequences daily Electronic Health Record updates and chest X-ray findings to predict 30-day hospital readmissions, achieving state-of-the-art results on MIMIC datasets.

This arXiv paper proposes a mathematical and engineering architecture, the Geometric Belief Interface, for secure EHR interoperability using logit boundaries and sheaf-enclave protocols, with a narrow benchmark showing zero coverage for a 4B model.

Introduces CliniCARE-Bench, a deployment-oriented benchmark for evaluating AI agents on clinical audit tasks over longitudinal EHR data, with 25 clinician-validated scenarios and 750 patient cases. It assesses verdict accuracy, evidence grounding, policy adherence, and calibrated abstention, finding that raw accuracy overstates investigation quality.

This paper compares five feature selection methods for EHR diagnosis codes in opioid use disorder prediction, finding that NTK sensitivity offers the best accuracy-stability balance while LLM-guided selection adds complementary clinical signals.

Proposes Patient Sampling, a pretraining sequence construction method for EHR foundation models that improves downstream performance over the standard Global Stream baseline on MIMIC-IV datasets, highlighting the importance of sequence construction in autoregressive health models.

CardioMeta is a calibrated multi-task framework for jointly predicting diabetes, hypertension, and cardiovascular disease across NHANES and MIMIC-IV data, emphasizing leakage control, calibration, and transparent reliability.

LLM4EHR proposes a clinical foundation model that temporally aligns Electronic Health Record time series with medical event sequences using a domain-adapted large language model and a regularized contrastive objective, improving downstream prediction tasks.

This paper proposes an explicit multimodal routing framework for clinical prediction using EHR data, enabling interpretable, robust, and auditable reasoning across structured variables, clinical notes, and chest X-rays via discrete unimodal, bimodal, and trimodal routes with inference-time route masking for missing modality simulation.

This paper evaluates retrieval-augmented generation (RAG) versus long-context prompting for clinical reasoning tasks over electronic health records, finding RAG to be token-efficient and competitive, especially for imaging extraction and antibiotic timeline reconstruction.

This paper systematically compares discrete, continuous, and hybrid value encoding strategies for transformers in electronic health record data, finding that hybrid token-based approaches with binning provide robust performance and are recommended as a practical default.

PORTER is a language-grounded structured EHR foundation model that represents clinical events through text descriptions and numeric values, enabling vocabulary-independent transfer across institutions without retraining. On pediatric prediction tasks, PORTER matches fixed-vocabulary models and recovers 97.1% of AUROC when transferred to unseen event descriptions.

Introduces PhysAssistBench, a benchmark for evaluating LLMs in interactive doctor-patient-EHR assistance. Experiments show current models are unreliable in this setting, highlighting the need for coordinated capabilities.

Introduces AIPatient Arena, an EHR-grounded evaluation framework for assessing LLMs across multiple dimensions of clinical competence. The study reveals strengths in interviewing and ethics but weaknesses in handling ambiguity and diagnostic accuracy.

This paper proposes a Machine-Learned Comorbidity Index (MLCI) that uses diagnosis codes and nonlinear learning to improve risk adjustment across multiple clinical outcomes, outperforming traditional mortality-centric indices.

This paper investigates explicit encoding of ICD-10-CM hierarchy in EHR foundation models, using hierarchical token augmentation and graph-based code representations. Experiments on MIMIC-IV and eICU show improvements over flat code representations for in-domain and cross-dataset prediction tasks.

This paper introduces a lightweight, end-to-end benchmarking framework for reproducible synthetic Electronic Health Record (EHR) generation, unifying multiple baselines (MedGAN, CorGAN, PromptEHR, HALO) and a GPT-2 baseline under a single pipeline with a rigorous privacy-utility evaluation suite.

This paper presents Traj-Evolve, a self-evolving multi-agent system that uses an experience pool and multi-agent reinforcement learning to model patient trajectories from longitudinal EHRs for lung cancer early detection, outperforming strong baselines.