ehr

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.

ChatHealthAI is a multimodal reasoning framework that aligns structured EHR representations with a frozen LLM to enable grounded clinical reasoning while maintaining predictive performance.

This paper introduces EHR-ReasonCon, a reasoning-intensive benchmark for consistency verification between clinical notes and structured tables in electronic health records, and EHR-Inspector, an LLM-based framework that achieves state-of-the-art performance in detecting discrepancies.

Introduces TreeText-CTS, a method that converts irregular EHR trajectories into compact, source-traceable tree-path evidence units without patient-level summarization. Achieves state-of-the-art AUROC and AUPRC among text-based EHR time-series interfaces on three clinical benchmarks.

GraphDiffMed is a medication recommendation framework that uses dual-scale differential attention and pharmacological graph priors to improve recommendation quality and safety on EHR data. Experiments on MIMIC-III show consistent improvements over baselines.

DT-Transformer is a foundation model trained on 57.1 million structured EHR entries from 1.7 million patients across 11 hospitals in the Mass General Brigham health system, achieving strong discrimination for next-event prediction across 896 disease categories.

This paper investigates the latent structure of multimodal embeddings from a masked autoencoder for pediatric sleep analysis. It shows that augmenting embeddings with geometric, topological, and clinical features improves prediction and calibration for sleep-related events.

COTCAgent is a hierarchical reasoning framework for longitudinal electronic health records that uses a probabilistic chain-of-thought completion approach, achieving 90.47% Top-1 accuracy on a self-built dataset and outperforming existing medical agents.

This paper presents a nationwide EHR-based chronic rhinosinusitis prediction model using demographic-stratified models and a hybrid feature-selection pipeline, achieving an overall AUC of 0.8461 on data from the All of Us Research Program.

A 9-week pilot at a Dutch academic hospital shows 58% of admissions used LLM-generated discharge drafts, with 87% of clinicians reporting reduced documentation time and 91% intending continued use.