256 papers
Health informatics sits at the vibrant intersection of medicine, data science, and technology, transforming how we store, analyze, and utilize health information. This rapidly evolving field empowers clinicians and researchers to uncover patterns in patient data, improve diagnostic accuracy, and personalize treatment plans without getting lost in complex databases. By turning raw medical records into actionable insights, these innovations are reshaping the future of healthcare delivery and population health management.
At Gist.Science, we bridge the gap between cutting-edge research and public understanding by curating the latest preprints from medRxiv specifically within this domain. Our team processes every new submission in this category, providing both accessible plain-language explanations and detailed technical summaries to ensure the science is clear for everyone, from policymakers to curious readers. Below are the latest papers in health informatics, freshly distilled and ready for you to explore.
This paper introduces a multidimensional prediction-context framework that enhances the interpretability of ICU mortality models by complementing calibrated risk scores with insights into model behavior, data availability, physiological trends, and feature attribution, thereby providing a more comprehensive understanding of prediction formation beyond simple risk estimates.
This paper proposes a Neural Gated Fusion architecture that dynamically balances visual chest X-ray data and clinical text using an adapted Gated Multimodal Unit, demonstrating superior performance in detecting thoracic pathologies—particularly those with subtle visual evidence—compared to static fusion and unimodal approaches on a clinically diverse MIMIC-CXR subset.
This paper presents findings from a dual-perspective survey revealing that while openEHR has achieved significant national-scale adoption across 26 countries through both vendor-driven regional implementations and government-led national systems, its growth remains driven by practitioner innovation rather than regulatory mandates, leaving it vulnerable to structural fragility and hindered by a widespread lack of specialized knowledge.
This study demonstrates that a transformer-based model can accurately identify clinical study designs in biomedical literature, revealing significant limitations in the National Library of Medicine's indexing—particularly for cohort studies—and highlighting the need for a new manually annotated corpus to serve as a reliable gold standard for training and evaluation.
This study reveals that while frontier large language models generate a significant proportion of unreliable medical citations—primarily through misattribution rather than fabrication—an automated Chain-of-Verification system can detect these errors with expert-level accuracy, offering a viable solution for ensuring citation integrity in AI-assisted medical writing.
This study demonstrates that a published polygenic risk score for hypercholesterolemia (PGS000936) effectively stratifies risk and distinguishes monogenic-negative cases from controls in an independent German cohort, while highlighting the necessity of population-specific calibration for accurate clinical interpretation.
This study demonstrates that auditing routine clinical labels for operator-level bias is critical before training, and that while standard fine-tuning and reinforcement learning on limited specialist data failed to outperform a traditional logistic regression, knowledge distillation from a frontier model into a local 4B parameter model achieved superior performance, establishing a practical deployment recipe for cognitive-screening programs.
This study demonstrates that machine learning analysis of baseline and early dynamic shifts in peripheral immune profiles, rather than absolute counts after two cycles, can predict survival outcomes in extensive-stage small-cell lung cancer patients treated with immunochemotherapy.
This paper introduces PCGS, an explainable graph neural network framework that integrates multi-modal omics and clinical data to identify biomarkers and risk groups for pediatric cancers like glioma and Wilms tumor by leveraging Shapley values for feature attribution.
This factorial study demonstrates that in medical LLM interactions, users are slightly more likely to adopt incorrect suggestions from a source with a specialist title but low stated accuracy than from a student with high stated accuracy, highlighting that answer revisions following user challenges should not be automatically treated as independent second opinions.