193 papers
This study demonstrates that a Nash bargaining-based multi-agent language model framework significantly improves the safety and efficiency of care plans for underserved Medicaid patients compared to a single-model baseline, while highlighting that equity requires explicit design attention rather than emerging automatically from multi-objective optimization.
The AutoscriberValidate study demonstrates that using the ambient digital scribe Autoscriber significantly reduces healthcare providers' typing workload and improves the overall quality of electronic medical notes by decreasing various types of documentation errors.
This study demonstrates that commercial vision-language models are critically vulnerable to adversarial attacks where OCR-readable text overlays override visual evidence in radiology tasks, causing widespread diagnostic failures that prompt-level defenses cannot reliably mitigate, thereby necessitating strict system-level safeguards before clinical deployment.
A cross-sectional survey of 500 patients in Jordan reveals that while there is conditional optimism regarding AI in healthcare, acceptance is heavily dependent on maintaining human physician involvement, ensuring transparency and privacy, and addressing disparities linked to education and digital literacy.
This study develops and validates a novel hybrid SARIMA-CNNAR model using ten years of national surveillance data to accurately forecast monthly tuberculosis incidence in Nepal, demonstrating superior performance over standalone statistical and machine learning benchmarks to support evidence-based public health planning.
MedOS is a general-purpose embodied world model that bridges the gap between abstract clinical reasoning and physical intervention by utilizing a dual-system architecture to autonomously execute complex medical procedures, simulate adverse events, and democratize clinical expertise by narrowing the performance gap between junior and senior physicians.
This study demonstrates that a fully automated pipeline using large language models can generate systematic reviews with citation accuracy and expert-rated quality surpassing human-authored counterparts, while simultaneously revealing critical limitations in information breadth and the urgent need for new verification standards and AI literacy in scientific publishing.
This paper proposes and evaluates two methods—a reliability score-based stratification and a Bayesian adjustment model—to resolve inconsistencies in self-reported onset ages within longitudinal healthcare surveys, demonstrating that both approaches significantly enhance data quality, predictive performance, and disease clustering coherence using data from the Canadian Partnership for Tomorrow's Health.
This study utilizes a graph neural network-based sentiment analysis model on over 340,000 app store reviews to identify that while therapeutic features like sound masking and sleep support drive positive user sentiment, issues regarding pricing, advertisements, and technical stability remain key areas for improvement in tinnitus management apps.
This protocol outlines a 90-day, two-arm, parallel-group randomized controlled trial designed to evaluate whether access to a native Epic generative AI chart summarization tool reduces cognitive task load and improves professional fulfillment among ambulatory providers compared to usual care.
This study of 30 clinicians reveals that edits to ambient AI-drafted clinical notes are primarily driven by the need to correct transcription errors, ensure clinical accuracy, mitigate liability risks, and meet billing standards, highlighting the necessity for improved AI customization, integration, and institutional support to enhance human-AI collaboration.
This paper presents an autonomous agentic framework that uses large language models to emulate clinical trials at scale and a Bayesian hierarchical model to calibrate EHR-derived results against RCTs, successfully reducing systematic discrepancies and improving the accuracy of institution-specific drug effect estimates.
This paper introduces CardioPulmoNet, a physiologically inspired neural network architecture that models cardiopulmonary dynamics to achieve stable, interpretable, and high-performing histopathological diagnosis, particularly under limited data conditions.
This study proposes an interpretable, two-stage machine learning pipeline using XGBoost and SHAP to enable cost-effective, non-invasive early detection of non-alcoholic fatty liver disease and its associated comorbidities through a user-friendly clinical interface.
This study validates "otto-SR," a large language model-based workflow that demonstrates high performance in automating article screening, data extraction, and risk of bias assessment, thereby enabling the rapid and reliable reproduction and updating of systematic reviews.
This study analyzes Reddit posts to reveal that while natural environments can alleviate social anxiety for many young people, they may also trigger symptoms in others due to fears of social judgment, suggesting that nature-based interventions require careful design to address these unique concerns.
This exploratory eye-tracking study reveals that primary care clinicians divert substantial visual attention to electronic health records to verify suicide risk indicators, often delaying patient discussions until confirming relevant data, thereby illustrating how EHR-embedded tools shape clinical attention and encounter flow.
By integrating clinician expertise with causal structure learning algorithms on 18,568 pediatric intensive care unit encounters, this study demonstrates that identifying plausible causal drivers of acute brain dysfunction enables the creation of parsimonious predictive models with only 14 biomarkers that achieve performance comparable to models using all 45 available variables.
This paper presents ED-Triage-Agent (ETA), a multi-agent AI framework that enhances Emergency Severity Index classification in emergency departments by mirroring clinical workflows through autonomous patient intake and collaborative, explainable decision-making that preserves clinician autonomy.
The paper presents Lilo Engine, a deterministic 5-layer therapeutic AI pipeline for elderly assisted living that replaces unreliable multi-agent systems with structural safety invariants to achieve 100% crisis recall, sub-30ms latency, and fully auditable execution paths across 3,720 test scenarios.