A5086954690- Electronic Health Records Systems
- Biomedical Text Mining and Ontologies
- Patient Safety and Medication Errors
- Healthcare Technology and Patient Monitoring
- Pharmaceutical Practices and Patient Outcomes
- Data Quality and Management
- Ethics in Clinical Research
- Artificial Intelligence in Healthcare and Education
- Health Sciences Research and Education
- Machine Learning in Healthcare
- Healthcare Systems and Technology
- Medical Coding and Health Information
- Kidney Stones and Urolithiasis Treatments
- Diabetes Management and Research
- Clinical practice guidelines implementation
- Pharmacovigilance and Adverse Drug Reactions
- Inflammatory Bowel Disease
- Pediatric Urology and Nephrology Studies
- Primary Care and Health Outcomes
- Topic Modeling
- Semantic Web and Ontologies
- COVID-19 Clinical Research Studies
- Cardiac Arrest and Resuscitation
- Telemedicine and Telehealth Implementation
- Medication Adherence and Compliance
Vanderbilt University
2008-2025
Vanderbilt University Medical Center
2009-2025
University of South Carolina
2023
Prevention of Organ Failure
2022
Tulane University
2013-2019
Dalhousie University
2018
Ochsner Health System
2014-2017
The University of Texas Health Science Center at Houston
2011-2013
Memorial Hermann
2011-2013
University of Houston
2013
To determine if ChatGPT can generate useful suggestions for improving clinical decision support (CDS) logic and to assess noninferiority compared human-generated suggestions.
To quantify and contextualize the risk for coronavirus disease 2019 (COVID-19)-related hospitalization illness severity in type 1 diabetes.
To determine if ChatGPT can generate useful suggestions for improving clinical decision support (CDS) logic and to assess noninferiority compared human-generated suggestions. We supplied summaries of CDS ChatGPT, an artificial intelligence (AI) tool question answering that uses a large language model, asked it human clinician reviewers review the AI-generated as well same alerts, rate their usefulness, acceptance, relevance, understanding, workflow, bias, inversion, redundancy. Five...
Abstract Objective This study aimed to develop and assess the performance of fine-tuned large language models for generating responses patient messages sent via an electronic health record portal. Materials Methods Utilizing a dataset extracted from portal at academic medical center, we developed model (CLAIR-Short) based on pre-trained (LLaMA-65B). In addition, used OpenAI API update physician open-source into format with informative paragraphs that offered education while emphasizing...
Alerting systems, a type of clinical decision support, are increasingly prevalent in healthcare, yet few studies have concurrently measured the appropriateness alerts with provider responses to alerts. Recent reports suboptimal alert system design and implementation highlight need for better evaluation inform future designs. The authors present comprehensive framework evaluating synchronous, interruptive medication safety alerts.Through literature review iterative testing, metrics were...
ABSTRACT Objective This study aimed to develop and assess the performance of fine-tuned large language models for generating responses patient messages sent via an electronic health record portal. Methods Utilizing a dataset extracted from portal at academic medical center, we developed model (CLAIR-Short) based on pre-trained (LLaMA-65B). In addition, used OpenAI API update physician open-source into format with informative paragraphs that offered education while emphasizing empathy...
Abstract Objective To develop and evaluate a data-driven process to generate suggestions for improving alert criteria using explainable artificial intelligence (XAI) approaches. Methods We extracted data on alerts generated from January 1, 2019 December 31, 2020, at Vanderbilt University Medical Center. developed machine learning models predict user responses alerts. applied XAI techniques global explanations local explanations. evaluated the by comparing with alert’s historical change logs...
Abstract Objectives To evaluate the capability of using generative artificial intelligence (AI) in summarizing alert comments and to determine if AI-generated summary could be used improve clinical decision support (CDS) alerts. Materials Methods We extracted user alerts generated from September 1, 2022 2023 at Vanderbilt University Medical Center. For a subset 8 alerts, comment summaries were independently by 2 physicians then separately GPT-4. surveyed 5 CDS experts rate human-generated on...
This study aims to develop and evaluate an approach using large language models (LLMs) a knowledge graph triage patient messages that need emergency care. The goal is notify patients when their indicate emergency, guiding them seek immediate help rather than the portal, improve safety. We selected 1020 sent Vanderbilt University Medical Center providers between January 1, 2022 March 7, 2023. developed four these for emergencies: (1) Prompt-Only: message was input with prompt directly into...
The impact of AKI on adverse drug events and therapeutic failures the medication errors leading to these have not been well described.A single-center observational study 396 hospitalized patients with a minimum 0.5 mg/dl change in serum creatinine who were prescribed nephrotoxic or renally eliminated was conducted. population stratified into two groups by direction their initial change: recovery. Adverse events, potential failures, for 148 drugs 46 outcomes retrospectively measured. Events...
The United States Office of the National Coordinator for Health Information Technology sponsored development a "high-priority" list drug-drug interactions (DDIs) to be used clinical decision support. We assessed current adoption this and alerting practice these DDIs with regard alert implementation (presence or absence an alert) display (alert appearance as interruptive passive).We conducted evaluations electronic health records (EHRs) at convenience sample care organizations across using...
Understanding the differences and potential synergies between traditional clinician assessment automated machine learning might enable more accurate useful suicide risk detection.
Chronic kidney disease (CKD) affects 37 million adults in the United States, and for patients with CKD, hypertension is a key risk factor adverse outcomes, such as failure, cardiovascular events, death.
Abstract Objective This study aims to investigate the feasibility of using Large Language Models (LLMs) engage with patients at time they are drafting a question their healthcare providers, and generate pertinent follow-up questions that patient can answer before sending message, goal ensuring provider receives all information need safely accurately patient’s question, eliminating back-and-forth messaging, associated delays frustrations. Methods We collected dataset messages sent between...
Objective To measure performance by eligible health care providers on CMS 's meaningful use measures. Data Source Medicare Electronic Health Record Incentive Program Eligible Professionals Public Use File ( PUF ), which contains data attestations 237,267 through May 31, 2013. Study Design Cross‐sectional analysis of the 15 core and 10 menu measures pertaining to EHR functions reported in . Principal Findings Providers dataset performed strongly all measures, with most frequent response for...
Objective To quantify the percentage of records with matching identifiers as an indicator for duplicate or potentially patient in electronic health five different healthcare organisations, describe safety issues that may arise, and present solutions managing identifiers. Methods For each institution, we retrieved deidentified counts exact match first last names dates birth determined number existing top 250 most frequently occurring name pairs. We also identified methods identifiers,...
The study sought to determine availability and use of structured override reasons for drug-drug interaction (DDI) alerts in electronic health records.