A5088607405- Artificial Intelligence in Healthcare and Education
- Machine Learning in Healthcare
- Topic Modeling
- Biomedical Text Mining and Ontologies
- Chronic Disease Management Strategies
- Radiology practices and education
- Emergency and Acute Care Studies
- Natural Language Processing Techniques
- Clinical Reasoning and Diagnostic Skills
- Electronic Health Records Systems
- Colorectal Cancer Screening and Detection
- Radiomics and Machine Learning in Medical Imaging
- Criminal Justice and Corrections Analysis
- Lung Cancer Treatments and Mutations
- Acute Myocardial Infarction Research
- Healthcare Decision-Making and Restraints
- Data-Driven Disease Surveillance
- Computational Drug Discovery Methods
- EEG and Brain-Computer Interfaces
- Text Readability and Simplification
- Urinary Tract Infections Management
- Functional Brain Connectivity Studies
- Pelvic floor disorders treatments
- Meta-analysis and systematic reviews
- Neural dynamics and brain function
Yale University
2021-2025
Case Western Reserve University
2017-2022
Harvard University
2019-2022
Background Chat Generative Pre-trained Transformer (ChatGPT) is a 175-billion-parameter natural language processing model that can generate conversation-style responses to user input. Objective This study aimed evaluate the performance of ChatGPT on questions within scope United States Medical Licensing Examination (USMLE) Step 1 and 2 exams, as well analyze for interpretability. Methods We used sets multiple-choice ChatGPT’s performance, each with pertaining 2. The first set was derived...
ABSTRACT Background ChatGPT is a 175 billion parameter natural language processing model which can generate conversation style responses to user input. Objective To evaluate the performance of on questions within scope United States Medical Licensing Examination (USMLE) Step 1 and 2 exams, as well analyze for interpretability. Methods We used two novel sets multiple choice ChatGPT’s performance, each with pertaining 2. The first was derived from AMBOSS, commonly question bank medical...
Abstract Objectives This study presents a design framework to enhance the accuracy by which large language models (LLMs), like ChatGPT can extract insights from clinical notes. We highlight this via prompt refinement for automated determination of HEART (History, ECG, Age, Risk factors, Troponin risk algorithm) scores in chest pain evaluation. Methods developed pipeline LLM testing, employing stochastic repeat testing and quantifying response errors relative physician assessment. evaluated...
To derive 7 proposed core electronic health record (EHR) use metrics across 2 healthcare systems with different EHR vendor product installations and examine factors associated time.
Abstract Background: Incarceration is a significant social determinant of health, contributing to high morbidity, mortality, and racialized health inequities. However, incarceration status largely invisible services research due inadequate clinical electronic record (EHR) capture. This study aims develop, train, validate natural language processing (NLP) techniques more effectively identify in the EHR. Methods: The population consisted adult patients (≥ 18 y.o.) who presented emergency...
Patients with a history of incarceration experience bias from health care team members, barriers to privacy, and multitude disparities. We aimed assess processes delivered in emergency departments (EDs) for people histories incarceration. utilized fine-tuned large language model identify patient status 480,374 notes the ED setting. compared socio-demographic characteristics, comorbidities, processes, including disposition, restraint use, sedation, between individuals without then conducted...
For emergency department (ED) patients, lung cancer may be detected early through incidental nodules (ILNs) discovered on chest CTs. However, there are significant errors in the communication and follow-up of findings ED imaging, particularly due to unstructured radiology reports. Natural language processing (NLP) can aid identifying ILNs requiring follow-up, potentially reducing from missed follow-up. We sought develop an open-access, three-step NLP pipeline specifically for this purpose....
Discharge instructions are a key form of documentation and patient communication in the time transition from emergency department (ED) to home. time-consuming often underprioritized, especially ED, leading discharge delays possibly impersonal instructions. Generative artificial intelligence large language models (LLMs) offer promising methods creating high-quality personalized instructions; however, there exists gap understanding perspectives LLM-generated
Natural language processing (NLP) tools including recently developed large models (LLMs) have myriad potential applications in medical care and research, the efficient labeling classification of unstructured text such as electronic health record (EHR) notes. This opens door to large-scale projects that rely on variables are not typically recorded a structured form, patient signs symptoms.
<u>A</u>dverse <u>D</u>rug <u>R</u>eactions (ADRs) are a major cause of morbidity and mortality worldwide, making post-market surveillance drugs vital for the protection public health. The <u>F</u>DA <u>E</u>vent <u>R</u>eporting <u>S</u>ystem (FAERS) is system online reporting (AE) incidents. Although FAERS data storage includes structural schema, entry large portions AE reports as unstructured free-form narratives prevalent, impeding automated monitoring these reports. To improve review...
Abstract Objectives Symptom characterization is critical to urinary tract infection (UTI) diagnosis, but identification of symptoms from the electronic health record (EHR) challenging, limiting large-scale research, public surveillance, and EHR-based clinical decision support. We therefore developed compared two natural language processing (NLP) models identify UTI unstructured emergency department (ED) notes. Methods The study population consisted patients aged ≥ 18 who presented in a...
<sec> <title>UNSTRUCTURED</title> </sec>
<sec> <title>BACKGROUND</title> Discharge instructions are a key form of documentation and patient communication in the time transition from emergency department (ED) to home. time-consuming often underprioritized, especially ED, leading discharge delays possibly impersonal instructions. Generative artificial intelligence large language models (LLMs) offer promising methods creating high-quality personalized instructions; however, there exists gap understanding perspectives LLM-generated...
<sec> <title>BACKGROUND</title> Polypharmacy, the concurrent use of multiple medications, is prevalent among older adults and associated with increased risks for adverse drug events including falls. Deprescribing, systematic process discontinuing potentially inappropriate medications (PIMs), aims to mitigate these risks. However, practical application deprescribing criteria in emergency settings remains limited due time constraints complexity criteria. </sec> <title>OBJECTIVE</title> This...
Polypharmacy, the concurrent use of multiple medications, is prevalent among older adults and associated with increased risks for adverse drug events including falls. Deprescribing, systematic process discontinuing potentially inappropriate aims to mitigate these risks. However, practical application deprescribing criteria in emergency settings remains limited due time constraints complexity. This study evaluate performance a large language model (LLM)-based pipeline identifying...
ABSTRACT Background Incarceration is a highly prevalent social determinant of health associated with high rates morbidity and mortality racialized inequities. Despite this, incarceration status largely invisible to services research due poor electronic record capture within clinical settings. Our primary objective develop assess natural language processing (NLP) techniques for identifying from notes improve sciences delivery care millions individuals impacted by incarceration. Methods We...