A5010096679- Machine Learning in Healthcare
- Topic Modeling
- Artificial Intelligence in Healthcare
- Hepatitis C virus research
- Interstitial Lung Diseases and Idiopathic Pulmonary Fibrosis
- Liver Disease and Transplantation
- Liver Disease Diagnosis and Treatment
- Technology Assessment and Management
- Biomedical Text Mining and Ontologies
- Sepsis Diagnosis and Treatment
- Healthcare Systems and Public Health
- Innovation Policy and R&D
- Hepatitis B Virus Studies
- Clinical Reasoning and Diagnostic Skills
- Lung Cancer Diagnosis and Treatment
- Pancreatic and Hepatic Oncology Research
- Healthcare Policy and Management
- Electronic Health Records Systems
- Chronic Obstructive Pulmonary Disease (COPD) Research
- Systemic Sclerosis and Related Diseases
- Emergency and Acute Care Studies
- Artificial Intelligence in Healthcare and Education
- Empathy and Medical Education
- Innovative Approaches in Technology and Social Development
- Chronic Disease Management Strategies
University of California, San Francisco
2014-2025
University of San Francisco
2023
Southern California Clinical and Translational Science Institute
2014-2020
Background and Aims: Large language models (LLMs) have significant capabilities in clinical information processing tasks. Commercially available LLMs, however, are not optimized for uses prone to generating hallucinatory information. Retrieval-augmented generation (RAG) is an enterprise architecture that allows the embedding of customized data into LLMs. This approach “specializes” LLMs thought reduce hallucinations. Approach Results We developed “LiVersa,” a liver disease–specific LLM, by...
Abstract Background Large language models (LLMs) have significant capabilities in clinical information processing tasks. Commercially available LLMs, however, are not optimized for uses and prone to generating incorrect or hallucinatory information. Retrieval-augmented generation (RAG) is an enterprise architecture that allows embedding of customized data into LLMs. This approach “specializes” the LLMs thought reduce hallucinations. Methods We developed “LiVersa,” a liver disease-specific...
PURPOSE The Electronic Medical Record Search Engine (EMERSE) is a software tool built to aid research spanning cohort discovery, population health, and data abstraction for clinical trials. EMERSE now live at three academic medical centers, with additional sites currently working on implementation. In this report, we describe how has been used support cancer based variety of metrics. METHODS We identified peer-reviewed publications that through online searches as well direct e-mails users...
Background and Aims: Diagnosis code classification is a common method for cohort identification in cirrhosis research, but it often inaccurate augmented by labor-intensive chart review. Natural language processing using large models (LLMs) potentially more accurate method. To assess LLMs’ potential identification, we compared code-based versus LLM-based with review as “gold standard.” Approach Results: We extracted conducted limited of 3788 discharge summaries admissions. engineered...
Background: Despite publicly available practice guidelines, in-hospital cirrhosis care remains highly variable. Prior studies of guideline-adherence have been limited by administrative claims data. We aimed to overcome these limitations using a novel multi-center electronic health record (EHR) database, the University California Health Data Warehouse (UCHDW), compare guideline adherence in five medical centers (UCH). Methods: identified adult patients with hospitalized 201 3 -2022. evaluated...
Ge, Jin; Sun, Steve; Owens, Joseph; Galvez, Victor; Gologorskaya, Oksana; Lai, Jennifer C.; Pletcher, Mark J.; Ki Author Information
To fully meet their mission, Clinical and Translational Science Awards (CTSAs) cannot rely on old models for supporting research. Sure, providing services that investigators have come to expect is a critical role CTSAs, but current systems of research are obviously broken, with costs only increasing barriers delivering health improvements growing over the past several decades. Clearly, new executing clinical translational required. Top-down approaches innovation advantages, including...
OBJECTIVES: To develop proof-of-concept algorithms using alternative approaches to capture provider sentiment in ICU notes. DESIGN: Retrospective observational cohort study. SETTING: The Multiparameter Intelligent Monitoring of Intensive Care III (MIMIC-III) and the University California, San Francisco (UCSF) deidentified notes databases. PATIENTS: Adult (≥18 yr old) patients admitted ICU. MEASUREMENTS AND MAIN RESULTS: We developed two models: 1) a keywords-based approach consensus-based...
Introduction: Despite publicly available practice guidelines published by AASLD and AGA, cirrhosis care remains highly variable. Investigating variations in guidelines-adherence could identify key opportunities to improve clinical outcomes, but prior studies have been limited the lack of patient-level data. We aimed overcome these limitations leveraging a novel multicenter EHR database, University California Health Data Warehouse (UCHDW), with high-dimensional data quantify adherence...