- Home
- 2025 Annual Symposium Gallery
- Rapid Review of Models Assessing Suicide Risk from Patient Portal and Crisis Text Line Messages
Custom CSS
double-click to edit, do not edit in source
11/16/2025 |
5:30 PM – 7:00 PM |
Room 13
Poster Session 1
Presentation Type: Posters
Decoding STEMI Team Performance: EHR Audit Log Insights
Poster Number: P01
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Workflow, Critical Care, Clinical Decision Support, Healthcare Quality, Informatics Implementation
Primary Track: Applications
Programmatic Theme: Clinical Informatics
Electronic healthcare record (EHR) audit logs data provides a scalable approach to measuring team-level contextual factors that influence care outcomes. Analysis of STEMI cases revealed that recent shared team experience reduced door-to-balloon times, while greater physical movement and charting activity delayed treatment. These results highlight opportunities to improve STEMI care efficiency through reducing task switching, minimizing physical dispersion, and leveraging shared team experience.
Speaker:
Antra Nakhasi, MS
School of Medicine, Stanford University
Authors:
Ariadna Garcia, MS - Department of Medicine, Quantitative Sciences Unit, Stanford University School of Medicine; Antra Nakhasi, MS - School of Medicine, Stanford University; Carl Preiksaitis, MD - Department of Emergency Medicine, Stanford University School of Medicine; Nidia Rodriguez-Ormaza, MD, PhD - Department of Medicine, Quantitative Sciences Unit, Stanford University School of Medicine; Christian Rose, MD - Stanford University, School of Medicine;
Poster Number: P01
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Workflow, Critical Care, Clinical Decision Support, Healthcare Quality, Informatics Implementation
Primary Track: Applications
Programmatic Theme: Clinical Informatics
Electronic healthcare record (EHR) audit logs data provides a scalable approach to measuring team-level contextual factors that influence care outcomes. Analysis of STEMI cases revealed that recent shared team experience reduced door-to-balloon times, while greater physical movement and charting activity delayed treatment. These results highlight opportunities to improve STEMI care efficiency through reducing task switching, minimizing physical dispersion, and leveraging shared team experience.
Speaker:
Antra Nakhasi, MS
School of Medicine, Stanford University
Authors:
Ariadna Garcia, MS - Department of Medicine, Quantitative Sciences Unit, Stanford University School of Medicine; Antra Nakhasi, MS - School of Medicine, Stanford University; Carl Preiksaitis, MD - Department of Emergency Medicine, Stanford University School of Medicine; Nidia Rodriguez-Ormaza, MD, PhD - Department of Medicine, Quantitative Sciences Unit, Stanford University School of Medicine; Christian Rose, MD - Stanford University, School of Medicine;
Antra
Nakhasi,
MS - School of Medicine, Stanford University
Effect of Physician-Directed Appointment Slots on EHR Workload: A Controlled Interrupted Time Series Study
Poster Number: P02
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Workflow, Patient Engagement and Preferences, Population Health
Primary Track: Policy
Programmatic Theme: Clinical Informatics
Allocating one appointment slot per half-day for asynchronous tasks reduced after-hours EHR time and patient message volume among ambulatory internal medicine physicians, with minimal impact on productivity(RVU). Using a controlled interrupted time series and difference-in-differences analysis across two health systems, this study found that structured asynchronous time may support physician well-being and workflow efficiency. These findings highlight the potential of operational changes to address digital burden while sustaining care delivery under existing reimbursement models.
Speaker:
Yu Sun, Master of Science
Yale
Authors:
Naga Sasidhar Kanaparthy, MD MPH - Yale University; Edward Melnick, MD - Yale University, School of Medicine; A J Holmgren, PhD - University of California, San Francisco; Yu Sun, Master of Science - Yale;
Poster Number: P02
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Workflow, Patient Engagement and Preferences, Population Health
Primary Track: Policy
Programmatic Theme: Clinical Informatics
Allocating one appointment slot per half-day for asynchronous tasks reduced after-hours EHR time and patient message volume among ambulatory internal medicine physicians, with minimal impact on productivity(RVU). Using a controlled interrupted time series and difference-in-differences analysis across two health systems, this study found that structured asynchronous time may support physician well-being and workflow efficiency. These findings highlight the potential of operational changes to address digital burden while sustaining care delivery under existing reimbursement models.
Speaker:
Yu Sun, Master of Science
Yale
Authors:
Naga Sasidhar Kanaparthy, MD MPH - Yale University; Edward Melnick, MD - Yale University, School of Medicine; A J Holmgren, PhD - University of California, San Francisco; Yu Sun, Master of Science - Yale;
Yu
Sun,
Master of Science - Yale
Design requirements of a clinical decision support tool interface for machine learning models that detect clinical deterioration for patients with COVID-19
Poster Number: P03
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: User-centered Design Methods, Information Visualization, Clinical Decision Support
Primary Track: Applications
Early detection of clinical deterioration of patients with COVID-19 can improve outcomes. Although machine learning models can aid with these tasks, it is unknown how to best present these models to clinicians in a user interface. Therefore, we conducted surveys and focus groups among physicians and nurses who care for critically ill patients to explore the design requirements for these tools. Our findings will be used to develop prototypes for usability testing.
Speaker:
Oliver Nguyen, MSHI
University of Wisconsin at Madison
Authors:
Arsalan Ahmad, MS - University of Wisconsin at Madison; Madeline Oguss, MS - University of Wisconsin at Madison; Jonathan Allan, BS - AgileMD; Joseph Reid, MSN, RN, CCRN - AgileMD; Dana Edelson, MD - University of Chicago; Douglas Wiegmann, PhD - University of Wisconsin at Madison; Sushant Govindan, MD - Kansas City Veterans Affairs Medical Center; Matthew Churpek, MD, MPH, PhD - University of Wisconsin-Madison;
Poster Number: P03
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: User-centered Design Methods, Information Visualization, Clinical Decision Support
Primary Track: Applications
Early detection of clinical deterioration of patients with COVID-19 can improve outcomes. Although machine learning models can aid with these tasks, it is unknown how to best present these models to clinicians in a user interface. Therefore, we conducted surveys and focus groups among physicians and nurses who care for critically ill patients to explore the design requirements for these tools. Our findings will be used to develop prototypes for usability testing.
Speaker:
Oliver Nguyen, MSHI
University of Wisconsin at Madison
Authors:
Arsalan Ahmad, MS - University of Wisconsin at Madison; Madeline Oguss, MS - University of Wisconsin at Madison; Jonathan Allan, BS - AgileMD; Joseph Reid, MSN, RN, CCRN - AgileMD; Dana Edelson, MD - University of Chicago; Douglas Wiegmann, PhD - University of Wisconsin at Madison; Sushant Govindan, MD - Kansas City Veterans Affairs Medical Center; Matthew Churpek, MD, MPH, PhD - University of Wisconsin-Madison;
Oliver
Nguyen,
MSHI - University of Wisconsin at Madison
Detecting and Limiting Fraudulent Survey Responses in REDCap
Poster Number: P04
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Surveys and Needs Analysis, Administrative Systems, Informatics Implementation
Primary Track: Applications
Programmatic Theme: Clinical Research Informatics
Fraudulent survey responses pose challenges to electronic data collection tools like REDCap Survey, especially when incentives are offered. This study reviews solutions for mitigating fraudulent responses, categorized as Technical Controls, Study Design & Process Controls, and Participant & Verification Controls. Each approach, such as Google reCAPTCHA and identity verification, has distinct advantages and drawbacks. Researchers are encouraged to balance these trade-offs to protect data integrity while minimizing participant deterrence.
Speaker:
David Hanauer, MD
University of Michigan
Authors:
Andrew Carroll, AS - University of Michigan; James Maszatics - MICHR/University of Michigan; David Hanauer, MD - University of Michigan;
Poster Number: P04
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Surveys and Needs Analysis, Administrative Systems, Informatics Implementation
Primary Track: Applications
Programmatic Theme: Clinical Research Informatics
Fraudulent survey responses pose challenges to electronic data collection tools like REDCap Survey, especially when incentives are offered. This study reviews solutions for mitigating fraudulent responses, categorized as Technical Controls, Study Design & Process Controls, and Participant & Verification Controls. Each approach, such as Google reCAPTCHA and identity verification, has distinct advantages and drawbacks. Researchers are encouraged to balance these trade-offs to protect data integrity while minimizing participant deterrence.
Speaker:
David Hanauer, MD
University of Michigan
Authors:
Andrew Carroll, AS - University of Michigan; James Maszatics - MICHR/University of Michigan; David Hanauer, MD - University of Michigan;
David
Hanauer,
MD - University of Michigan
Exploring User Needs and Feature Preferences for the Development of Virtual Study Assistants
Poster Number: P05
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Surveys and Needs Analysis, Artificial Intelligence, User-centered Design Methods
Primary Track: Foundations
The purpose of the study was to inform the development of a Virtual Study Assistant (VSA) by exploring user needs and preferences through focus groups and surveys. The study identified the highest-priority features and assessed acceptability and user preferences. The results offer information on which features are most valued, thereby providing a foundation for developing the VSA prototype.
Speaker:
Chi-shan Tsai, MSN
University of Washington
Authors:
Hyunhae Lee, MSN - University of Washington; Warren Szewczyk, BA - University of Washington; Julia Palmer, Research Coordinator - University of Washington; Sophie Putnam, Student - University of Washington; Sean Munson, PhD - University of Washington; Jaimee Heffner, PhD - Fred Hutch Cancer Center; Alexi Vasbinder, PhD - University of Washington; Amandalynne Paullada; Weichao Yuwen, PhD, RN - University of Washington Tacoma; Kerry Reding, PhD - University of Washington;
Poster Number: P05
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Surveys and Needs Analysis, Artificial Intelligence, User-centered Design Methods
Primary Track: Foundations
The purpose of the study was to inform the development of a Virtual Study Assistant (VSA) by exploring user needs and preferences through focus groups and surveys. The study identified the highest-priority features and assessed acceptability and user preferences. The results offer information on which features are most valued, thereby providing a foundation for developing the VSA prototype.
Speaker:
Chi-shan Tsai, MSN
University of Washington
Authors:
Hyunhae Lee, MSN - University of Washington; Warren Szewczyk, BA - University of Washington; Julia Palmer, Research Coordinator - University of Washington; Sophie Putnam, Student - University of Washington; Sean Munson, PhD - University of Washington; Jaimee Heffner, PhD - Fred Hutch Cancer Center; Alexi Vasbinder, PhD - University of Washington; Amandalynne Paullada; Weichao Yuwen, PhD, RN - University of Washington Tacoma; Kerry Reding, PhD - University of Washington;
Chi-shan
Tsai,
MSN - University of Washington
Adapting NICU Documentation through Process Engineering
Poster Number: P06
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Surveys and Needs Analysis, Workflow, User-centered Design Methods, Nursing Informatics, Pediatrics
Primary Track: Applications
Programmatic Theme: Clinical Informatics
Aims to understand the impact of migrating multiple EHR systems to a single system within the Neonatal Intensive Care Unit (NICU) by identifying key processes and clinical data to integrate roles and core documentation. A process engineering approach mapped current state, future state, gaps, opportunities for optimization and ongoing performance measurement. Findings illuminate need to optimize the EHR to improve usability, optimize clinical data visualization, and integrate clinical decision support tailored to the NICU population.
Speaker:
Rachel Buchleiter, MSN, RN, RN-BC
University of Utah
Authors:
Rachel Buchleiter, MSN, RN, RN-BC - University of Utah; Kathryn Price, MEng - HCA Healthcare; Jani Bowen, MISE - HCA Healthcare;
Poster Number: P06
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Surveys and Needs Analysis, Workflow, User-centered Design Methods, Nursing Informatics, Pediatrics
Primary Track: Applications
Programmatic Theme: Clinical Informatics
Aims to understand the impact of migrating multiple EHR systems to a single system within the Neonatal Intensive Care Unit (NICU) by identifying key processes and clinical data to integrate roles and core documentation. A process engineering approach mapped current state, future state, gaps, opportunities for optimization and ongoing performance measurement. Findings illuminate need to optimize the EHR to improve usability, optimize clinical data visualization, and integrate clinical decision support tailored to the NICU population.
Speaker:
Rachel Buchleiter, MSN, RN, RN-BC
University of Utah
Authors:
Rachel Buchleiter, MSN, RN, RN-BC - University of Utah; Kathryn Price, MEng - HCA Healthcare; Jani Bowen, MISE - HCA Healthcare;
Rachel
Buchleiter,
MSN, RN, RN-BC - University of Utah
Towards Reporting Standards for Digital Health Technology-enabled Randomized Controlled Trials: A Modified-Delphi Study
Poster Number: P07
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Standards, Clinical Guidelines, Health Equity
Primary Track: Foundations
Programmatic Theme: Clinical Informatics
Clinical trial reporting guidelines aim to improve transparent and reproducible documentation of findings. However,
the rapid growth and diverse innovation of digital health technology-enabled randomized controlled trials present
reporting standardization challenges. To address this gap, we generated candidate reporting guideline items and
conducted a modified-Delphi process at the AMIA 2024 Annual Symposium. Twenty-five experts participated and the 80% consensus threshold was achieved for eight digital health technology-enabled randomized controlled trial items.
Speaker:
Taylor Harrison, MS, MBS
Mayo Clinic
Authors:
Hongfang Liu, PhD - University of Texas Health Science Center at Houston; Sunyang Fu, PhD, MHI - UTHealth;
Poster Number: P07
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Standards, Clinical Guidelines, Health Equity
Primary Track: Foundations
Programmatic Theme: Clinical Informatics
Clinical trial reporting guidelines aim to improve transparent and reproducible documentation of findings. However,
the rapid growth and diverse innovation of digital health technology-enabled randomized controlled trials present
reporting standardization challenges. To address this gap, we generated candidate reporting guideline items and
conducted a modified-Delphi process at the AMIA 2024 Annual Symposium. Twenty-five experts participated and the 80% consensus threshold was achieved for eight digital health technology-enabled randomized controlled trial items.
Speaker:
Taylor Harrison, MS, MBS
Mayo Clinic
Authors:
Hongfang Liu, PhD - University of Texas Health Science Center at Houston; Sunyang Fu, PhD, MHI - UTHealth;
Taylor
Harrison,
MS, MBS - Mayo Clinic
Performance of a Computable Phenotype to Identify Patients with Stimulant Use Disorder
Poster Number: P08
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Real-World Evidence Generation, Chronic Care Management, Population Health
Primary Track: Foundations
Programmatic Theme: Clinical Research Informatics
People with stimulant use disorder (StimUD) often struggle with co-occurring use disorders that increase the risk of overdose and poor treatment outcomes. Study of use disorder treatment is limited by the lack of validated algorithms to identify those with StimUD in real world data. Thus, we used electronic health record data to validate an algorithm to identify patients with StimUD. Findings from this study provide a method for accurately identifying and studying patients with StimUD.
Speaker:
Thomas Reese
Department of Biomedical Informatics, Vanderbilt University
Authors:
Mauli Shah, MPH - Vanderbilt University Medical Center; Stephen Patrick, MD, MS - Emory University; Ashley Leech, PhD - Vanderbilt University Medical Center; Bryan Steitz, PhD - Vanderbilt University Medical Center; Michael Matheny, MD, MS, MPH, FACMI, FAMIA - Vanderbilt University Medical Center; Adam Wright, PhD - Vanderbilt University Medical Center; Andrew Wiese, PhD - Vanderbilt University Medical Center;
Poster Number: P08
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Real-World Evidence Generation, Chronic Care Management, Population Health
Primary Track: Foundations
Programmatic Theme: Clinical Research Informatics
People with stimulant use disorder (StimUD) often struggle with co-occurring use disorders that increase the risk of overdose and poor treatment outcomes. Study of use disorder treatment is limited by the lack of validated algorithms to identify those with StimUD in real world data. Thus, we used electronic health record data to validate an algorithm to identify patients with StimUD. Findings from this study provide a method for accurately identifying and studying patients with StimUD.
Speaker:
Thomas Reese
Department of Biomedical Informatics, Vanderbilt University
Authors:
Mauli Shah, MPH - Vanderbilt University Medical Center; Stephen Patrick, MD, MS - Emory University; Ashley Leech, PhD - Vanderbilt University Medical Center; Bryan Steitz, PhD - Vanderbilt University Medical Center; Michael Matheny, MD, MS, MPH, FACMI, FAMIA - Vanderbilt University Medical Center; Adam Wright, PhD - Vanderbilt University Medical Center; Andrew Wiese, PhD - Vanderbilt University Medical Center;
Thomas
Reese - Department of Biomedical Informatics, Vanderbilt University
A Real-World Examination of CAR-T Patient Characteristics and Treatment-related Outcomes
Poster Number: P09
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Real-World Evidence Generation, Clinical Guidelines, Drug Discoveries, Repurposing, and Side-effect
Primary Track: Foundations
Programmatic Theme: Clinical Research Informatics
To date, information related to CAR-T-related adverse events has been limited to case reports and retrospective studies, the majority of which have not tracked post-treatment second primary malignancies or mortality. This exploratory analysis draws on real-world data to address the knowledge gap, profiling key characteristics of the “average” CAR-T patient and comparing rates of key outcomes between and among CAR-T therapies.
Speaker:
Nathan Markward, PhD, MPH
PurpleLab
Authors:
Nathan Markward, PhD, MPH - PurpleLab; Douglas Londono, PhD - PurpleLab; Daniel Lemberg, BS - PurpleLab; Allison Brosso, BA - PurpleLab; Diane Faraone, PharmD - PurpleLab;
Poster Number: P09
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Real-World Evidence Generation, Clinical Guidelines, Drug Discoveries, Repurposing, and Side-effect
Primary Track: Foundations
Programmatic Theme: Clinical Research Informatics
To date, information related to CAR-T-related adverse events has been limited to case reports and retrospective studies, the majority of which have not tracked post-treatment second primary malignancies or mortality. This exploratory analysis draws on real-world data to address the knowledge gap, profiling key characteristics of the “average” CAR-T patient and comparing rates of key outcomes between and among CAR-T therapies.
Speaker:
Nathan Markward, PhD, MPH
PurpleLab
Authors:
Nathan Markward, PhD, MPH - PurpleLab; Douglas Londono, PhD - PurpleLab; Daniel Lemberg, BS - PurpleLab; Allison Brosso, BA - PurpleLab; Diane Faraone, PharmD - PurpleLab;
Nathan
Markward,
PhD, MPH - PurpleLab
Boosting Suicide Risk Prediction with Social and Behavioral Factors
Poster Number: P10
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Real-World Evidence Generation, Information Extraction, Natural Language Processing, Machine Learning
Primary Track: Applications
Programmatic Theme: Clinical Informatics
This study examined the impact of integrating individual-level social and behavioral factors (SBFs) into suicide risk prediction models for patients presenting to the emergency department with suicidal ideation. Using NLP-extracted SBFs from clinical notes, models achieved up to a three-fold increase in AUPRC and two-fold increase in PPV compared to clinical-only model. Homelessness, chronic stress, and adverse childhood experiences were the strongest predictors. Findings highlight the value of SBFs in improving suicide risk prediction.
Speaker:
Hyunjoon Lee, MS
Vanderbilt University Medical Center DBMI
Authors:
Hyunjoon Lee, MS - Vanderbilt University Medical Center DBMI; Michael Ripperger - Vanderbilt University Medical Center; Ketan Jadhav, MPS - Vanderbilt University Medical Center; Samuel Palmer, BS - Vanderbilt University; Peyton Coleman, BS - Vanderbilt University; Cosmin Bejan, PhD - Vanderbilt University Medical Center; Douglas Ruderfer - Vanderbilt University Medical Center; Colin Walsh, MD MA - Department of Biomedical Informatics, Vanderbilt University;
Poster Number: P10
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Real-World Evidence Generation, Information Extraction, Natural Language Processing, Machine Learning
Primary Track: Applications
Programmatic Theme: Clinical Informatics
This study examined the impact of integrating individual-level social and behavioral factors (SBFs) into suicide risk prediction models for patients presenting to the emergency department with suicidal ideation. Using NLP-extracted SBFs from clinical notes, models achieved up to a three-fold increase in AUPRC and two-fold increase in PPV compared to clinical-only model. Homelessness, chronic stress, and adverse childhood experiences were the strongest predictors. Findings highlight the value of SBFs in improving suicide risk prediction.
Speaker:
Hyunjoon Lee, MS
Vanderbilt University Medical Center DBMI
Authors:
Hyunjoon Lee, MS - Vanderbilt University Medical Center DBMI; Michael Ripperger - Vanderbilt University Medical Center; Ketan Jadhav, MPS - Vanderbilt University Medical Center; Samuel Palmer, BS - Vanderbilt University; Peyton Coleman, BS - Vanderbilt University; Cosmin Bejan, PhD - Vanderbilt University Medical Center; Douglas Ruderfer - Vanderbilt University Medical Center; Colin Walsh, MD MA - Department of Biomedical Informatics, Vanderbilt University;
Hyunjoon
Lee,
MS - Vanderbilt University Medical Center DBMI
A Computer Vision Approach for Melanopsin-Derived Pupillary Light Reflex Analysis in Parkinson’s Disease Detection
Poster Number: P11
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Quantitative Methods, Evaluation, Imaging Informatics, Diagnostic Systems, Clinical Decision Support
Primary Track: Applications
Programmatic Theme: Clinical Informatics
This study presents a novel computer vision algorithm that automates the detection and quantification of melanopsin-mediated pupillary light reflex (PLR) abnormalities. By analyzing high-resolution video recordings, the algorithm accurately measures pupil dynamics and identifies key metrics like dilation velocity, and recovery latency. This non-invasive tool demonstrates potential for improving Parkinson’s disease diagnostics, monitoring disease progression, and guiding personalized treatment strategies in both research and clinical settings.
Speaker:
Dhruvin Patel, Computer Science
Loyola University Chicago
Authors:
Dhruvin Patel, BSc - Loyola University Chicago; Alexandria Umbarger, BS - Edward Hines Jr. VA Medical Center; Sandra Kletzel, PhD - Edward Hines Jr. VA Medical Center; Bruce Gaynes, O.D., PharmD - Loyola University Chicago; Samie Tootooni, PhD - Loyola University Chicago;
Poster Number: P11
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Quantitative Methods, Evaluation, Imaging Informatics, Diagnostic Systems, Clinical Decision Support
Primary Track: Applications
Programmatic Theme: Clinical Informatics
This study presents a novel computer vision algorithm that automates the detection and quantification of melanopsin-mediated pupillary light reflex (PLR) abnormalities. By analyzing high-resolution video recordings, the algorithm accurately measures pupil dynamics and identifies key metrics like dilation velocity, and recovery latency. This non-invasive tool demonstrates potential for improving Parkinson’s disease diagnostics, monitoring disease progression, and guiding personalized treatment strategies in both research and clinical settings.
Speaker:
Dhruvin Patel, Computer Science
Loyola University Chicago
Authors:
Dhruvin Patel, BSc - Loyola University Chicago; Alexandria Umbarger, BS - Edward Hines Jr. VA Medical Center; Sandra Kletzel, PhD - Edward Hines Jr. VA Medical Center; Bruce Gaynes, O.D., PharmD - Loyola University Chicago; Samie Tootooni, PhD - Loyola University Chicago;
Dhruvin
Patel,
Computer Science - Loyola University Chicago
Detecting Deterioration: Qualitative Study of Emergency Department Communication Patterns
Poster Number: P12
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Qualitative Methods, Clinical Decision Support, Healthcare Quality, Artificial Intelligence
Primary Track: Applications
Programmatic Theme: Clinical Informatics
Timely recognition and response to patient deterioration in the emergency department (ED) depend on effective team communication, yet clinicians’ preferences and systemic barriers often hinder this communication. Through focus groups and interviews, we identified differences in communication methods, documentation gaps, and resource constraints, such as limited monitored beds and staff, that impact ED care. Our findings highlight opportunities to optimize communication channels and support detection of deterioration through an early warning system.
Speaker:
Yu-Hsiang Lo, MD
NewYork-Presbyterian/Columbia University Irving Medical Center
Authors:
Laura Henze, MD, MA - Columbia University; Temiloluwa Daramola, BA - Columbia University Irving Medical Center - Department of Biomedical Informatics; Rachel Lee, PhD, RN - Columbia University; Xuhai Xu, PhD - Columbia University; Richard Trepp, MD - NewYork-Presbyterian; Kenrick Cato, PhD, RN, CPHIMS, FAAN, FACMI - University of Pennsylvania/ Children's Hospital of Philadelphia; Sarah Rossetti, RN, PhD - Columbia University Department of Biomedical Informatics;
Poster Number: P12
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Qualitative Methods, Clinical Decision Support, Healthcare Quality, Artificial Intelligence
Primary Track: Applications
Programmatic Theme: Clinical Informatics
Timely recognition and response to patient deterioration in the emergency department (ED) depend on effective team communication, yet clinicians’ preferences and systemic barriers often hinder this communication. Through focus groups and interviews, we identified differences in communication methods, documentation gaps, and resource constraints, such as limited monitored beds and staff, that impact ED care. Our findings highlight opportunities to optimize communication channels and support detection of deterioration through an early warning system.
Speaker:
Yu-Hsiang Lo, MD
NewYork-Presbyterian/Columbia University Irving Medical Center
Authors:
Laura Henze, MD, MA - Columbia University; Temiloluwa Daramola, BA - Columbia University Irving Medical Center - Department of Biomedical Informatics; Rachel Lee, PhD, RN - Columbia University; Xuhai Xu, PhD - Columbia University; Richard Trepp, MD - NewYork-Presbyterian; Kenrick Cato, PhD, RN, CPHIMS, FAAN, FACMI - University of Pennsylvania/ Children's Hospital of Philadelphia; Sarah Rossetti, RN, PhD - Columbia University Department of Biomedical Informatics;
Yu-Hsiang
Lo,
MD - NewYork-Presbyterian/Columbia University Irving Medical Center
Almost Everything is Related to Oral Health: A Systematic Investigation of Systemic-Oral Health Connections
Poster Number: P13
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Public Health, Informatics Implementation, Data Mining
Primary Track: Applications
Programmatic Theme: Public Health Informatics
The connection between systemic and oral health is well established, yet research remains fragmented, often focusing on isolated conditions. This study systematically investigated the extent of published evidence linking systemic diseases to dental conditions. Using AHRQ’s Clinical Classification Software (CCS), 258 distinct medical conditions were identified, and structured PubMed searches were conducted for each, pairing CCS-coded conditions with common dental conditions (periodontitis, dental caries, tooth loss, oral ulcers). A total of 6,364 publications were retrieved; 191 CCS-coded conditions had documented associations with oral health. The most frequently studied were diabetes (757 articles), Parkinson’s disease (455), and metabolic disorders (197). No literature was found for 67 conditions, including limb fractures and acute bronchitis. These findings highlight the broad systemic–oral health linkages reported in the literature and underscore the need for integrated research approaches and further exploration of under-studied conditions.
Speaker:
Bhumi Patel, Ph.D. in Health Informatics
George Mason University
Author:
Janusz Wojtusiak, PhD - George Mason University;
Poster Number: P13
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Public Health, Informatics Implementation, Data Mining
Primary Track: Applications
Programmatic Theme: Public Health Informatics
The connection between systemic and oral health is well established, yet research remains fragmented, often focusing on isolated conditions. This study systematically investigated the extent of published evidence linking systemic diseases to dental conditions. Using AHRQ’s Clinical Classification Software (CCS), 258 distinct medical conditions were identified, and structured PubMed searches were conducted for each, pairing CCS-coded conditions with common dental conditions (periodontitis, dental caries, tooth loss, oral ulcers). A total of 6,364 publications were retrieved; 191 CCS-coded conditions had documented associations with oral health. The most frequently studied were diabetes (757 articles), Parkinson’s disease (455), and metabolic disorders (197). No literature was found for 67 conditions, including limb fractures and acute bronchitis. These findings highlight the broad systemic–oral health linkages reported in the literature and underscore the need for integrated research approaches and further exploration of under-studied conditions.
Speaker:
Bhumi Patel, Ph.D. in Health Informatics
George Mason University
Author:
Janusz Wojtusiak, PhD - George Mason University;
Bhumi
Patel,
Ph.D. in Health Informatics - George Mason University
AI chatbot for connecting low-income patients and caregivers with community resources
Poster Number: P14
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Public Health, Large Language Models (LLMs), Pediatrics, Surveys and Needs Analysis, Personal Health Informatics
Primary Track: Applications
Programmatic Theme: Public Health Informatics
This study evaluates an AI chatbot designed to help low-income caregivers navigate community resources for shelter, food, transportation, and financial aid. Built with rule-based logic and generative AI, the chatbot provided real-time, location-based recommendations. Findings show high usability, low cognitive load and strong user trust. Participants valued its efficiency but suggested more conversational customization. Results highlight its potential to enhance access to essential services and inform future clinical integration
Speaker:
Emre Sezgin, PhD
Nationwide Children's Hospital / The Ohio State University College of Medicine
Authors:
Daniel Jackson, B.Sc. - Nationwide Childrens Hospital at Abigail Wexner Research Institute; Syed-Amad Hussain, BSE - Nationwide Children's Hospital; A. Baki Kocaballi, PhD - University of Technology Sidney, Sidney, Australia;
Poster Number: P14
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Public Health, Large Language Models (LLMs), Pediatrics, Surveys and Needs Analysis, Personal Health Informatics
Primary Track: Applications
Programmatic Theme: Public Health Informatics
This study evaluates an AI chatbot designed to help low-income caregivers navigate community resources for shelter, food, transportation, and financial aid. Built with rule-based logic and generative AI, the chatbot provided real-time, location-based recommendations. Findings show high usability, low cognitive load and strong user trust. Participants valued its efficiency but suggested more conversational customization. Results highlight its potential to enhance access to essential services and inform future clinical integration
Speaker:
Emre Sezgin, PhD
Nationwide Children's Hospital / The Ohio State University College of Medicine
Authors:
Daniel Jackson, B.Sc. - Nationwide Childrens Hospital at Abigail Wexner Research Institute; Syed-Amad Hussain, BSE - Nationwide Children's Hospital; A. Baki Kocaballi, PhD - University of Technology Sidney, Sidney, Australia;
Emre
Sezgin,
PhD - Nationwide Children's Hospital / The Ohio State University College of Medicine
Supporting Cybersecurity-aware Technology Adoption Decision-making among Patients in an Increasingly Complex Digital Health Ecosystem
Poster Number: P15
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Privacy and Security, Human-computer Interaction, Chronic Care Management, Large Language Models (LLMs), Personal Health Informatics
Primary Track: Foundations
Programmatic Theme: Consumer Health Informatics
We examine cybersecurity implications in the digital health ecosystem (DHE) that patients rely on for their health, including health apps and wearables. Our findings, based on a literature review and interviews with five health information technology (HIT) professionals and 25 patients, highlight cybersecurity risks in the DHE and how these risks affect patients' decisions to adopt HIT. This study is a first step towards co-creating a toolkit to help patients make secure and informed decisions.
Speaker:
Zainab Balogun, MS, MA
University of Maryland Baltimore County
Authors:
Melissa Carraway, MS - University of Maryland Baltimore County; Tera Reynolds, PhD, MPH, MS - University of Maryland Baltimore County;
Poster Number: P15
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Privacy and Security, Human-computer Interaction, Chronic Care Management, Large Language Models (LLMs), Personal Health Informatics
Primary Track: Foundations
Programmatic Theme: Consumer Health Informatics
We examine cybersecurity implications in the digital health ecosystem (DHE) that patients rely on for their health, including health apps and wearables. Our findings, based on a literature review and interviews with five health information technology (HIT) professionals and 25 patients, highlight cybersecurity risks in the DHE and how these risks affect patients' decisions to adopt HIT. This study is a first step towards co-creating a toolkit to help patients make secure and informed decisions.
Speaker:
Zainab Balogun, MS, MA
University of Maryland Baltimore County
Authors:
Melissa Carraway, MS - University of Maryland Baltimore County; Tera Reynolds, PhD, MPH, MS - University of Maryland Baltimore County;
Zainab
Balogun,
MS, MA - University of Maryland Baltimore County
The Role of Digital Biomarkers for Insulin Resistance in Predicting Fatty Liver, and Type 2 Diabetes: Evidence from the Taiwan Biobank
Poster Number: P16
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Precision Medicine, Clinical Guidelines, Public Health, Machine Learning, Artificial Intelligence
Primary Track: Applications
Programmatic Theme: Public Health Informatics
Introduction and Objective: Insulin resistance (IR) plays a crucial role in the development of fatty liver disease (FLD) and type 2 diabetes mellitus (T2DM). This study aims to assess the predictive accuracy of an Artificial Intelligence-based IR index (AI-IR) as a digital biomarker for fatty liver (FL) and T2DM, and compares its performance with that of established tools such as the Fatty Liver Score (FLS) and IR-related indices.
Methods: Data from 23,517 participants in the Taiwan Biobank, free of diabetes mellitus (DM) at baseline and undergoing abdominal ultrasound during follow-up, were analyzed. A novel AI-based IR biomarker (AI-IR) was derived using the XGBoost algorithm. FLD was diagnosed via ultrasound, and FLS was evaluated using indices such as FLI, HSI, and FIB-4. Seven IR-related indices, including TyG-BMI, TyG-WC, and TG/HDL, were assessed. Diagnostic performance was evaluated using receiver operating characteristic (ROC) curves and the area under the curve (AUC).
Results: The AI-IR digital biomarker outperformed other FLS and IR-related indices for predicting incident T2DM (AUC: AI-IR, 0.80; K-NAFLD, 0.76; TyG-WC, 0.74; FLI, 0.74; TyG-BMI, 0.74; LAP, 0.73; TyG, 0.73; all p < 0.001). For predicting FL, AI-IR (AUC = 0.734) performed similarly to FLI and TyG-BMI (AUC = 0.737 and 0.734, respectively; p > 0.05), and surpassed other indices like TyG-WC, LAP, HSI, and KNAFLD (all p < 0.001).
Conclusion: AI-IR, a digital biomarker using nine accessible features, excels in predicting FL and T2DM, supporting early detection and personalized risk evaluation.
Speaker:
Chia-Lin Lee, Attending physician/MD.PhD
Division of Artificial Intelligence, Department of Digital Medicine, Taichung Veterans General Hospital, Taichung, Taiwan.
Authors:
Wei-Ju Liu, Research assistant/Master - Taichung Veterans General Hospital, Taichung, Taiwan; I-Hsin Huang, Research assistant/Master - Taichung Veterans General Hospital, Taichung, Taiwan.;
Poster Number: P16
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Precision Medicine, Clinical Guidelines, Public Health, Machine Learning, Artificial Intelligence
Primary Track: Applications
Programmatic Theme: Public Health Informatics
Introduction and Objective: Insulin resistance (IR) plays a crucial role in the development of fatty liver disease (FLD) and type 2 diabetes mellitus (T2DM). This study aims to assess the predictive accuracy of an Artificial Intelligence-based IR index (AI-IR) as a digital biomarker for fatty liver (FL) and T2DM, and compares its performance with that of established tools such as the Fatty Liver Score (FLS) and IR-related indices.
Methods: Data from 23,517 participants in the Taiwan Biobank, free of diabetes mellitus (DM) at baseline and undergoing abdominal ultrasound during follow-up, were analyzed. A novel AI-based IR biomarker (AI-IR) was derived using the XGBoost algorithm. FLD was diagnosed via ultrasound, and FLS was evaluated using indices such as FLI, HSI, and FIB-4. Seven IR-related indices, including TyG-BMI, TyG-WC, and TG/HDL, were assessed. Diagnostic performance was evaluated using receiver operating characteristic (ROC) curves and the area under the curve (AUC).
Results: The AI-IR digital biomarker outperformed other FLS and IR-related indices for predicting incident T2DM (AUC: AI-IR, 0.80; K-NAFLD, 0.76; TyG-WC, 0.74; FLI, 0.74; TyG-BMI, 0.74; LAP, 0.73; TyG, 0.73; all p < 0.001). For predicting FL, AI-IR (AUC = 0.734) performed similarly to FLI and TyG-BMI (AUC = 0.737 and 0.734, respectively; p > 0.05), and surpassed other indices like TyG-WC, LAP, HSI, and KNAFLD (all p < 0.001).
Conclusion: AI-IR, a digital biomarker using nine accessible features, excels in predicting FL and T2DM, supporting early detection and personalized risk evaluation.
Speaker:
Chia-Lin Lee, Attending physician/MD.PhD
Division of Artificial Intelligence, Department of Digital Medicine, Taichung Veterans General Hospital, Taichung, Taiwan.
Authors:
Wei-Ju Liu, Research assistant/Master - Taichung Veterans General Hospital, Taichung, Taiwan; I-Hsin Huang, Research assistant/Master - Taichung Veterans General Hospital, Taichung, Taiwan.;
Chia-Lin
Lee,
Attending physician/MD.PhD - Division of Artificial Intelligence, Department of Digital Medicine, Taichung Veterans General Hospital, Taichung, Taiwan.
Navigating SDoH Z-Code Integration in Coding: Challenges, Innovations and Outcomes
Poster Number: P17
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Population Health, Healthcare Quality, Data Standards, Health Equity, Natural Language Processing, Nursing Informatics
Primary Track: Applications
Programmatic Theme: Public Health Informatics
Healthcare data capture now includes non-medical factors impacting health outcomes. Social risks significantly affect patient outcomes, making it essential for hospitals to address health disparities. A program was implemented to systematically capture SDoH data from ancillary documents and flowsheet data in collaboration with healthcare organizations, increasing SDoH Z code capture. This initiative substantially improves auto-suggested, accepted, and final coded SDoH data, emphasizing the importance of precise coding for better healthcare outcomes.
Speaker:
Tiffany Harman, MSN
Solventum
Author:
Rachael Howe, MS, MBA, BSN, RN, CCDS-O - Solventum HIS;
Poster Number: P17
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Population Health, Healthcare Quality, Data Standards, Health Equity, Natural Language Processing, Nursing Informatics
Primary Track: Applications
Programmatic Theme: Public Health Informatics
Healthcare data capture now includes non-medical factors impacting health outcomes. Social risks significantly affect patient outcomes, making it essential for hospitals to address health disparities. A program was implemented to systematically capture SDoH data from ancillary documents and flowsheet data in collaboration with healthcare organizations, increasing SDoH Z code capture. This initiative substantially improves auto-suggested, accepted, and final coded SDoH data, emphasizing the importance of precise coding for better healthcare outcomes.
Speaker:
Tiffany Harman, MSN
Solventum
Author:
Rachael Howe, MS, MBA, BSN, RN, CCDS-O - Solventum HIS;
Tiffany
Harman,
MSN - Solventum
Translating HbA1c Lab Test Name and Results from English to Arabic for Enhanced Patient Access and Workflow Efficiency
Poster Number: P18
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Personal Health Informatics, Controlled Terminologies, Ontologies, and Vocabularies, Delivering Health Information and Knowledge to the Public, Controlled Terminologies, Ontologies, and Vocabularies, Global Health, Healthcare Quality
Working Group: Consumer Health Informatics Working Group
Primary Track: Applications
Programmatic Theme: Consumer Health Informatics
In Jordan, where the healthcare system primarily operates in English, Arabic-speaking patients face challenges in understanding laboratory test results, including HbA1c. A study found that 27.1% of the population had limited health literacy, increasing reliance on healthcare staff for translation assistance. This project aligns with global healthcare accessibility efforts, such as the United States 21st Century Cures Act and the World Health Organization’s Framework on Integrated, People-Centered Health Services (IPCHS), emphasizing clear, patient-centered health communication.
This translational informatics project aimed to translate the HbA1c lab test name and results from English to Arabic to improve patient comprehension, digital engagement, and health literacy. Given the 30% prevalence of type 2 diabetes in Jordan, HbA1c was chosen for translation. The process utilized SNOMED CT, ICD-11, LOINC, and CPT, ensuring clinical accuracy, while Arabization, Arabicization, and descriptive translation methods provided cultural and linguistic adaptation.
Guided by the Sittig and Singh Eight-Dimensional Model for Health IT and the Plan-Do-Study-Act (PDSA) cycle, the study ensured technical and usability validation through a scientific committee of linguistic experts, clinicians, and health informaticians. The study targeted Arabic-speaking adults (18+) who had undergone HbA1c testing, integrating translated results into the patient portal.
Evaluation results demonstrated a 12% increase in patient portal engagement and an 84% reduction in lab staff translation assistance, improving workflow efficiency and patient autonomy. The initiative also contributed to establishing standardized Arabic medical terminology, supporting consistent health communication across institutions.
This project presents a scalable model for multilingual medical translation, enhancing patient engagement and workflow efficiency.
Speaker:
Aiman Alrawabdeh
The University of Texas Health Science Center at Houston (UTHealth) School of Biomedical Informatics
Authors:
Aiman Alrawabdeh - The University of Texas Health Science Center at Houston (UTHealth) School of Biomedical Informatics; Robert Murphy, MD - UTHealth School of Biomedical Informatics;
Poster Number: P18
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Personal Health Informatics, Controlled Terminologies, Ontologies, and Vocabularies, Delivering Health Information and Knowledge to the Public, Controlled Terminologies, Ontologies, and Vocabularies, Global Health, Healthcare Quality
Working Group: Consumer Health Informatics Working Group
Primary Track: Applications
Programmatic Theme: Consumer Health Informatics
In Jordan, where the healthcare system primarily operates in English, Arabic-speaking patients face challenges in understanding laboratory test results, including HbA1c. A study found that 27.1% of the population had limited health literacy, increasing reliance on healthcare staff for translation assistance. This project aligns with global healthcare accessibility efforts, such as the United States 21st Century Cures Act and the World Health Organization’s Framework on Integrated, People-Centered Health Services (IPCHS), emphasizing clear, patient-centered health communication.
This translational informatics project aimed to translate the HbA1c lab test name and results from English to Arabic to improve patient comprehension, digital engagement, and health literacy. Given the 30% prevalence of type 2 diabetes in Jordan, HbA1c was chosen for translation. The process utilized SNOMED CT, ICD-11, LOINC, and CPT, ensuring clinical accuracy, while Arabization, Arabicization, and descriptive translation methods provided cultural and linguistic adaptation.
Guided by the Sittig and Singh Eight-Dimensional Model for Health IT and the Plan-Do-Study-Act (PDSA) cycle, the study ensured technical and usability validation through a scientific committee of linguistic experts, clinicians, and health informaticians. The study targeted Arabic-speaking adults (18+) who had undergone HbA1c testing, integrating translated results into the patient portal.
Evaluation results demonstrated a 12% increase in patient portal engagement and an 84% reduction in lab staff translation assistance, improving workflow efficiency and patient autonomy. The initiative also contributed to establishing standardized Arabic medical terminology, supporting consistent health communication across institutions.
This project presents a scalable model for multilingual medical translation, enhancing patient engagement and workflow efficiency.
Speaker:
Aiman Alrawabdeh
The University of Texas Health Science Center at Houston (UTHealth) School of Biomedical Informatics
Authors:
Aiman Alrawabdeh - The University of Texas Health Science Center at Houston (UTHealth) School of Biomedical Informatics; Robert Murphy, MD - UTHealth School of Biomedical Informatics;
Aiman
Alrawabdeh - The University of Texas Health Science Center at Houston (UTHealth) School of Biomedical Informatics
Longitudinal Assessment of Model Performance and Bias in Glaucoma Predictions: Insights from Six Years of All of Us Data
Poster Number: P19
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Personal Health Informatics, Machine Learning, Health Equity, Diversity, Equity, Inclusion, and Accessibility, Racial disparities, Fairness and elimination of bias
Primary Track: Applications
Programmatic Theme: Clinical Informatics
Glaucoma is a leading cause of blindness, necessitating accurate risk prediction for early intervention. This study longitudinally evaluates machine learning models for glaucoma intervention prediction using six years of data from the All of Us Research Program. Models exhibited reduced performance on the latest dataset version with disparities across racial and gender subgroups. Results highlight the impact of class imbalance on generalizability and emphasize the need for diverse datasets to ensure equitable predictive accuracy.
Speaker:
Nick Souligne, M.S.
University of Arizona
Author:
Vignesh Subbian, PhD - University of Arizona;
Poster Number: P19
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Personal Health Informatics, Machine Learning, Health Equity, Diversity, Equity, Inclusion, and Accessibility, Racial disparities, Fairness and elimination of bias
Primary Track: Applications
Programmatic Theme: Clinical Informatics
Glaucoma is a leading cause of blindness, necessitating accurate risk prediction for early intervention. This study longitudinally evaluates machine learning models for glaucoma intervention prediction using six years of data from the All of Us Research Program. Models exhibited reduced performance on the latest dataset version with disparities across racial and gender subgroups. Results highlight the impact of class imbalance on generalizability and emphasize the need for diverse datasets to ensure equitable predictive accuracy.
Speaker:
Nick Souligne, M.S.
University of Arizona
Author:
Vignesh Subbian, PhD - University of Arizona;
Nick
Souligne,
M.S. - University of Arizona
Predictors of Preferences for Receiving and Sharing Cognitive Health Information Among ADRD Clinical Research Participants
Poster Number: P20
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Personal Health Informatics, Nursing Informatics, Health Equity
Primary Track: Applications
Programmatic Theme: Consumer Health Informatics
This preliminary study examined predictors of preferences for receiving and sharing cognitive health information among Alzheimer’s disease and related dementias (ADRD) clinical research participants. Older age was associated with decreased preference for digital formats (patient portals, tele-medicine) and in-person visits. Higher health literacy was associated with decreased preference for in-person visits. Lower healthcare access was associated with decreased preference for sharing their results with policymakers. Findings highlight the need for tailored health information delivery strategies.
Speaker:
Jeong Eun Kim, BSN
University of Pittsburgh School of Nursing
Authors:
Dianxu Ren, MD, PhD - University of Pittsburgh; Young Ji Lee, PhD, MSN , RN - University of Pittsburgh; Joshua Grill, PhD - University of California Irvine; Jennifer Lingler, PhD, MA, CRNP, FAAN - University of Pittsburgh;
Poster Number: P20
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Personal Health Informatics, Nursing Informatics, Health Equity
Primary Track: Applications
Programmatic Theme: Consumer Health Informatics
This preliminary study examined predictors of preferences for receiving and sharing cognitive health information among Alzheimer’s disease and related dementias (ADRD) clinical research participants. Older age was associated with decreased preference for digital formats (patient portals, tele-medicine) and in-person visits. Higher health literacy was associated with decreased preference for in-person visits. Lower healthcare access was associated with decreased preference for sharing their results with policymakers. Findings highlight the need for tailored health information delivery strategies.
Speaker:
Jeong Eun Kim, BSN
University of Pittsburgh School of Nursing
Authors:
Dianxu Ren, MD, PhD - University of Pittsburgh; Young Ji Lee, PhD, MSN , RN - University of Pittsburgh; Joshua Grill, PhD - University of California Irvine; Jennifer Lingler, PhD, MA, CRNP, FAAN - University of Pittsburgh;
Jeong Eun
Kim,
BSN - University of Pittsburgh School of Nursing
The Impact of Patient Portals on Medication Adherence in the Elderly: A Pilot Scoping Review
Poster Number: P21
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Personal Health Informatics, Patient Safety, Telemedicine, Real-World Evidence Generation, Healthcare Economics/Cost of Care, Chronic Care Management
Primary Track: Applications
Programmatic Theme: Consumer Health Informatics
This pilot scoping review addressed importance of medication adherence in elderly population since elderly are more prone to less adhere to their medications despite of having multiple disease conditions and polypharmacy. Following the PRISAM guideline and the PCC model, our literature search included 6 articles between 2017 and 2024, many of which focused on chronical conditions and had an experimental study design. These studies consistently reported a positive relationship between patient portal use and medical adherence. Portal features such as medication reminders, prescription refill functionalities, and note reading play a critical role in adherence.
Speaker:
suguna Kotte, PharmD
University of North Carolina at Chapel Hill
Authors:
Suguna Kotte, PharmD - University of North Carolina at Chapel Hill; Roxy Huang, MSIS - University of North Carolina at Chapel Hill; Danny Wu, PhD - University of North Carolina at Chapel Hill;
Poster Number: P21
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Personal Health Informatics, Patient Safety, Telemedicine, Real-World Evidence Generation, Healthcare Economics/Cost of Care, Chronic Care Management
Primary Track: Applications
Programmatic Theme: Consumer Health Informatics
This pilot scoping review addressed importance of medication adherence in elderly population since elderly are more prone to less adhere to their medications despite of having multiple disease conditions and polypharmacy. Following the PRISAM guideline and the PCC model, our literature search included 6 articles between 2017 and 2024, many of which focused on chronical conditions and had an experimental study design. These studies consistently reported a positive relationship between patient portal use and medical adherence. Portal features such as medication reminders, prescription refill functionalities, and note reading play a critical role in adherence.
Speaker:
suguna Kotte, PharmD
University of North Carolina at Chapel Hill
Authors:
Suguna Kotte, PharmD - University of North Carolina at Chapel Hill; Roxy Huang, MSIS - University of North Carolina at Chapel Hill; Danny Wu, PhD - University of North Carolina at Chapel Hill;
suguna
Kotte,
PharmD - University of North Carolina at Chapel Hill
Analysis of electronic health record data illuminates heterogeneity of pediatric allergic trajectories
Poster Number: P22
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Pediatrics, Bioinformatics, Phenomics and Phenome-wide Association Studies
Primary Track: Applications
Programmatic Theme: Clinical Research Informatics
We used longitudinal electronic health record data to develop a primary care-based birth cohort to study pediatric allergic disease trajectories. Focusing on atopic dermatitis, IgE-mediated food allergy, asthma, allergic rhinitis, and eosinophilic esophagitis, we identified heterogenous 2-to-5-step trajectories, including ones distinct from the classical allergic march sequence. Atopic dermatitis-to-asthma and atopic dermatitis-to-allergic rhinitis trajectories were enriched among Black and female children, respectively. Our study highlights a framework for nuanced study of health trajectories in children.
Speaker:
Stanislaw Gabryszewski, MD, PhD
Children's Hospital of Philadelphia & University of Pennsylvania
Authors:
Jesse Dudley, MS - Children's Hospital of Philadelphia; Robert Grundmeier, MD - Children's Hospital of Philadelphia; Jonathan Spergel, MD, PhD - Children's Hospital of Philadelphia; David Hill, MD, PhD - Children's Hospital of Philadelphia;
Poster Number: P22
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Pediatrics, Bioinformatics, Phenomics and Phenome-wide Association Studies
Primary Track: Applications
Programmatic Theme: Clinical Research Informatics
We used longitudinal electronic health record data to develop a primary care-based birth cohort to study pediatric allergic disease trajectories. Focusing on atopic dermatitis, IgE-mediated food allergy, asthma, allergic rhinitis, and eosinophilic esophagitis, we identified heterogenous 2-to-5-step trajectories, including ones distinct from the classical allergic march sequence. Atopic dermatitis-to-asthma and atopic dermatitis-to-allergic rhinitis trajectories were enriched among Black and female children, respectively. Our study highlights a framework for nuanced study of health trajectories in children.
Speaker:
Stanislaw Gabryszewski, MD, PhD
Children's Hospital of Philadelphia & University of Pennsylvania
Authors:
Jesse Dudley, MS - Children's Hospital of Philadelphia; Robert Grundmeier, MD - Children's Hospital of Philadelphia; Jonathan Spergel, MD, PhD - Children's Hospital of Philadelphia; David Hill, MD, PhD - Children's Hospital of Philadelphia;
Stanislaw
Gabryszewski,
MD, PhD - Children's Hospital of Philadelphia & University of Pennsylvania
Identifying Substance Use Information in Pediatric EHRs Using a Hybrid Rule-based and Deep Learning Model
Poster Number: P23
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Pediatrics, Machine Learning, Natural Language Processing, Information Extraction, Population Health, Public Health
Primary Track: Applications
Programmatic Theme: Clinical Informatics
This study examined the generalizability of a previously developed substance use screening detection system for a general pediatric dataset. Structured and unstructured EHR data were screened for 15 substance use categories with deep learning and rule-based modules. ASUDS achieved an average AUC, specificity and sensitivity of 0.66, 0.97 and 0.36 respectively, a significant reduction in performance compared to the original study. Further work is needed to develop better generalizability for a larger clinical setting.
Speaker:
Clare Treutel, MS
Cincinnati Children's Hospital Medical Center
Authors:
Clare Treutel, MS - Cincinnati Children's Hospital Medical Center; Judith Dexheimer, PhD - Cincinnati Children's Hospital; Eneida Mendonca, MD, PhD - Cincinnati Children's Hospital / University of Cincinnati; Katie Fox, BS - Cincinnati Children's Hospital Medical Center; Sarah Beal, PhD - Cincinnati Children's Hospital Medical Center;
Poster Number: P23
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Pediatrics, Machine Learning, Natural Language Processing, Information Extraction, Population Health, Public Health
Primary Track: Applications
Programmatic Theme: Clinical Informatics
This study examined the generalizability of a previously developed substance use screening detection system for a general pediatric dataset. Structured and unstructured EHR data were screened for 15 substance use categories with deep learning and rule-based modules. ASUDS achieved an average AUC, specificity and sensitivity of 0.66, 0.97 and 0.36 respectively, a significant reduction in performance compared to the original study. Further work is needed to develop better generalizability for a larger clinical setting.
Speaker:
Clare Treutel, MS
Cincinnati Children's Hospital Medical Center
Authors:
Clare Treutel, MS - Cincinnati Children's Hospital Medical Center; Judith Dexheimer, PhD - Cincinnati Children's Hospital; Eneida Mendonca, MD, PhD - Cincinnati Children's Hospital / University of Cincinnati; Katie Fox, BS - Cincinnati Children's Hospital Medical Center; Sarah Beal, PhD - Cincinnati Children's Hospital Medical Center;
Clare
Treutel,
MS - Cincinnati Children's Hospital Medical Center
Enhancing Medication Safety with System Approach to Verifying Electronic Prescriptions (SAV E-Rx): Pharmacists' Review of Product Selection Mismatches Between Prescribed and Dispensed Medications
Poster Number: P24
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Patient Safety, Clinical Decision Support, Healthcare Quality
Primary Track: Applications
Programmatic Theme: Clinical Informatics
Electronic prescriptions improve safety but introduce drug selection errors. SAV E-Rx detects mismatches in ingredient, strength, and dosage form between prescribed and dispensed medications using RxNorm and alerts pharmacists. A retrospective analysis (2023–2024) across 14 pharmacies found 662 flagged records, with 75 (11.3%) unintended mismatches, mostly from human factors. Pharmacists supported future alerts for 96% of unintended mismatches (p < 0.001). SAV E-Rx enhances medication safety through automated verification and clinical decision support.
Speaker:
Jun Gong, MPharm
University of Michigan
Authors:
Jun Gong, MPharm - University of Michigan; Vincent Marshall, MS - Univeristy of Michigan; Megan Whitaker, MHI - University of Michigan; Brigid Rowell, MA - University of Michigan; Corey Lester, PharmD, PhD - University of Michigan;
Poster Number: P24
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Patient Safety, Clinical Decision Support, Healthcare Quality
Primary Track: Applications
Programmatic Theme: Clinical Informatics
Electronic prescriptions improve safety but introduce drug selection errors. SAV E-Rx detects mismatches in ingredient, strength, and dosage form between prescribed and dispensed medications using RxNorm and alerts pharmacists. A retrospective analysis (2023–2024) across 14 pharmacies found 662 flagged records, with 75 (11.3%) unintended mismatches, mostly from human factors. Pharmacists supported future alerts for 96% of unintended mismatches (p < 0.001). SAV E-Rx enhances medication safety through automated verification and clinical decision support.
Speaker:
Jun Gong, MPharm
University of Michigan
Authors:
Jun Gong, MPharm - University of Michigan; Vincent Marshall, MS - Univeristy of Michigan; Megan Whitaker, MHI - University of Michigan; Brigid Rowell, MA - University of Michigan; Corey Lester, PharmD, PhD - University of Michigan;
Jun
Gong,
MPharm - University of Michigan
Enhancing Patient Safety through Improved Imaging Contrast-related Adverse Event Documentation
Poster Number: P25
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Patient Safety, Clinical Guidelines, Healthcare Quality, Controlled Terminologies, Ontologies, and Vocabularies
Primary Track: Applications
Programmatic Theme: Clinical Informatics
Gaps exist in documentation of contrast-related adverse events(AEs). This study aims to use case reports to create a comprehensive list of symptoms/signs categorized by physiologic system from the Common Terminology Criteria for Adverse Events(CTCAE) guide to design a patient reporting tool. Nervous system AEs were the most prevalent, however, not listed on the American College of Radiology reaction card. This study emphasizes the need for expansion of the ACR reaction card for cancer patients.
Speaker:
Niveditha Chandrakanth, Undergraduate Student
University of South Florida
Authors:
Niveditha Chandrakanth, N/A - University of South Florida; Christina Eldredge, MD, PhD, FAMIA - University of South Florida; Cesar Lam, MD - H. Lee Moffitt Cancer Center & Research Institute;
Poster Number: P25
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Patient Safety, Clinical Guidelines, Healthcare Quality, Controlled Terminologies, Ontologies, and Vocabularies
Primary Track: Applications
Programmatic Theme: Clinical Informatics
Gaps exist in documentation of contrast-related adverse events(AEs). This study aims to use case reports to create a comprehensive list of symptoms/signs categorized by physiologic system from the Common Terminology Criteria for Adverse Events(CTCAE) guide to design a patient reporting tool. Nervous system AEs were the most prevalent, however, not listed on the American College of Radiology reaction card. This study emphasizes the need for expansion of the ACR reaction card for cancer patients.
Speaker:
Niveditha Chandrakanth, Undergraduate Student
University of South Florida
Authors:
Niveditha Chandrakanth, N/A - University of South Florida; Christina Eldredge, MD, PhD, FAMIA - University of South Florida; Cesar Lam, MD - H. Lee Moffitt Cancer Center & Research Institute;
Niveditha
Chandrakanth,
Undergraduate Student - University of South Florida
Improving Algorithms for Detecting Inpatient Deterioration with a Remote Sensing Device to Document Real-Time Changes to Supplemental Oxygen
Poster Number: P26
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Patient Safety, Critical Care, Workflow
Primary Track: Applications
Programmatic Theme: Clinical Informatics
Increased reliance on supplemental oxygen is highly predictive of clinical deterioration, but less than 30% of patients have oxygen accurately documented in their record. We developed a small, low-cost, sensor that automatically detects and documents changes to supplemental oxygen. User testing of our prototype found that the sensor detected 100% of oxygen changes without false positive readings. Improved oxygen documentation is a promising opportunity to improve clinical documentation and patient safety without increasing clinical workload.
Speaker:
Bryan Steitz, PhD
Vanderbilt University Medical Center
Authors:
Adam Wright, PhD - Vanderbilt University Medical Center; Daniel Fabbri, Ph.D. - VUMC/Brim Analytics;
Poster Number: P26
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Patient Safety, Critical Care, Workflow
Primary Track: Applications
Programmatic Theme: Clinical Informatics
Increased reliance on supplemental oxygen is highly predictive of clinical deterioration, but less than 30% of patients have oxygen accurately documented in their record. We developed a small, low-cost, sensor that automatically detects and documents changes to supplemental oxygen. User testing of our prototype found that the sensor detected 100% of oxygen changes without false positive readings. Improved oxygen documentation is a promising opportunity to improve clinical documentation and patient safety without increasing clinical workload.
Speaker:
Bryan Steitz, PhD
Vanderbilt University Medical Center
Authors:
Adam Wright, PhD - Vanderbilt University Medical Center; Daniel Fabbri, Ph.D. - VUMC/Brim Analytics;
Bryan
Steitz,
PhD - Vanderbilt University Medical Center
Cognitive Load Theory as a Framework for EHR Downtime
Poster Number: P27
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Patient Safety, Workflow, Administrative Systems, Human-computer Interaction
Primary Track: Applications
Programmatic Theme: Clinical Informatics
Cognitive Load Theory (CLT) explains how cognitive resources are managed during complex tasks. In healthcare, electronic health records (EHRs) help reduce cognitive burden, but downtime increases extraneous load, disrupting workflow. Applying CLT principles can enhance downtime preparedness through structured workflows, targeted training, and optimized EHR design. Future research should explore how EHR reliance impacts cognitive load during downtime to improve workflow resilience, clinician adaptability, and patient safety.
Speaker:
Amber Massey, MSN
Vanderbilt Unviersity Medical Center/Vanderbilt University
Author:
Amber Massey, MSN - Vanderbilt University Medical Center;
Poster Number: P27
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Patient Safety, Workflow, Administrative Systems, Human-computer Interaction
Primary Track: Applications
Programmatic Theme: Clinical Informatics
Cognitive Load Theory (CLT) explains how cognitive resources are managed during complex tasks. In healthcare, electronic health records (EHRs) help reduce cognitive burden, but downtime increases extraneous load, disrupting workflow. Applying CLT principles can enhance downtime preparedness through structured workflows, targeted training, and optimized EHR design. Future research should explore how EHR reliance impacts cognitive load during downtime to improve workflow resilience, clinician adaptability, and patient safety.
Speaker:
Amber Massey, MSN
Vanderbilt Unviersity Medical Center/Vanderbilt University
Author:
Amber Massey, MSN - Vanderbilt University Medical Center;
Amber
Massey,
MSN - Vanderbilt Unviersity Medical Center/Vanderbilt University
Whose Data, Whose Decision? Participant-Centered Approaches to Data Sharing in Brain Health Research
Poster Number: P28
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Patient Engagement and Preferences, Data Sharing, Policy, Artificial Intelligence
Primary Track: Policy
Programmatic Theme: Clinical Research Informatics
This study explores public perspectives on data sharing in brain health research using a nationally representative survey of 603 U.S. adults. We assessed participants' willingness to participate in research under various data-sharing scenarios, including use of data for artificial intelligence and machine learning. Findings highlight the importance of transparency, data sensitivity, and recipient type, with strong support for returning data to participants and concerns around commercial and AI-related uses—informing more equitable, participant-centered data governance.
Speaker:
Stephanie Nino de Rivera, BA
Columbia University
Authors:
Stephanie Nino de Rivera, BA - Columbia University; Sarah Eslami, Bachelors of Science - Columbia University; Yihong Zhao, PhD, MPhil - Columbia University School of Nursing; Natalie Benda, PhD - Columbia University School of Nursing; Ruth Masterson Creber, PhD, MSc, RN - Columbia University;
Poster Number: P28
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Patient Engagement and Preferences, Data Sharing, Policy, Artificial Intelligence
Primary Track: Policy
Programmatic Theme: Clinical Research Informatics
This study explores public perspectives on data sharing in brain health research using a nationally representative survey of 603 U.S. adults. We assessed participants' willingness to participate in research under various data-sharing scenarios, including use of data for artificial intelligence and machine learning. Findings highlight the importance of transparency, data sensitivity, and recipient type, with strong support for returning data to participants and concerns around commercial and AI-related uses—informing more equitable, participant-centered data governance.
Speaker:
Stephanie Nino de Rivera, BA
Columbia University
Authors:
Stephanie Nino de Rivera, BA - Columbia University; Sarah Eslami, Bachelors of Science - Columbia University; Yihong Zhao, PhD, MPhil - Columbia University School of Nursing; Natalie Benda, PhD - Columbia University School of Nursing; Ruth Masterson Creber, PhD, MSc, RN - Columbia University;
Stephanie
Nino de Rivera,
BA - Columbia University
Reaching the Right Population: Study Relevance Outweighs Time Commitment and Compensation Concerns
Poster Number: P29
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Patient Engagement and Preferences, Evaluation, Diversity, Equity, Inclusion, and Accessibility, Delivering Health Information and Knowledge to the Public
Primary Track: Applications
Programmatic Theme: Clinical Research Informatics
Research recruitment at Emory Healthcare utilized Epic Clarity and MyChart to target eligible participants across 13 studies over 1.5 years. Study-specific SQL queries identified participants, followed by MyChart invitations. The impact of study characteristics on engagement was assessed, showing financial incentives had minimal effect, while relevance and study type influenced response and interest rates. Interventional studies targeting specific conditions were most effective, highlighting the importance of aligning recruitment strategies with participant needs and study goals.
Speaker:
Megan Schwinne, MPH
Emory University
Authors:
Megan Schwinne, MPH - Emory University; Barney Chan, MS CIS - Emory University; Chad Robichaux, MPH - Emory University; Mugisha Niyibizi, MPH - Emory University; Neal Dickert, MD, PhD - Emory University; Gabriel Najarro, PA-C - Emory Healthcare;
Poster Number: P29
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Patient Engagement and Preferences, Evaluation, Diversity, Equity, Inclusion, and Accessibility, Delivering Health Information and Knowledge to the Public
Primary Track: Applications
Programmatic Theme: Clinical Research Informatics
Research recruitment at Emory Healthcare utilized Epic Clarity and MyChart to target eligible participants across 13 studies over 1.5 years. Study-specific SQL queries identified participants, followed by MyChart invitations. The impact of study characteristics on engagement was assessed, showing financial incentives had minimal effect, while relevance and study type influenced response and interest rates. Interventional studies targeting specific conditions were most effective, highlighting the importance of aligning recruitment strategies with participant needs and study goals.
Speaker:
Megan Schwinne, MPH
Emory University
Authors:
Megan Schwinne, MPH - Emory University; Barney Chan, MS CIS - Emory University; Chad Robichaux, MPH - Emory University; Mugisha Niyibizi, MPH - Emory University; Neal Dickert, MD, PhD - Emory University; Gabriel Najarro, PA-C - Emory Healthcare;
Megan
Schwinne,
MPH - Emory University
Electronic Patient-Reported Outcome Measures for Total Joint Arthroplasty: AHRQ-Supported Research to Inform Real-World Implementation
Poster Number: P30
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Patient / Person Generated Health Data (Patient Reported Outcomes), Surgery, Patient Engagement and Preferences, Healthcare Quality, Legal, Ethical, Social and Regulatory Issues
Primary Track: Applications
Programmatic Theme: Clinical Informatics
This analysis of Agency for Healthcare Research and Quality-funded research supports implementation of electronic patient-reported outcome measures (ePROMs) to collect patient-reported outcomes (PROs) to support compliance with Centers for Medicare & Medicaid Services mandated hospital-level reporting of PROs for total hip arthroplasty and total knee arthroplasty starting in 2027. Results focus on ePROM collection system design, patient communication and education, and clinician PROs use.
Speaker:
Stephanie Pitts, PhD, CPH
AHRQ
Authors:
Stephanie Pitts, PhD, CPH - AHRQ; Christine Dymek, EdD - AHRQ;
Poster Number: P30
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Patient / Person Generated Health Data (Patient Reported Outcomes), Surgery, Patient Engagement and Preferences, Healthcare Quality, Legal, Ethical, Social and Regulatory Issues
Primary Track: Applications
Programmatic Theme: Clinical Informatics
This analysis of Agency for Healthcare Research and Quality-funded research supports implementation of electronic patient-reported outcome measures (ePROMs) to collect patient-reported outcomes (PROs) to support compliance with Centers for Medicare & Medicaid Services mandated hospital-level reporting of PROs for total hip arthroplasty and total knee arthroplasty starting in 2027. Results focus on ePROM collection system design, patient communication and education, and clinician PROs use.
Speaker:
Stephanie Pitts, PhD, CPH
AHRQ
Authors:
Stephanie Pitts, PhD, CPH - AHRQ; Christine Dymek, EdD - AHRQ;
Stephanie
Pitts,
PhD, CPH - AHRQ
Development and Modifications to the CONCERN Implementation Toolkit
Poster Number: P31
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Nursing Informatics, Clinical Decision Support, Informatics Implementation, Bioinformatics
Primary Track: Applications
Programmatic Theme: Clinical Research Informatics
The COmmunicating Narrative Concerns Entered by RNs (CONCERN) is a machine-learning based early warning system that leverages nursing documentation patterns to track risk of patient deterioration. A recent clinical trial that implemented this support tool at various sites showed a significant reduction in in-patient mortality and length of stay. To facilitate broader implementation, the CONCERN Implementation Toolkit (CIT) was developed and refined to guide hospitals in adopting the system, ensuring it fits site-specific contexts.
Speaker:
Temiloluwa Daramola, BA
Columbia University Irving Medical Center - Department of Biomedical Informatics
Authors:
Temiloluwa Daramola, BA - Columbia University Irving Medical Center - Department of Biomedical Informatics; Rachel Lee, PhD, RN - Columbia University; Sarah Rossetti, RN, PhD - Columbia University Department of Biomedical Informatics; PATRICIA C DYKES, PhD, MA, RN - Brigham and Women's Hospital/Harvard Medical School; Graham Lowenthal, BA - Brigham and Women's Hospital; Po-Yin Yen, PhD, RN - Washington University in St. Louis; Catherine Ivory, PhD, NI-BC, NEA-BC, FAAN - Vanderbilt University Medical Center; Brian Douthit, PhD, RN, NI-BC - Department of Biomedical Informatics, Vanderbilt University; Kenrick Cato, PhD, RN, CPHIMS, FAAN, FACMI - University of Pennsylvania/ Children's Hospital of Philadelphia;
Poster Number: P31
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Nursing Informatics, Clinical Decision Support, Informatics Implementation, Bioinformatics
Primary Track: Applications
Programmatic Theme: Clinical Research Informatics
The COmmunicating Narrative Concerns Entered by RNs (CONCERN) is a machine-learning based early warning system that leverages nursing documentation patterns to track risk of patient deterioration. A recent clinical trial that implemented this support tool at various sites showed a significant reduction in in-patient mortality and length of stay. To facilitate broader implementation, the CONCERN Implementation Toolkit (CIT) was developed and refined to guide hospitals in adopting the system, ensuring it fits site-specific contexts.
Speaker:
Temiloluwa Daramola, BA
Columbia University Irving Medical Center - Department of Biomedical Informatics
Authors:
Temiloluwa Daramola, BA - Columbia University Irving Medical Center - Department of Biomedical Informatics; Rachel Lee, PhD, RN - Columbia University; Sarah Rossetti, RN, PhD - Columbia University Department of Biomedical Informatics; PATRICIA C DYKES, PhD, MA, RN - Brigham and Women's Hospital/Harvard Medical School; Graham Lowenthal, BA - Brigham and Women's Hospital; Po-Yin Yen, PhD, RN - Washington University in St. Louis; Catherine Ivory, PhD, NI-BC, NEA-BC, FAAN - Vanderbilt University Medical Center; Brian Douthit, PhD, RN, NI-BC - Department of Biomedical Informatics, Vanderbilt University; Kenrick Cato, PhD, RN, CPHIMS, FAAN, FACMI - University of Pennsylvania/ Children's Hospital of Philadelphia;
Temiloluwa
Daramola,
BA - Columbia University Irving Medical Center - Department of Biomedical Informatics
Nurses’ Frustration with the Documentation of Care Plan
Poster Number: P32
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Nursing Informatics, Documentation Burden, Qualitative Methods, Surveys and Needs Analysis
Primary Track: Applications
Nursing care plans, though central to the profession, have become standardized templates, failing to capture individualized care. This leads to nurse frustration, reduced engagement, and a diminished perception of their value. Compliance-driven documentation has further distances nurses from the true intent of the nursing process. To restore meaning and relevance, it is essential to re-evaluate EHR-care plan generation and documentation, ensuring it reflects patient-centered care and highlights the genuine value of the care nurses provide.
Speaker:
Rosie Mugoya, Bsn
Goldfarb School of Nursing and Washington University of St. Louis
Authors:
Jennifer Thate, PhD, RN, CNE - Siena College; Hao Fan, MBBS - Washington University School of Medicine in St Louis; Sarah Rossetti, RN, PhD - Columbia University Department of Biomedical Informatics; Po-Yin Yen, PhD, RN - Washington University in St. Louis;
Poster Number: P32
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Nursing Informatics, Documentation Burden, Qualitative Methods, Surveys and Needs Analysis
Primary Track: Applications
Nursing care plans, though central to the profession, have become standardized templates, failing to capture individualized care. This leads to nurse frustration, reduced engagement, and a diminished perception of their value. Compliance-driven documentation has further distances nurses from the true intent of the nursing process. To restore meaning and relevance, it is essential to re-evaluate EHR-care plan generation and documentation, ensuring it reflects patient-centered care and highlights the genuine value of the care nurses provide.
Speaker:
Rosie Mugoya, Bsn
Goldfarb School of Nursing and Washington University of St. Louis
Authors:
Jennifer Thate, PhD, RN, CNE - Siena College; Hao Fan, MBBS - Washington University School of Medicine in St Louis; Sarah Rossetti, RN, PhD - Columbia University Department of Biomedical Informatics; Po-Yin Yen, PhD, RN - Washington University in St. Louis;
Rosie
Mugoya,
Bsn - Goldfarb School of Nursing and Washington University of St. Louis
Leveraging Clinical Language Models for Fall Risk Concept Extraction from Nursing Notes
Poster Number: P33
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Nursing Informatics, Information Extraction, Natural Language Processing, Patient Safety, Artificial Intelligence, Clinical Decision Support, Controlled Terminologies, Ontologies, and Vocabularies
Primary Track: Applications
Programmatic Theme: Clinical Informatics
This study evaluates clinical language models for extracting fall risk concepts from nursing notes, an underutilized yet rich data source. Analyzing 25,913 notes, we compared Bio+Clinical BERT and KLUE BERT (Korean BERT), achieving up to 89% accuracy. Results underscore the need for nursing-specific language models to enhance fall risk prediction and clinical decision support.
Speaker:
INSOOK CHO, PhD
Inha University
Authors:
Hyunchul Park, MBA - aSSIST University; Byeong Sun Park, MS - Inha University; Hyekeyong Shin, MS - Inha University;
Poster Number: P33
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Nursing Informatics, Information Extraction, Natural Language Processing, Patient Safety, Artificial Intelligence, Clinical Decision Support, Controlled Terminologies, Ontologies, and Vocabularies
Primary Track: Applications
Programmatic Theme: Clinical Informatics
This study evaluates clinical language models for extracting fall risk concepts from nursing notes, an underutilized yet rich data source. Analyzing 25,913 notes, we compared Bio+Clinical BERT and KLUE BERT (Korean BERT), achieving up to 89% accuracy. Results underscore the need for nursing-specific language models to enhance fall risk prediction and clinical decision support.
Speaker:
INSOOK CHO, PhD
Inha University
Authors:
Hyunchul Park, MBA - aSSIST University; Byeong Sun Park, MS - Inha University; Hyekeyong Shin, MS - Inha University;
INSOOK
CHO,
PhD - Inha University
Perspectives of Nursing Staff on the Role of Care Robots in Long-Term Care
Poster Number: P34
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Nursing Informatics, User-centered Design Methods, Human-computer Interaction, Workforce Development, Workflow
Primary Track: Applications
Programmatic Theme: Clinical Informatics
This survey study, using the socioecological model, explores nursing staff’s perspectives on care robots in long-term care facilities and identifies key predictors of these views on assistive and social roles. Results revealed neutral views and a preference for robots in assistive roles, like monitoring and reminders. Workload was a significant predictor of positive attitudes, suggesting robots could alleviate stress in high-burden environments. Findings emphasize robots should complement, not replace, human care, calling for further research.
Speaker:
Katie Trainum, BSN, RN
University of Texas at Austin
Authors:
Elizabeth Heitkemper, PhD, RN - The University of Texas at Austin; Elliott Hauser, PhD - The University of Texas at Austin; Karen Johnson, PhD, RN - The University of Texas at Austin; Bo Xie, PhD - University of Texas at Austin;
Poster Number: P34
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Nursing Informatics, User-centered Design Methods, Human-computer Interaction, Workforce Development, Workflow
Primary Track: Applications
Programmatic Theme: Clinical Informatics
This survey study, using the socioecological model, explores nursing staff’s perspectives on care robots in long-term care facilities and identifies key predictors of these views on assistive and social roles. Results revealed neutral views and a preference for robots in assistive roles, like monitoring and reminders. Workload was a significant predictor of positive attitudes, suggesting robots could alleviate stress in high-burden environments. Findings emphasize robots should complement, not replace, human care, calling for further research.
Speaker:
Katie Trainum, BSN, RN
University of Texas at Austin
Authors:
Elizabeth Heitkemper, PhD, RN - The University of Texas at Austin; Elliott Hauser, PhD - The University of Texas at Austin; Karen Johnson, PhD, RN - The University of Texas at Austin; Bo Xie, PhD - University of Texas at Austin;
Katie
Trainum,
BSN, RN - University of Texas at Austin
Multi-Task Learning for Extracting Menstrual Characteristics from Clinical Notes
Poster Number: P35
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Natural Language Processing, Information Extraction, Large Language Models (LLMs), Information Retrieval, Clinical Decision Support
Primary Track: Foundations
Programmatic Theme: Clinical Research Informatics
Menstrual health is a critical yet often overlooked aspect of women’s healthcare. Despite its clinical relevance, detailed data on menstrual characteristics is rarely available in structured medical records. To address this gap, we propose a novel Natural Language Processing pipeline to extract key menstrual cycle attributes - dysmenorrhea, regularity, flow volume, and intermenstrual bleeding. Our approach utilizes the GatorTron model with Multi-Task Prompt-based Learning, enhanced by a hybrid retrieval preprocessing step to identify relevant text segments. It outperforms baseline methods, achieving an average F1-score of 90% across all menstrual characteristics, despite being trained on fewer than 100 annotated clinical notes. The retrieval step consistently improves performance across all approaches, allowing the model to focus on the most relevant segments of lengthy clinical notes. These results show that combining multi-task learning with retrieval improves generalization and performance across menstrual characteristics, advancing automated extraction from clinical notes and supporting women's health research.
Speaker:
Anna Shopova, BS
Hasso Plattner Institute, Digital Engineering Faculty, University of Potsdam, Potsdam, Germany
Authors:
Eugenia Alleva, MD, MSc; Leslee Shaw, PhD - Blavatnik Family Women's Health Research Institute, Icahn School of Medicine at Mount Sinai, New York, USA; Christoph Lippert, PhD - Hasso Plattner Institute, Digital Engineering Faculty, University of Potsdam, Potsdam, Germany; Anna Shopova, BS - Hasso Plattner Institute, Digital Engineering Faculty, University of Potsdam, Potsdam, Germany;
Poster Number: P35
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Natural Language Processing, Information Extraction, Large Language Models (LLMs), Information Retrieval, Clinical Decision Support
Primary Track: Foundations
Programmatic Theme: Clinical Research Informatics
Menstrual health is a critical yet often overlooked aspect of women’s healthcare. Despite its clinical relevance, detailed data on menstrual characteristics is rarely available in structured medical records. To address this gap, we propose a novel Natural Language Processing pipeline to extract key menstrual cycle attributes - dysmenorrhea, regularity, flow volume, and intermenstrual bleeding. Our approach utilizes the GatorTron model with Multi-Task Prompt-based Learning, enhanced by a hybrid retrieval preprocessing step to identify relevant text segments. It outperforms baseline methods, achieving an average F1-score of 90% across all menstrual characteristics, despite being trained on fewer than 100 annotated clinical notes. The retrieval step consistently improves performance across all approaches, allowing the model to focus on the most relevant segments of lengthy clinical notes. These results show that combining multi-task learning with retrieval improves generalization and performance across menstrual characteristics, advancing automated extraction from clinical notes and supporting women's health research.
Speaker:
Anna Shopova, BS
Hasso Plattner Institute, Digital Engineering Faculty, University of Potsdam, Potsdam, Germany
Authors:
Eugenia Alleva, MD, MSc; Leslee Shaw, PhD - Blavatnik Family Women's Health Research Institute, Icahn School of Medicine at Mount Sinai, New York, USA; Christoph Lippert, PhD - Hasso Plattner Institute, Digital Engineering Faculty, University of Potsdam, Potsdam, Germany; Anna Shopova, BS - Hasso Plattner Institute, Digital Engineering Faculty, University of Potsdam, Potsdam, Germany;
Anna
Shopova,
BS - Hasso Plattner Institute, Digital Engineering Faculty, University of Potsdam, Potsdam, Germany
A Preliminary Study of LoRA Experts for Personalized Clinical Summarization
Poster Number: P36
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Natural Language Processing, Large Language Models (LLMs), Artificial Intelligence, User-centered Design Methods
Primary Track: Applications
Clinicians document and summarize critical patient information in Electronic Health Records (EHRs), but existing summarization models fail to address specialty-specific needs. Single-task models lack scalability, while multi-task models suffer from task interference. To improve personalized clinical summarization, we propose a novel approach leveraging the Llama3.1 model with Mixture of Low-Rank Adaptation (LoRA) Experts. This method enhances accuracy and flexibility, enabling tailored summaries for different specialties without compromising performance or manageability.
Speaker:
Mengxian Lyu, Master
University of Florida
Authors:
Mengxian Lyu, Master - University of Florida; Yonghui Wu, PhD - University of Florida;
Poster Number: P36
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Natural Language Processing, Large Language Models (LLMs), Artificial Intelligence, User-centered Design Methods
Primary Track: Applications
Clinicians document and summarize critical patient information in Electronic Health Records (EHRs), but existing summarization models fail to address specialty-specific needs. Single-task models lack scalability, while multi-task models suffer from task interference. To improve personalized clinical summarization, we propose a novel approach leveraging the Llama3.1 model with Mixture of Low-Rank Adaptation (LoRA) Experts. This method enhances accuracy and flexibility, enabling tailored summaries for different specialties without compromising performance or manageability.
Speaker:
Mengxian Lyu, Master
University of Florida
Authors:
Mengxian Lyu, Master - University of Florida; Yonghui Wu, PhD - University of Florida;
Mengxian
Lyu,
Master - University of Florida
Schema-Free LLM-Based Extraction of Clinical Measurements
Poster Number: P37
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Natural Language Processing, Large Language Models (LLMs), Information Extraction, Information Retrieval
Primary Track: Applications
Programmatic Theme: Clinical Informatics
We have developed a software tool using large language models (LLMs) for the extraction of structured data from
unstructured clinical text. The structured output is achieved via custom prompts without a supporting schema, which
improves accuracy and performance vs. schema-based methods. We present results for a value extraction task
involving measurement-entity resolution. Preparations are underway to test our system on real-world patient data at
Grady Hospital in Atlanta.
Speaker:
Richard Boyd, Ph.D.
Georgia Tech Research Institute
Authors:
Micaela Siraj; Jon Duke, MD - Georgia Tech Research Institute;
Poster Number: P37
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Natural Language Processing, Large Language Models (LLMs), Information Extraction, Information Retrieval
Primary Track: Applications
Programmatic Theme: Clinical Informatics
We have developed a software tool using large language models (LLMs) for the extraction of structured data from
unstructured clinical text. The structured output is achieved via custom prompts without a supporting schema, which
improves accuracy and performance vs. schema-based methods. We present results for a value extraction task
involving measurement-entity resolution. Preparations are underway to test our system on real-world patient data at
Grady Hospital in Atlanta.
Speaker:
Richard Boyd, Ph.D.
Georgia Tech Research Institute
Authors:
Micaela Siraj; Jon Duke, MD - Georgia Tech Research Institute;
Richard
Boyd,
Ph.D. - Georgia Tech Research Institute
Automatic Generation of Medical Mermaid Flowcharts Based on DeepSeek-r1 Pre-trained Model
Poster Number: P38
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Natural Language Processing, Large Language Models (LLMs), Deep Learning, Artificial Intelligence, Information Extraction, Information Retrieval, Information Visualization, Workflow
Primary Track: Applications
Programmatic Theme: Public Health Informatics
We propose an automated approach for generating Mermaid flowcharts from medical literature, utilizing the DeepSeek-r1 model enhanced by TF-IDF keyword selection and LoRA fine-tuning. Evaluation results (F1=0.86) demonstrate a 56% reduction in reading time compared to traditional methods. This structured workflow extraction significantly improves the efficiency of medical document processing and reading.
Speaker:
Xianghan Tan, MS in Health Informatics
Weill Cornell Medicine
Authors:
Xianghan Tan, MS in Health Informatics - Weill Cornell Medicine; Haoxin Chen, MS - Weill Cornell Medical College; Ziyu Liu, Master of Science - Health Care and Social Assistance; Xuan Gao, Master - Weil Cornell Medicine; Shiqin Tong, Master - Cornell; Yunshu Yang, Ph.D. - University of Minnesota; Weijian Qin, Master - Weill Cornell Medicine; Jose Florez-Arango, MD MS PhD - Weill Cornell Medicine;
Poster Number: P38
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Natural Language Processing, Large Language Models (LLMs), Deep Learning, Artificial Intelligence, Information Extraction, Information Retrieval, Information Visualization, Workflow
Primary Track: Applications
Programmatic Theme: Public Health Informatics
We propose an automated approach for generating Mermaid flowcharts from medical literature, utilizing the DeepSeek-r1 model enhanced by TF-IDF keyword selection and LoRA fine-tuning. Evaluation results (F1=0.86) demonstrate a 56% reduction in reading time compared to traditional methods. This structured workflow extraction significantly improves the efficiency of medical document processing and reading.
Speaker:
Xianghan Tan, MS in Health Informatics
Weill Cornell Medicine
Authors:
Xianghan Tan, MS in Health Informatics - Weill Cornell Medicine; Haoxin Chen, MS - Weill Cornell Medical College; Ziyu Liu, Master of Science - Health Care and Social Assistance; Xuan Gao, Master - Weil Cornell Medicine; Shiqin Tong, Master - Cornell; Yunshu Yang, Ph.D. - University of Minnesota; Weijian Qin, Master - Weill Cornell Medicine; Jose Florez-Arango, MD MS PhD - Weill Cornell Medicine;
Xianghan
Tan,
MS in Health Informatics - Weill Cornell Medicine
Annotation and Information Extraction of Social Determinants of Health from Social Worker Notes of Pediatric Transplantation
Poster Number: P39
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Natural Language Processing, Large Language Models (LLMs), Pediatrics, Information Extraction
Primary Track: Applications
Programmatic Theme: Clinical Informatics
Social determinants of health (SDoH) play a critical role in pediatric transplant outcomes, influencing every stage from eligibility to long-term graft survival. Factors such as socioeconomic status, healthcare access, education, and family support directly impact whether a child receives a transplant in a timely manner and can follow essential post-operative care. Understanding these social factors allows healthcare systems to create targeted interventions that improve outcomes and promote equity. In this research, we use de-identified social worker notes from pediatric transplant patients’ electronic health records (EHRs) at UF Health Shands Children’s Hospital to annotate SDoH data. We developed an annotation framework to train a model capable of identifying key SDoH factors.
Speaker:
Xiaoyu Wang, MS
Florida State University
Authors:
Xiaoyu Wang, MS - Florida State University; Luis Sanchez, BS - Florida State University; Harjith Pradeep, BS - Florida State University; Dipankar Gupta, MD; Michael Killian, PhD - Florida State University; Zhe He, PhD, FAMIA - Florida State University;
Poster Number: P39
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Natural Language Processing, Large Language Models (LLMs), Pediatrics, Information Extraction
Primary Track: Applications
Programmatic Theme: Clinical Informatics
Social determinants of health (SDoH) play a critical role in pediatric transplant outcomes, influencing every stage from eligibility to long-term graft survival. Factors such as socioeconomic status, healthcare access, education, and family support directly impact whether a child receives a transplant in a timely manner and can follow essential post-operative care. Understanding these social factors allows healthcare systems to create targeted interventions that improve outcomes and promote equity. In this research, we use de-identified social worker notes from pediatric transplant patients’ electronic health records (EHRs) at UF Health Shands Children’s Hospital to annotate SDoH data. We developed an annotation framework to train a model capable of identifying key SDoH factors.
Speaker:
Xiaoyu Wang, MS
Florida State University
Authors:
Xiaoyu Wang, MS - Florida State University; Luis Sanchez, BS - Florida State University; Harjith Pradeep, BS - Florida State University; Dipankar Gupta, MD; Michael Killian, PhD - Florida State University; Zhe He, PhD, FAMIA - Florida State University;
Xiaoyu
Wang,
MS - Florida State University
Developing and Validating Natural Language Processing Algorithms to Extract Bleeding Concepts and Assess Bleeding Outcomes from Unstructured Clinical Notes
Poster Number: P40
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Natural Language Processing, Large Language Models (LLMs), Machine Learning, Information Extraction, Precision Medicine
Primary Track: Foundations
Programmatic Theme: Clinical Research Informatics
This study trained machine learning, deep learning, and transformer-based models on more than 4,000 expertly annotated clinical note snippets with the goal of automating bleeding event annotation. We demonstrate that extracted keywords were specific to bleeding concepts and that natural language processing algorithms effectively identified both positive and negative bleeding cases. This method enables robust extraction of bleeding outcomes from unstructured text, which is essential for EHR-based risk stratification, clinical decision making, and research.
Speaker:
Cameron Thomas, PharmD
University of Florida
Authors:
Cameron Thomas, PharmD - University of Florida; Caitrin McDonough, PhD - University of Florida; Makayla Kapalczynski, BS - University of Florida; Ellen Keeley, MD, MS - University of Florida; Yan Gong, PhD - University of Florida; Larisa Cavallari, PharmD - University of Florida; Masoud Rouhizadeh, PhD, MSc, MA - University of Florida;
Poster Number: P40
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Natural Language Processing, Large Language Models (LLMs), Machine Learning, Information Extraction, Precision Medicine
Primary Track: Foundations
Programmatic Theme: Clinical Research Informatics
This study trained machine learning, deep learning, and transformer-based models on more than 4,000 expertly annotated clinical note snippets with the goal of automating bleeding event annotation. We demonstrate that extracted keywords were specific to bleeding concepts and that natural language processing algorithms effectively identified both positive and negative bleeding cases. This method enables robust extraction of bleeding outcomes from unstructured text, which is essential for EHR-based risk stratification, clinical decision making, and research.
Speaker:
Cameron Thomas, PharmD
University of Florida
Authors:
Cameron Thomas, PharmD - University of Florida; Caitrin McDonough, PhD - University of Florida; Makayla Kapalczynski, BS - University of Florida; Ellen Keeley, MD, MS - University of Florida; Yan Gong, PhD - University of Florida; Larisa Cavallari, PharmD - University of Florida; Masoud Rouhizadeh, PhD, MSc, MA - University of Florida;
Cameron
Thomas,
PharmD - University of Florida
Experiences Using Smart Health Devices for Hypertension Management Among Racial and Ethnic Minority Older Adults
Poster Number: P41
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Mobile Health, Chronic Care Management, Patient Engagement and Preferences, Health Equity, Telemedicine
Primary Track: Applications
Programmatic Theme: Consumer Health Informatics
Smart health devices (SHD) are increasingly used to manage chronic health conditions among older adults. Yet, less than 4% users are African American and Hispanic/Latino. To promote the use of SHD, we aim to understand racial and ethnic minority older adults’ experiences using SHD. While 95% expressed positive outcomes, 30% noted frustration with device limitations. Our results highlight the need for more inclusive, accessible, and supportive approaches to adopting health technology for minority older adults.
Speaker:
Jany Sun, BS
Rush Medical College
Authors:
Yangjun Dong, MA - School of Social Welfare, University at Albany, SUNY; Jimmie Boliboun - Rush University Medical Center; Katherine Koo; Valeria Vazquez-Trejo, BS - Rush Medical College; Michael Cui, MD - Rush University Medical Center; Victoria Rizzo, PhD, LCSW-R - School of Social Welfare, University at Albany, SUNY; Jeannine Rowe, PhD, MSW - Department of Social Work, University of Wisconsin-Whitewater;
Poster Number: P41
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Mobile Health, Chronic Care Management, Patient Engagement and Preferences, Health Equity, Telemedicine
Primary Track: Applications
Programmatic Theme: Consumer Health Informatics
Smart health devices (SHD) are increasingly used to manage chronic health conditions among older adults. Yet, less than 4% users are African American and Hispanic/Latino. To promote the use of SHD, we aim to understand racial and ethnic minority older adults’ experiences using SHD. While 95% expressed positive outcomes, 30% noted frustration with device limitations. Our results highlight the need for more inclusive, accessible, and supportive approaches to adopting health technology for minority older adults.
Speaker:
Jany Sun, BS
Rush Medical College
Authors:
Yangjun Dong, MA - School of Social Welfare, University at Albany, SUNY; Jimmie Boliboun - Rush University Medical Center; Katherine Koo; Valeria Vazquez-Trejo, BS - Rush Medical College; Michael Cui, MD - Rush University Medical Center; Victoria Rizzo, PhD, LCSW-R - School of Social Welfare, University at Albany, SUNY; Jeannine Rowe, PhD, MSW - Department of Social Work, University of Wisconsin-Whitewater;
Jany
Sun,
BS - Rush Medical College
Remote Patient Monitoring for Hypertension: Experiences of English and Spanish-speaking Patients
Poster Number: P42
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Mobile Health, Chronic Care Management, Qualitative Methods, Patient Engagement and Preferences
Primary Track: Applications
Programmatic Theme: Clinical Informatics
This study explores the use of remote patient monitoring (RPM) for hypertension in diverse and underserved populations. Through semi-structured interviews with 20 Spanish and English-speaking Family Health Center patients, the research identifies barriers and facilitators of RPM use. Thematic analysis found variations in adherence to RPM protocol, impact of collaborative care, barriers to effective RPM utilization, and facilitators of RPM success.
Speaker:
Lisa Groom, PhD, RN
NYU
Authors:
Moroni Fernandez Cajavilca, MSN, RN - NYU Meyers; Rishika Budhrani, NP - NYU Langone; Lily Russell, BA - Boston College; Luchy Gonzalez, BSN - NYU Meyers; Antoinette Schoenthaler, EdD - NYU Langone; Devin Mann, MD - NYU Grossman School of Medicine; Abraham Brody, FAAN, PhD, RN - NYU Meyers;
Poster Number: P42
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Mobile Health, Chronic Care Management, Qualitative Methods, Patient Engagement and Preferences
Primary Track: Applications
Programmatic Theme: Clinical Informatics
This study explores the use of remote patient monitoring (RPM) for hypertension in diverse and underserved populations. Through semi-structured interviews with 20 Spanish and English-speaking Family Health Center patients, the research identifies barriers and facilitators of RPM use. Thematic analysis found variations in adherence to RPM protocol, impact of collaborative care, barriers to effective RPM utilization, and facilitators of RPM success.
Speaker:
Lisa Groom, PhD, RN
NYU
Authors:
Moroni Fernandez Cajavilca, MSN, RN - NYU Meyers; Rishika Budhrani, NP - NYU Langone; Lily Russell, BA - Boston College; Luchy Gonzalez, BSN - NYU Meyers; Antoinette Schoenthaler, EdD - NYU Langone; Devin Mann, MD - NYU Grossman School of Medicine; Abraham Brody, FAAN, PhD, RN - NYU Meyers;
Lisa
Groom,
PhD, RN - NYU
Accessible and Scalable Closed-loop Neuromotor Rehabilitation Using Mobile Computer Vision and Transcutaneous Vagus Nerve Stimulation
Poster Number: P43
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Mobile Health, Disability, Accessibility, and Human Function, Telemedicine
Primary Track: Applications
Programmatic Theme: Clinical Research Informatics
Pairing transcutaneous vagus nerve stimulation (tVNS) with good quality movements is an emerging therapy following neurological injury or disease. To enhance at-home rehabilitation we developed a real-time closed-loop system, where mobile computer vision model detects successful weight transfer during tango backward step therapy to deliver tVNS stimulation. Our system was able to stimulate good weight transfer with 71% overall stimulation success rate.
Speaker:
Joshua Posen, B.S.
Georgia Institute of Technology
Authors:
Arya Mohan, B.S. - Georgia Institute of Technology; Nathaniel Green, M.S. - Independent Researcher; Milka Trajkova, PhD - Georgia Institute of Technology; Minoru Shinohara, PhD - Georgia Institute of Technology; Hyeokhyen Kwon, Ph.D. - Emory University;
Poster Number: P43
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Mobile Health, Disability, Accessibility, and Human Function, Telemedicine
Primary Track: Applications
Programmatic Theme: Clinical Research Informatics
Pairing transcutaneous vagus nerve stimulation (tVNS) with good quality movements is an emerging therapy following neurological injury or disease. To enhance at-home rehabilitation we developed a real-time closed-loop system, where mobile computer vision model detects successful weight transfer during tango backward step therapy to deliver tVNS stimulation. Our system was able to stimulate good weight transfer with 71% overall stimulation success rate.
Speaker:
Joshua Posen, B.S.
Georgia Institute of Technology
Authors:
Arya Mohan, B.S. - Georgia Institute of Technology; Nathaniel Green, M.S. - Independent Researcher; Milka Trajkova, PhD - Georgia Institute of Technology; Minoru Shinohara, PhD - Georgia Institute of Technology; Hyeokhyen Kwon, Ph.D. - Emory University;
Joshua
Posen,
B.S. - Georgia Institute of Technology
MyCap Mobile App Strategic Evolution to Support Scalability for Hybrid, Remote, and Decentralized Trials
Poster Number: P44
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Mobile Health, Patient / Person Generated Health Data (Patient Reported Outcomes), Change Management
Primary Track: Applications
Programmatic Theme: Clinical Research Informatics
The MyCap Mobile App, launched in 2018, is a customizable participant-facing mobile application for remote data collection. Rapid adoption necessitated changes for scalability. We implemented three enhancements to support long-term growth: integration into REDCap, a complete app rewrite using the Flutter framework, and incorporation of NIH’s Mobile Toolbox. Continued growth coinciding with these changes suggest MyCap will continue to grow as a platform to enable decentralized, remote, and hybrid trials.
Speaker:
Mike Enger, M.S.
RTI International
Authors:
Alex Cheng, PhD - Vanderbilt University Medical Center; Mike Enger, M.S. - RTI International; Emily Serdoz, MPA - Vanderbilt University Medical Center; Jessica Eidenmuller, N/A - Vanderbilt University Medical Center; Richard Gershon, PhD - Northwestern University; Elizabeth Dworak, PhD, MS, MA - Northwestern University; Aaron Kaat, PhD - Northwestern University; Paul Harris, PhD - Vanderbilt University;
Poster Number: P44
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Mobile Health, Patient / Person Generated Health Data (Patient Reported Outcomes), Change Management
Primary Track: Applications
Programmatic Theme: Clinical Research Informatics
The MyCap Mobile App, launched in 2018, is a customizable participant-facing mobile application for remote data collection. Rapid adoption necessitated changes for scalability. We implemented three enhancements to support long-term growth: integration into REDCap, a complete app rewrite using the Flutter framework, and incorporation of NIH’s Mobile Toolbox. Continued growth coinciding with these changes suggest MyCap will continue to grow as a platform to enable decentralized, remote, and hybrid trials.
Speaker:
Mike Enger, M.S.
RTI International
Authors:
Alex Cheng, PhD - Vanderbilt University Medical Center; Mike Enger, M.S. - RTI International; Emily Serdoz, MPA - Vanderbilt University Medical Center; Jessica Eidenmuller, N/A - Vanderbilt University Medical Center; Richard Gershon, PhD - Northwestern University; Elizabeth Dworak, PhD, MS, MA - Northwestern University; Aaron Kaat, PhD - Northwestern University; Paul Harris, PhD - Vanderbilt University;
Mike
Enger,
M.S. - RTI International
Psycho-social and behavioral trends in Type 2 Diabetes self-management amongst medically underserved communities during the COVID-19 pandemic
Poster Number: P45
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Mobile Health, Quantitative Methods, Health Equity
Primary Track: Applications
Programmatic Theme: Public Health Informatics
This study examined how the COVID-19 pandemic affected Type 2 Diabetes-Mellitus (T2DM) self-management among medically underserved patients in New York City. Most participants were Hispanic, female, and foreign-born, with high rates of food insecurity and poor glycemic control. Pandemic-related disruptions to diet and general health were linked to greater diabetes distress and poorer self-care. Depression lowered self-efficacy, highlighting the influence of psychosocial factors. Findings inform targeted interventions to support T2DM management in vulnerable populations.
Speaker:
Elizabeth Campbell, MS, MSPH, PhD
Johns Hopkins Bloomberg School of Public Health
Authors:
Haomiao Jia, PhD - Columbia University Irving Medical Center; Andrea Cassells, MPH - Clinical Directors Network; TJ Lin, MPH - Clinical Directors Network; Arlene Smaldone, PhD - Columbia University Irving Medical Center; Jonathan Tobin, PhD - Clinical Directors Network; Pooja Desai, BA, MA, MPhil - Columbia University Irving Medical Center; George Hripcsak, MD - Columbia University Irving Medical Center; Lena Mamykina, PhD - Columbia University;
Poster Number: P45
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Mobile Health, Quantitative Methods, Health Equity
Primary Track: Applications
Programmatic Theme: Public Health Informatics
This study examined how the COVID-19 pandemic affected Type 2 Diabetes-Mellitus (T2DM) self-management among medically underserved patients in New York City. Most participants were Hispanic, female, and foreign-born, with high rates of food insecurity and poor glycemic control. Pandemic-related disruptions to diet and general health were linked to greater diabetes distress and poorer self-care. Depression lowered self-efficacy, highlighting the influence of psychosocial factors. Findings inform targeted interventions to support T2DM management in vulnerable populations.
Speaker:
Elizabeth Campbell, MS, MSPH, PhD
Johns Hopkins Bloomberg School of Public Health
Authors:
Haomiao Jia, PhD - Columbia University Irving Medical Center; Andrea Cassells, MPH - Clinical Directors Network; TJ Lin, MPH - Clinical Directors Network; Arlene Smaldone, PhD - Columbia University Irving Medical Center; Jonathan Tobin, PhD - Clinical Directors Network; Pooja Desai, BA, MA, MPhil - Columbia University Irving Medical Center; George Hripcsak, MD - Columbia University Irving Medical Center; Lena Mamykina, PhD - Columbia University;
Elizabeth
Campbell,
MS, MSPH, PhD - Johns Hopkins Bloomberg School of Public Health
Developing A Novel Approach to Video-Based Fall Risk Assessment in Home Healthcare Using Multimodal Large Language Models: A Pilot Study
Poster Number: P46
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Machine Learning, Artificial Intelligence, Large Language Models (LLMs)
Primary Track: Applications
Programmatic Theme: Clinical Research Informatics
Falls are a major concern in home healthcare. This study evaluates LLaVA-NeXT-Video7B-hf, a compact Multimodal Large Language Model (MLLM), for fall risk assessment using in-home video data. Structured prompts were developed from twelve literature-based risk factors. The model achieved 85.7% accuracy on seven simple factors, 100% on two complex ones, and failed on three requiring clinical expertise. Findings highlight both the promise and limitations of MLLMs for scalable, prompt-driven fall prevention in resource-constrained settings.
Speaker:
Pallavi Gupta, PhD
Columbia University
Authors:
Pallavi Gupta, PhD - Columbia University; Zhihong Zhang, PhD - Columbia University; Meijia Song, BSN - University of Minnesota; Martin Michalowski, PhD, FAMIA, FIAHSI - University of Minnesota; Xiao Hu, PhD - Emory University; Max Topaz, PhD, RN, MA, FAAN, FIAHSI, FACMI - Columbia University School of Nursing;
Poster Number: P46
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Machine Learning, Artificial Intelligence, Large Language Models (LLMs)
Primary Track: Applications
Programmatic Theme: Clinical Research Informatics
Falls are a major concern in home healthcare. This study evaluates LLaVA-NeXT-Video7B-hf, a compact Multimodal Large Language Model (MLLM), for fall risk assessment using in-home video data. Structured prompts were developed from twelve literature-based risk factors. The model achieved 85.7% accuracy on seven simple factors, 100% on two complex ones, and failed on three requiring clinical expertise. Findings highlight both the promise and limitations of MLLMs for scalable, prompt-driven fall prevention in resource-constrained settings.
Speaker:
Pallavi Gupta, PhD
Columbia University
Authors:
Pallavi Gupta, PhD - Columbia University; Zhihong Zhang, PhD - Columbia University; Meijia Song, BSN - University of Minnesota; Martin Michalowski, PhD, FAMIA, FIAHSI - University of Minnesota; Xiao Hu, PhD - Emory University; Max Topaz, PhD, RN, MA, FAAN, FIAHSI, FACMI - Columbia University School of Nursing;
Pallavi
Gupta,
PhD - Columbia University
Deep Learning-Based Model to Predict Burkitt Lymphoma Recurrence
Poster Number: P47
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Machine Learning, Artificial Intelligence, Bioinformatics, Omics (genomics, metabolomics, proteomics, transcriptomics, etc.) and Integrative Analyses
Working Group: Genomics and Translational Bioinformatics Working Group
Primary Track: Applications
Programmatic Theme: Translational Bioinformatics
Burkitt Lymphoma (BL) is an aggressive B Cell Lymphoma. Although the disease pathology is well characterized, there is currently a lack of literature specifically on BL recurrence and its prediction. We developed a deep learning (DL) model to predict BL recurrence by combining clinical and gene expression data. Our approach achieved an AUC of 0.812, outperforming traditional machine learning (ML) models. These results showed the effectiveness of DL-based models for BL recurrence prediction.
Speaker:
Avery Maytin, High School
Brown University
Authors:
Jessica Patricoski, PhD Candidate, Computational Biology - Brown University Center for Computational Molecular Biology; Ece Uzun, PhD - Brown University Health/Brown University;
Poster Number: P47
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Machine Learning, Artificial Intelligence, Bioinformatics, Omics (genomics, metabolomics, proteomics, transcriptomics, etc.) and Integrative Analyses
Working Group: Genomics and Translational Bioinformatics Working Group
Primary Track: Applications
Programmatic Theme: Translational Bioinformatics
Burkitt Lymphoma (BL) is an aggressive B Cell Lymphoma. Although the disease pathology is well characterized, there is currently a lack of literature specifically on BL recurrence and its prediction. We developed a deep learning (DL) model to predict BL recurrence by combining clinical and gene expression data. Our approach achieved an AUC of 0.812, outperforming traditional machine learning (ML) models. These results showed the effectiveness of DL-based models for BL recurrence prediction.
Speaker:
Avery Maytin, High School
Brown University
Authors:
Jessica Patricoski, PhD Candidate, Computational Biology - Brown University Center for Computational Molecular Biology; Ece Uzun, PhD - Brown University Health/Brown University;
Avery
Maytin,
High School - Brown University
Unraveling Social and Behavioral Determinants of Late-Stage Lung Cancer Diagnosis among Blacks and Whites via Bayesian Network Analysis
Poster Number: P48
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Machine Learning, Causal Inference, Health Equity, Population Health, Cancer Prevention
Primary Track: Applications
Programmatic Theme: Public Health Informatics
Significant disparities in lung cancer outcomes, particularly late-stage diagnosis (LSD), persist between Black and White populations in the United States. This study develops a Bayesian Network model to assess the influence of county-level neighborhood factors on LSD rates among these groups. Findings will inform targeted interventions aimed at reducing LSD rates and bridging racial gaps in lung cancer diagnosis.
Speaker:
Piyawan Conahan, Ph.D.
Moffitt Cancer Center
Authors:
Yi Luo, PhD - H. Lee Moffitt Cancer Center & Research Institute; Lary Robinson, M.D. - Thoracic Oncology Program (Surgery), H. Lee Moffitt Cancer Center & Research Institute, FL, USA; Trung Le, Ph.D. - Department of Industrial and Management Systems Engineering, University of South Florida, USA; Matthew B. Schabath, Ph.D. - Department of Cancer Epidemiology, H. Lee Moffitt Cancer Center & Research Institute, FL, USA; Margaret B. Byrne, Ph.D. - Department of Health Outcomes and Behavior, H. Lee Moffitt Cancer Center & Research Institute, FL, USA; B Lee Green, PhD; Issam El Naqa, PhD;
Poster Number: P48
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Machine Learning, Causal Inference, Health Equity, Population Health, Cancer Prevention
Primary Track: Applications
Programmatic Theme: Public Health Informatics
Significant disparities in lung cancer outcomes, particularly late-stage diagnosis (LSD), persist between Black and White populations in the United States. This study develops a Bayesian Network model to assess the influence of county-level neighborhood factors on LSD rates among these groups. Findings will inform targeted interventions aimed at reducing LSD rates and bridging racial gaps in lung cancer diagnosis.
Speaker:
Piyawan Conahan, Ph.D.
Moffitt Cancer Center
Authors:
Yi Luo, PhD - H. Lee Moffitt Cancer Center & Research Institute; Lary Robinson, M.D. - Thoracic Oncology Program (Surgery), H. Lee Moffitt Cancer Center & Research Institute, FL, USA; Trung Le, Ph.D. - Department of Industrial and Management Systems Engineering, University of South Florida, USA; Matthew B. Schabath, Ph.D. - Department of Cancer Epidemiology, H. Lee Moffitt Cancer Center & Research Institute, FL, USA; Margaret B. Byrne, Ph.D. - Department of Health Outcomes and Behavior, H. Lee Moffitt Cancer Center & Research Institute, FL, USA; B Lee Green, PhD; Issam El Naqa, PhD;
Piyawan
Conahan,
Ph.D. - Moffitt Cancer Center
Evaluating Machine Learning Models for Prehospital Stroke Triage in Emergency Medical Services
Poster Number: P49
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Machine Learning, Clinical Decision Support, Patient Safety, Artificial Intelligence, Healthcare Quality
Primary Track: Applications
Programmatic Theme: Clinical Informatics
Timely stroke triage is critical, yet common stroke scales used by Emergency Medical Services (EMS) often lead to misdiagnoses and costly transfers. We developed machine learning models (XGBoost, Neural Network, Random Forest) for detection of stroke and severe stroke using EMS data. XGBoost achieved the best performance, with an AUC of 0.795 for stroke detection and 0.834 for severe stroke detection. These findings highlight AI’s potential to enhance EMS stroke triage and support real-time decision-making.
Speaker:
Michael Saban, MS
Loyola University Chicago
Authors:
Samie Tootooni, PhD - Loyola University Chicago; Paula de la Peña, PhD, RN - Loyola University Chicago; Daniel Heiferman, MD - Edward-Elmhurst Health; Mark Cichon, DO, FACEP, FACOEP - Loyola University Medical Center;
Poster Number: P49
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Machine Learning, Clinical Decision Support, Patient Safety, Artificial Intelligence, Healthcare Quality
Primary Track: Applications
Programmatic Theme: Clinical Informatics
Timely stroke triage is critical, yet common stroke scales used by Emergency Medical Services (EMS) often lead to misdiagnoses and costly transfers. We developed machine learning models (XGBoost, Neural Network, Random Forest) for detection of stroke and severe stroke using EMS data. XGBoost achieved the best performance, with an AUC of 0.795 for stroke detection and 0.834 for severe stroke detection. These findings highlight AI’s potential to enhance EMS stroke triage and support real-time decision-making.
Speaker:
Michael Saban, MS
Loyola University Chicago
Authors:
Samie Tootooni, PhD - Loyola University Chicago; Paula de la Peña, PhD, RN - Loyola University Chicago; Daniel Heiferman, MD - Edward-Elmhurst Health; Mark Cichon, DO, FACEP, FACOEP - Loyola University Medical Center;
Michael
Saban,
MS - Loyola University Chicago
MOSIC: Model-Agnostic Optimal Subgroup Identification with Multi-Constraint for Improved Reliability
Poster Number: P50
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Machine Learning, Clinical Decision Support, Precision Medicine
Primary Track: Foundations
Identifying subgroups that benefit from specific treatments using observational data is challenging due to practical constraints like subgroup size and confounder balance. Existing methods often overlook these key constraints or address them separately. We propose a unified optimization framework that simultaneously identifies subgroups and enforces key constraints, with theoretical guarantees of convergence to feasible and locally optimal solutions. Its effectiveness is demonstrated on both synthetic and real-world datasets.
Speaker:
Wenxin Chen, MBI
Cornell University
Authors:
Wenxin Chen, MBI - Cornell University; Weishen Pan, PhD - Weill Cornell Medicine; Kyra Gan, PhD - Cornell University; Fei Wang, PhD - Weill Cornell Medicine;
Poster Number: P50
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Machine Learning, Clinical Decision Support, Precision Medicine
Primary Track: Foundations
Identifying subgroups that benefit from specific treatments using observational data is challenging due to practical constraints like subgroup size and confounder balance. Existing methods often overlook these key constraints or address them separately. We propose a unified optimization framework that simultaneously identifies subgroups and enforces key constraints, with theoretical guarantees of convergence to feasible and locally optimal solutions. Its effectiveness is demonstrated on both synthetic and real-world datasets.
Speaker:
Wenxin Chen, MBI
Cornell University
Authors:
Wenxin Chen, MBI - Cornell University; Weishen Pan, PhD - Weill Cornell Medicine; Kyra Gan, PhD - Cornell University; Fei Wang, PhD - Weill Cornell Medicine;
Wenxin
Chen,
MBI - Cornell University
Machine Learning based Hybrid Feature Selection Approach for the Detection of Biomarkers in Usher Syndrome using mRNA Expression Data
Poster Number: P51
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Machine Learning, Computational Biology, Data Mining, Artificial Intelligence
Primary Track: Applications
Programmatic Theme: Clinical Informatics
Usher syndrome, a rare genetic disorder causing hearing and vision loss, remains difficult to diagnose. This study employs a machine learning hybrid sequential feature selection approach to identify key mRNA biomarkers from 42,000 features, reducing them to 58 critical markers. Using nested cross-validation and machine learning models, we achieve robust classification performance. Pathway analysis highlights associations with auditory and visual functions, underscoring the potential of mRNA biomarkers for improving Usher syndrome diagnosis and treatment
Speaker:
Rama krishna Thelagathoti
Boys town National Research Hospital
Authors:
Wesley Tom, PhD - Boys Town National Research Hospital; Dinesh S. Chandel, PhD - Boys Town National Research Hospital; Chao Jiang, PhD - Boys Town National Research Hospital; Gary Krzyzanowski, MS - Boys Town National Research Hospital; Appolinaire Olou, PhD - Boys Town National Research Hospital; M Rohan Fernando, PhD - Boys Town National Research Hospital;
Poster Number: P51
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Machine Learning, Computational Biology, Data Mining, Artificial Intelligence
Primary Track: Applications
Programmatic Theme: Clinical Informatics
Usher syndrome, a rare genetic disorder causing hearing and vision loss, remains difficult to diagnose. This study employs a machine learning hybrid sequential feature selection approach to identify key mRNA biomarkers from 42,000 features, reducing them to 58 critical markers. Using nested cross-validation and machine learning models, we achieve robust classification performance. Pathway analysis highlights associations with auditory and visual functions, underscoring the potential of mRNA biomarkers for improving Usher syndrome diagnosis and treatment
Speaker:
Rama krishna Thelagathoti
Boys town National Research Hospital
Authors:
Wesley Tom, PhD - Boys Town National Research Hospital; Dinesh S. Chandel, PhD - Boys Town National Research Hospital; Chao Jiang, PhD - Boys Town National Research Hospital; Gary Krzyzanowski, MS - Boys Town National Research Hospital; Appolinaire Olou, PhD - Boys Town National Research Hospital; M Rohan Fernando, PhD - Boys Town National Research Hospital;
Rama krishna
Thelagathoti - Boys town National Research Hospital
Beyond the Bedside: Leveraging ICD Data and Machine Learning for Enhanced Febrile Neutropenia Risk Stratification
Poster Number: P52
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Machine Learning, Data Mining, Clinical Decision Support
Primary Track: Applications
Programmatic Theme: Clinical Informatics
This study refines risk stratification in febrile neutropenia by comparing a simplified MASCC score with a refined MASCC score that incorporates ICD‐derived flags for hypotension, sepsis, respiratory failure, COPD, dehydration, and cancer. In parallel, machine learning (ML) models were built using basic demographics and then extended with ICD‐features. The refined MASCC score improved discrimination (AUC 0.775 vs. 0.651) while ML models with ICD features achieved an AUC of ~0.87, suggesting enhanced detection of high‐risk patients.
Speaker:
Benedict Amalraj, MD
Louisiana State University Shreveport
Authors:
Benedict Amalraj, MD - Louisiana State University Shreveport; Mariana Marrero Castillo, MD - Louisiana State University Shreveport;
Poster Number: P52
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Machine Learning, Data Mining, Clinical Decision Support
Primary Track: Applications
Programmatic Theme: Clinical Informatics
This study refines risk stratification in febrile neutropenia by comparing a simplified MASCC score with a refined MASCC score that incorporates ICD‐derived flags for hypotension, sepsis, respiratory failure, COPD, dehydration, and cancer. In parallel, machine learning (ML) models were built using basic demographics and then extended with ICD‐features. The refined MASCC score improved discrimination (AUC 0.775 vs. 0.651) while ML models with ICD features achieved an AUC of ~0.87, suggesting enhanced detection of high‐risk patients.
Speaker:
Benedict Amalraj, MD
Louisiana State University Shreveport
Authors:
Benedict Amalraj, MD - Louisiana State University Shreveport; Mariana Marrero Castillo, MD - Louisiana State University Shreveport;
Benedict
Amalraj,
MD - Louisiana State University Shreveport
PrEP Adherence Prediction among Key Populations in Thailand: Classic Machine Learning vs. Deep Learning
Poster Number: P53
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Machine Learning, Deep Learning, Health Equity, Public Health
Primary Track: Applications
Programmatic Theme: Public Health Informatics
This study compares machine learning models to predict lost to follow-up (LTFU) in PrEP users among MSM and transgender women in Thailand. Using data from 7,680 clients, CatBoost outperformed others with an AUC of 0.684. Key risk factors for LTFU include younger age and engaging in condomless sex. Insights highlight the need for targeted interventions to enhance adherence and reduce HIV transmission.
Speaker:
Chandra Harsha Rachabathuni, MSc
Umass Chan Medical School
Authors:
ChandraHarsha Rachabathuni, MSc - Umass Chan Medical School; Feifan Liu, PhD - University of Massachusetts Chan Medical School; Artit Wongsa, MS - Institute of HIV Research and Innovation; Rena Janamnuaysook, MBA - 2Institute of HIV Research and Innovation; Nittaya Phanuphak, PhD - Institute of HIV Research and Innovation; Bo Wang, PhD - Umass Chan Medical School;
Poster Number: P53
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Machine Learning, Deep Learning, Health Equity, Public Health
Primary Track: Applications
Programmatic Theme: Public Health Informatics
This study compares machine learning models to predict lost to follow-up (LTFU) in PrEP users among MSM and transgender women in Thailand. Using data from 7,680 clients, CatBoost outperformed others with an AUC of 0.684. Key risk factors for LTFU include younger age and engaging in condomless sex. Insights highlight the need for targeted interventions to enhance adherence and reduce HIV transmission.
Speaker:
Chandra Harsha Rachabathuni, MSc
Umass Chan Medical School
Authors:
ChandraHarsha Rachabathuni, MSc - Umass Chan Medical School; Feifan Liu, PhD - University of Massachusetts Chan Medical School; Artit Wongsa, MS - Institute of HIV Research and Innovation; Rena Janamnuaysook, MBA - 2Institute of HIV Research and Innovation; Nittaya Phanuphak, PhD - Institute of HIV Research and Innovation; Bo Wang, PhD - Umass Chan Medical School;
Chandra Harsha
Rachabathuni,
MSc - Umass Chan Medical School
Protecting Patient Privacy Through Controlled Text Generation
Poster Number: P54
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Large Language Models (LLMs), Artificial Intelligence, Machine Learning
Primary Track: Foundations
We propose a controlled text generation (CTG) method to mitigate memorization and enhance patient privacy in medical large language models, preventing inadvertent leakage of sensitive patient information. Our method leverages classifiers during inference to guide token generation without additional LLM training. Evaluations demonstrate significant reduction in sensitive data leakage, confirming CTG's effectiveness in enhancing privacy protection.
Speaker:
Yifan Yang, B.S.
NCBI, NLM/NIH
Authors:
Yifan Yang, B.S. - NCBI, NLM/NIH; Yuancheng Xu, MS - University of Maryland, College Park; Qiao Jin, M.D. - National Institutes of Health; Anran Li, PhD - Yale University; Qingyu Chen, PhD - Yale University; Furong Huang, PhD - University of Maryland, College Park; Zhiyong Lu, PhD - National Library of Medicine, NIH;
Poster Number: P54
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Large Language Models (LLMs), Artificial Intelligence, Machine Learning
Primary Track: Foundations
We propose a controlled text generation (CTG) method to mitigate memorization and enhance patient privacy in medical large language models, preventing inadvertent leakage of sensitive patient information. Our method leverages classifiers during inference to guide token generation without additional LLM training. Evaluations demonstrate significant reduction in sensitive data leakage, confirming CTG's effectiveness in enhancing privacy protection.
Speaker:
Yifan Yang, B.S.
NCBI, NLM/NIH
Authors:
Yifan Yang, B.S. - NCBI, NLM/NIH; Yuancheng Xu, MS - University of Maryland, College Park; Qiao Jin, M.D. - National Institutes of Health; Anran Li, PhD - Yale University; Qingyu Chen, PhD - Yale University; Furong Huang, PhD - University of Maryland, College Park; Zhiyong Lu, PhD - National Library of Medicine, NIH;
Yifan
Yang,
B.S. - NCBI, NLM/NIH
Improving Cell Type Annotation with Context-aware Large Reasoning Models
Poster Number: P55
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Large Language Models (LLMs), Artificial Intelligence, Natural Language Processing
Primary Track: Foundations
Programmatic Theme: Translational Bioinformatics
Cell type annotation is essential in single-cell analysis for understanding cellular heterogeneity in health and disease. Existing methods typically lack interpretability and generalize poorly to new biological contexts. To overcome these limitations, we introduce a context-aware large reasoning model that integrates gene-expression profiles with biological metadata, emulating expert annotation processes. Our method demonstrates improved accuracy, robustness, and interpretability, effectively generalizing across diverse tissues and diseases without additional context-specific training.
Speaker:
Yin Fang, Ph.D.
National Institutes of Health
Authors:
Yin Fang, Ph.D. - National Institutes of Health; Qiao Jin, M.D. - National Institutes of Health; Guangzhi Xiong, BA - University of Virginia; Aidong Zhang, Ph.D. - University of Virginia; Zhiyong Lu, PhD - National Library of Medicine, NIH;
Poster Number: P55
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Large Language Models (LLMs), Artificial Intelligence, Natural Language Processing
Primary Track: Foundations
Programmatic Theme: Translational Bioinformatics
Cell type annotation is essential in single-cell analysis for understanding cellular heterogeneity in health and disease. Existing methods typically lack interpretability and generalize poorly to new biological contexts. To overcome these limitations, we introduce a context-aware large reasoning model that integrates gene-expression profiles with biological metadata, emulating expert annotation processes. Our method demonstrates improved accuracy, robustness, and interpretability, effectively generalizing across diverse tissues and diseases without additional context-specific training.
Speaker:
Yin Fang, Ph.D.
National Institutes of Health
Authors:
Yin Fang, Ph.D. - National Institutes of Health; Qiao Jin, M.D. - National Institutes of Health; Guangzhi Xiong, BA - University of Virginia; Aidong Zhang, Ph.D. - University of Virginia; Zhiyong Lu, PhD - National Library of Medicine, NIH;
Yin
Fang,
Ph.D. - National Institutes of Health
Comparing Llama3 and DeepSeekR1 on Biomedical Text Classification Tasks
Poster Number: P56
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Large Language Models (LLMs), Artificial Intelligence, Bioinformatics, Natural Language Processing
Primary Track: Applications
Programmatic Theme: Academic Informatics / LIEAF
This study benchmarks Llama3-70B and DeepSeekR1-distill-Llama3-70B on six biomedical text classification tasks from social media and clinical notes. Results show that DeepSeekR1-distill-Llama3-70B improves precision in most tasks, while Llama3-70B performs better in others. No model consistently outperforms the other, highlighting trade-offs in precision and recall. Findings emphasize the importance of task-specific model selection for healthcare applications.
Speaker:
Yuting Guo, MS
Emory University
Authors:
Yuting Guo, MS - Emory University; Abeed Sarker, PhD - Emory University School of Medicine;
Poster Number: P56
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Large Language Models (LLMs), Artificial Intelligence, Bioinformatics, Natural Language Processing
Primary Track: Applications
Programmatic Theme: Academic Informatics / LIEAF
This study benchmarks Llama3-70B and DeepSeekR1-distill-Llama3-70B on six biomedical text classification tasks from social media and clinical notes. Results show that DeepSeekR1-distill-Llama3-70B improves precision in most tasks, while Llama3-70B performs better in others. No model consistently outperforms the other, highlighting trade-offs in precision and recall. Findings emphasize the importance of task-specific model selection for healthcare applications.
Speaker:
Yuting Guo, MS
Emory University
Authors:
Yuting Guo, MS - Emory University; Abeed Sarker, PhD - Emory University School of Medicine;
Yuting
Guo,
MS - Emory University
A Question-Based Approach for Eligibility Representation in a Patient-Trial Matching Chat Application
Poster Number: P57
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Large Language Models (LLMs), Artificial Intelligence, Informatics Implementation, Information Extraction, Natural Language Processing
Primary Track: Applications
Programmatic Theme: Clinical Informatics
TBD
Speaker:
Jorge Barrios-Ginart, Ph.D.
Moffitt Cancer Center
Authors:
Jorge Barrios-Ginart, Ph.D. - Moffitt Cancer Center & Research Institute; Carolyn Rich, LPN - Moffitt Cancer Center & Research Institute; Noemi Feliciano, Mrs - Moffitt Cancer Center & Research Institute; Rodrigo Carvajal, B.Sc. - Moffitt Cancer Center & Research Institute; Steven Eschrich, PhD - Moffitt Cancer Center;
Poster Number: P57
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Large Language Models (LLMs), Artificial Intelligence, Informatics Implementation, Information Extraction, Natural Language Processing
Primary Track: Applications
Programmatic Theme: Clinical Informatics
TBD
Speaker:
Jorge Barrios-Ginart, Ph.D.
Moffitt Cancer Center
Authors:
Jorge Barrios-Ginart, Ph.D. - Moffitt Cancer Center & Research Institute; Carolyn Rich, LPN - Moffitt Cancer Center & Research Institute; Noemi Feliciano, Mrs - Moffitt Cancer Center & Research Institute; Rodrigo Carvajal, B.Sc. - Moffitt Cancer Center & Research Institute; Steven Eschrich, PhD - Moffitt Cancer Center;
Jorge
Barrios-Ginart,
Ph.D. - Moffitt Cancer Center
TrialGPT: Matching Patients to Clinical Trials with Large Language Models
Poster Number: P58
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Large Language Models (LLMs), Artificial Intelligence, Natural Language Processing
Primary Track: Foundations
Clinical trial recruitment remains challenging due to complex eligibility criteria and fragmented patient data. We introduce TrialGPT, an AI-driven framework that leverages large language models to filter trials, interpret patient eligibility, and rank suitable studies. Evaluated on three available cohorts of synthetic patients, TrialGPT achieved 87.3% accuracy with criterion-level explanations, outperforming baseline methods by up to 51.6%. Our pilot user study demonstrated a 42.6% reduction in screening time, highlighting TrialGPT’s potential for streamlined patient recruitment.
Speaker:
Qiao Jin, M.D.
National Institutes of Health
Authors:
Qiao Jin, M.D. - National Institutes of Health; Zifeng Wang, M.S. - University of Illinois Urbana-Champaign; Charalampos Floudas, MD, DMSc, MS - NIH/NCI; Nicholas Wan, Bachelor of Engineering - National Institutes of Health; Joey Chan, M.S. - National Library of Medicine; Fangyuan Chen, MD - University of Pittsburgh; Changlin Gong, MD - Albert Einstein College of Medicine; Dara Bracken-Clarke, MD - National Cancer Institute; Elisabetta Xue, MD - National Cancer Institute; Yin Fang, Ph.D. - National Institutes of Health; Shubo Tian, PhD - National Institutes of Health; Yifan Yang, B.S. - NCBI, NLM/NIH; Jimeng Sun - University of Illinois at Urbana Champaign; Zhiyong Lu, PhD - National Library of Medicine, NIH;
Poster Number: P58
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Large Language Models (LLMs), Artificial Intelligence, Natural Language Processing
Primary Track: Foundations
Clinical trial recruitment remains challenging due to complex eligibility criteria and fragmented patient data. We introduce TrialGPT, an AI-driven framework that leverages large language models to filter trials, interpret patient eligibility, and rank suitable studies. Evaluated on three available cohorts of synthetic patients, TrialGPT achieved 87.3% accuracy with criterion-level explanations, outperforming baseline methods by up to 51.6%. Our pilot user study demonstrated a 42.6% reduction in screening time, highlighting TrialGPT’s potential for streamlined patient recruitment.
Speaker:
Qiao Jin, M.D.
National Institutes of Health
Authors:
Qiao Jin, M.D. - National Institutes of Health; Zifeng Wang, M.S. - University of Illinois Urbana-Champaign; Charalampos Floudas, MD, DMSc, MS - NIH/NCI; Nicholas Wan, Bachelor of Engineering - National Institutes of Health; Joey Chan, M.S. - National Library of Medicine; Fangyuan Chen, MD - University of Pittsburgh; Changlin Gong, MD - Albert Einstein College of Medicine; Dara Bracken-Clarke, MD - National Cancer Institute; Elisabetta Xue, MD - National Cancer Institute; Yin Fang, Ph.D. - National Institutes of Health; Shubo Tian, PhD - National Institutes of Health; Yifan Yang, B.S. - NCBI, NLM/NIH; Jimeng Sun - University of Illinois at Urbana Champaign; Zhiyong Lu, PhD - National Library of Medicine, NIH;
Qiao
Jin,
M.D. - National Institutes of Health
Leveraging Large Language Models for Depression Detection in Palliative Care Patient Messages
Poster Number: P59
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Large Language Models (LLMs), Chronic Care Management, Bioinformatics
Primary Track: Applications
Programmatic Theme: Clinical Research Informatics
This study explores using patient-initiated portal messages for passive depression detection via large language models (LLMs). We applied LLaMA 3.1 with chain-of-thought prompting, alongside transformer-based models, to train a neural classifier. Expert-annotated validation and 5-fold cross-validation showed that few-shot learning improved accuracy. Flagged messages aligned with increased mental health referrals, highlighting the feasibility of LLM-based passive monitoring for integrating mental health assessment into clinical workflows.
Speaker:
Sheida Habibi, Masters
Emory University
Authors:
Selen Bozkurt Watson, PhD, MS - Emory University; Ravi Pathak, MD - Emory School of Medicine; Dio Kavalieratos, Phd - Emory School of Medicine; Carina Oltmann, MSSW, LCSW - Emory School of Medicine;
Poster Number: P59
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Large Language Models (LLMs), Chronic Care Management, Bioinformatics
Primary Track: Applications
Programmatic Theme: Clinical Research Informatics
This study explores using patient-initiated portal messages for passive depression detection via large language models (LLMs). We applied LLaMA 3.1 with chain-of-thought prompting, alongside transformer-based models, to train a neural classifier. Expert-annotated validation and 5-fold cross-validation showed that few-shot learning improved accuracy. Flagged messages aligned with increased mental health referrals, highlighting the feasibility of LLM-based passive monitoring for integrating mental health assessment into clinical workflows.
Speaker:
Sheida Habibi, Masters
Emory University
Authors:
Selen Bozkurt Watson, PhD, MS - Emory University; Ravi Pathak, MD - Emory School of Medicine; Dio Kavalieratos, Phd - Emory School of Medicine; Carina Oltmann, MSSW, LCSW - Emory School of Medicine;
Sheida
Habibi,
Masters - Emory University
Large Language Models Identify Errors in Clinical Value Sets
Poster Number: P60
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Large Language Models (LLMs), Clinical Decision Support, Interoperability and Health Information Exchange, Workflow
Primary Track: Applications
Programmatic Theme: Clinical Informatics
Clinical Value sets are essential for clinical decision support, quality measurement, and interoperability but are challenging to maintain due to evolving knowledge and variations in development. Errors in value sets may cause CDS malfunctions. This study tested ChatGPT o3-mini as a tool for auditing value sets, identifying potential issues in data extracted from VUMC and VSAC. Expert review validated 69% of the LLM-detected issues as actionable and over 50% of value sets containing validated errors.
Speaker:
Adam Wright, PhD
Vanderbilt University Medical Center
Authors:
Adam Wright, PhD - Vanderbilt University Medical Center; Laura Zahn, MS; Elise Russo - Vanderbilt University Medical Center; Dean Sittig, PhD - University of Texas Health Science Center at Houston;
Poster Number: P60
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Large Language Models (LLMs), Clinical Decision Support, Interoperability and Health Information Exchange, Workflow
Primary Track: Applications
Programmatic Theme: Clinical Informatics
Clinical Value sets are essential for clinical decision support, quality measurement, and interoperability but are challenging to maintain due to evolving knowledge and variations in development. Errors in value sets may cause CDS malfunctions. This study tested ChatGPT o3-mini as a tool for auditing value sets, identifying potential issues in data extracted from VUMC and VSAC. Expert review validated 69% of the LLM-detected issues as actionable and over 50% of value sets containing validated errors.
Speaker:
Adam Wright, PhD
Vanderbilt University Medical Center
Authors:
Adam Wright, PhD - Vanderbilt University Medical Center; Laura Zahn, MS; Elise Russo - Vanderbilt University Medical Center; Dean Sittig, PhD - University of Texas Health Science Center at Houston;
Adam
Wright,
PhD - Vanderbilt University Medical Center
Extracting Social Determinants of Health from Clinical Notes Using LLMs
Poster Number: P61
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Large Language Models (LLMs), Clinical Decision Support, Personal Health Informatics, Artificial Intelligence, Natural Language Processing, Fairness and elimination of bias
Primary Track: Applications
We evaluated various Large Language Models (LLMs) for extracting Social Determinants of Health (SDoH) from clinical notes using multi-label classification. Fine-tuned GPT-4o performed best, with Mistral-7B leading among open-source models. Our note-level approach simplifies implementation, eliminates error propagation, and better captures implicit SDoH references compared to traditional extraction methods. Class imbalance remains challenging in the minorities, highlighting opportunities for further optimization.
Speaker:
Biao Yin, PhD
UMass Chan Medical School
Authors:
Ben Gerber, MD, MPH - University of Massachusetts Chan Medical School; Huanmei Wu, FAMIA, PhD - Temple University; Omar Martinez, JD, MPH - University of Central Florida (UCF); Teresa Schmidt, PhD - OCHIN; Feifan Liu, PhD - University of Massachusetts Chan Medical School;
Poster Number: P61
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Large Language Models (LLMs), Clinical Decision Support, Personal Health Informatics, Artificial Intelligence, Natural Language Processing, Fairness and elimination of bias
Primary Track: Applications
We evaluated various Large Language Models (LLMs) for extracting Social Determinants of Health (SDoH) from clinical notes using multi-label classification. Fine-tuned GPT-4o performed best, with Mistral-7B leading among open-source models. Our note-level approach simplifies implementation, eliminates error propagation, and better captures implicit SDoH references compared to traditional extraction methods. Class imbalance remains challenging in the minorities, highlighting opportunities for further optimization.
Speaker:
Biao Yin, PhD
UMass Chan Medical School
Authors:
Ben Gerber, MD, MPH - University of Massachusetts Chan Medical School; Huanmei Wu, FAMIA, PhD - Temple University; Omar Martinez, JD, MPH - University of Central Florida (UCF); Teresa Schmidt, PhD - OCHIN; Feifan Liu, PhD - University of Massachusetts Chan Medical School;
Biao
Yin,
PhD - UMass Chan Medical School
Automated Risk Categorization of Metastatic Prostate Cancer Using Large Language Models
Poster Number: P62
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Large Language Models (LLMs), Clinical Decision Support, Information Extraction, Artificial Intelligence
Primary Track: Applications
Programmatic Theme: Clinical Informatics
We developed a large language model-based framework designed to directly enhance clinical practice by automating the prognostic classification of patients with metastatic hormone-sensitive prostate cancer. This approach accurately categorizes patients into clinically meaningful synchronous/metachronous and high/low-volume subgroups directly from electronic health records. Iterative clinician-driven error analyses and sophisticated prompt decomposition strategies were key in achieving clinically significant improvements in performance (weighted F1-score: 0.905), ultimately enabling faster, more precise decision-making in routine patient care.
Speaker:
Ji-Eun Yum, B.S.
Mayo Clinic Alix School of Medicine - Arizona
Authors:
Syed Arsalan Ahmed Naqvi, M.B.B.S - Mayo Clinic; Prateek Jain, MBBS - Mayo Clinic; Umair Ayub, PhD, MS - Mayo Clinic; Ben Zhou, PhD - Arizona State University; Huan He, Ph.D. - Yale University; Chitta Baral, PhD - Arizona State University; Neeraj Agarwal, MD, FASCO - University of Utah; Alan Bryce, MD - City of Hope; Cassandra Moore, MD - Mayo Clinic; Mark Waddle, MD - Mayo Clinic; Parminder Singh, MD - Mayo Clinic; Yousef Zakharia, MD - Mayo Clinic; Irbaz Riaz, MBBS, PhD, MS - Mayo Clinic;
Poster Number: P62
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Large Language Models (LLMs), Clinical Decision Support, Information Extraction, Artificial Intelligence
Primary Track: Applications
Programmatic Theme: Clinical Informatics
We developed a large language model-based framework designed to directly enhance clinical practice by automating the prognostic classification of patients with metastatic hormone-sensitive prostate cancer. This approach accurately categorizes patients into clinically meaningful synchronous/metachronous and high/low-volume subgroups directly from electronic health records. Iterative clinician-driven error analyses and sophisticated prompt decomposition strategies were key in achieving clinically significant improvements in performance (weighted F1-score: 0.905), ultimately enabling faster, more precise decision-making in routine patient care.
Speaker:
Ji-Eun Yum, B.S.
Mayo Clinic Alix School of Medicine - Arizona
Authors:
Syed Arsalan Ahmed Naqvi, M.B.B.S - Mayo Clinic; Prateek Jain, MBBS - Mayo Clinic; Umair Ayub, PhD, MS - Mayo Clinic; Ben Zhou, PhD - Arizona State University; Huan He, Ph.D. - Yale University; Chitta Baral, PhD - Arizona State University; Neeraj Agarwal, MD, FASCO - University of Utah; Alan Bryce, MD - City of Hope; Cassandra Moore, MD - Mayo Clinic; Mark Waddle, MD - Mayo Clinic; Parminder Singh, MD - Mayo Clinic; Yousef Zakharia, MD - Mayo Clinic; Irbaz Riaz, MBBS, PhD, MS - Mayo Clinic;
Ji-Eun
Yum,
B.S. - Mayo Clinic Alix School of Medicine - Arizona
Gene Set Analysis with Large Language Models
Poster Number: P63
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Large Language Models (LLMs), Computational Biology, Systems Biology
Primary Track: Applications
Programmatic Theme: Academic Informatics / LIEAF
Gene set analysis (GSA) is essential in genomic research, yet traditional methods often lack transparency and produce contextually irrelevant results, making interpretation challenging. While large language models (LLMs) offer a promising solution for result interpretation, they frequently hallucinate, reducing reliability. To address this, we develop a self-verifying language agent that autonomously interacts with biological databases to enhance accuracy and interpretability. Benchmarking on diverse datasets demonstrates significant improvements over vanilla LLMs.
Speaker:
Zhizheng Wang, Ph.D
National Institutes of Health
Authors:
Zhizheng Wang, Ph.D - National Institutes of Health; Qiao Jin, M.D. - National Institutes of Health; Chih-Hsuan Wei - NCBI; Shubo Tian, Ph.D. - National Institutes of Health; Po-Ting Lai; Qingqing Zhu, PHD - National Institutes of Health; Chi-Ping Day, Ph.D. - National Institutes of Health; Christina Ross, Ph.D. - National Institutes of Health; Robert Leaman - NCBI/NLM/NIH; Yifan Yang, B.S. - NCBI, NLM/NIH; Zhiyong Lu, PhD - National Library of Medicine, NIH;
Poster Number: P63
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Large Language Models (LLMs), Computational Biology, Systems Biology
Primary Track: Applications
Programmatic Theme: Academic Informatics / LIEAF
Gene set analysis (GSA) is essential in genomic research, yet traditional methods often lack transparency and produce contextually irrelevant results, making interpretation challenging. While large language models (LLMs) offer a promising solution for result interpretation, they frequently hallucinate, reducing reliability. To address this, we develop a self-verifying language agent that autonomously interacts with biological databases to enhance accuracy and interpretability. Benchmarking on diverse datasets demonstrates significant improvements over vanilla LLMs.
Speaker:
Zhizheng Wang, Ph.D
National Institutes of Health
Authors:
Zhizheng Wang, Ph.D - National Institutes of Health; Qiao Jin, M.D. - National Institutes of Health; Chih-Hsuan Wei - NCBI; Shubo Tian, Ph.D. - National Institutes of Health; Po-Ting Lai; Qingqing Zhu, PHD - National Institutes of Health; Chi-Ping Day, Ph.D. - National Institutes of Health; Christina Ross, Ph.D. - National Institutes of Health; Robert Leaman - NCBI/NLM/NIH; Yifan Yang, B.S. - NCBI, NLM/NIH; Zhiyong Lu, PhD - National Library of Medicine, NIH;
Zhizheng
Wang,
Ph.D - National Institutes of Health
Toward Agentic Workflow to Automate LOINC Coding of Laboratory Tests
Poster Number: P64
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Large Language Models (LLMs), Data Standards, Interoperability and Health Information Exchange, Natural Language Processing
Primary Track: Applications
Programmatic Theme: Clinical Research Informatics
We propose a multi-agent framework using GPT-4o to automate LOINC coding of laboratory tests in clinical notes. It involves an "align" agent for context alignment, a “select” agent for choosing the most suitable candidate, a “judge” agent for final selection, and a “suggest” agent for proposing revisions (if no code is selected). The system achieved high precision (96.27%) (evaluated on 181 entities), refraining from assigning any code as opposed to predicting an incorrect code.
Speaker:
Surabhi Datta, PhD
IMO Health
Authors:
Surabhi Datta, PhD - IMO Health; Joseph Cook, MS - IMO Health; Vidhya Sivakumaran, PhD - IMO Health; Chuck Levecke, BS - IMO Health; Xiaoyan Wang, PhD - IMO Health; Jingqi Wang, PhD - IMO Health;
Poster Number: P64
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Large Language Models (LLMs), Data Standards, Interoperability and Health Information Exchange, Natural Language Processing
Primary Track: Applications
Programmatic Theme: Clinical Research Informatics
We propose a multi-agent framework using GPT-4o to automate LOINC coding of laboratory tests in clinical notes. It involves an "align" agent for context alignment, a “select” agent for choosing the most suitable candidate, a “judge” agent for final selection, and a “suggest” agent for proposing revisions (if no code is selected). The system achieved high precision (96.27%) (evaluated on 181 entities), refraining from assigning any code as opposed to predicting an incorrect code.
Speaker:
Surabhi Datta, PhD
IMO Health
Authors:
Surabhi Datta, PhD - IMO Health; Joseph Cook, MS - IMO Health; Vidhya Sivakumaran, PhD - IMO Health; Chuck Levecke, BS - IMO Health; Xiaoyan Wang, PhD - IMO Health; Jingqi Wang, PhD - IMO Health;
Surabhi
Datta,
PhD - IMO Health
BioPulse-QA: A Novel Biomedical Question-Answering Benchmark for Evaluating Factuality, Robustness, and Bias in Large Language Models
Poster Number: P65
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Large Language Models (LLMs), Evaluation, Natural Language Processing
Primary Track: Applications
Programmatic Theme: Clinical Informatics
We introduce BioPulse-QA, a semi-automated biomedical Question Answering (QA) benchmark with drug labels, clinical trials, and clinical guidelines, designed for continual updates to evaluate emerging large language models (LLMs) on unseen data. It includes 350 QA pairs and evaluates three LLMs on 156 human-validated samples. Gemini-2.0-Flash evaluated with BioPulse-QA achieved the highest F1-score (82.8%) outperforming other models. The benchmark supports both extractive and abstractive QA considering retrieval accuracy, robustness and bias evaluations in LLM outputs.
Speaker:
Kriti Bhattarai, PhD in Computer Science
Yale University
Authors:
Kriti Bhattarai, PhD in Computer Science - Yale University; Vipina K. Keloth, PhD - Yale University; Yang Ren, Ph.D. - Yale University; Hua Xu, Ph.D - Yale University;
Poster Number: P65
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Large Language Models (LLMs), Evaluation, Natural Language Processing
Primary Track: Applications
Programmatic Theme: Clinical Informatics
We introduce BioPulse-QA, a semi-automated biomedical Question Answering (QA) benchmark with drug labels, clinical trials, and clinical guidelines, designed for continual updates to evaluate emerging large language models (LLMs) on unseen data. It includes 350 QA pairs and evaluates three LLMs on 156 human-validated samples. Gemini-2.0-Flash evaluated with BioPulse-QA achieved the highest F1-score (82.8%) outperforming other models. The benchmark supports both extractive and abstractive QA considering retrieval accuracy, robustness and bias evaluations in LLM outputs.
Speaker:
Kriti Bhattarai, PhD in Computer Science
Yale University
Authors:
Kriti Bhattarai, PhD in Computer Science - Yale University; Vipina K. Keloth, PhD - Yale University; Yang Ren, Ph.D. - Yale University; Hua Xu, Ph.D - Yale University;
Kriti
Bhattarai,
PhD in Computer Science - Yale University
Evaluating the Diagnostic Performance of State-of-the-Art Large Language Models on Psychiatry Case Vignettes
Poster Number: P66
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Large Language Models (LLMs), Evaluation, Diagnostic Systems, Clinical Decision Support, Natural Language Processing
Primary Track: Applications
Programmatic Theme: Clinical Informatics
We systematically evaluated 8 state-of-the-art LLMs on psychiatric diagnosis using 155 clinician-curated vignettes. We also assessed 3 of the LLMs' diagnostic reasoning on a 30-case subset. Results showed strong accuracy (83.9% at top-5) and high-quality reasoning (3.91/5 at the highest), with a significant correlation between reasoning quality and diagnostic accuracy. Despite promising results, these models struggle with clinical nuance and require oversight; further and deeper evaluation of a wider suite of LLMs is needed.
Speaker:
Kevin Jin, BS
Yale University
Authors:
Kevin Jin, BS - Yale University; Margaret Garrett, MD - The University of Texas Health Science Center at Houston; Ashley Huang, MD - The University of Texas Health Science Center at Houston; Mario Montelongo, MD - The University of Texas Health Science Center at Houston; Caesa Nagpal, MD - The University of Texas Health Science Center at Houston; Jasperina Shei, MD - The University of Texas Health Science Center at Houston; Judah Weathers, MD, DPhil - Yale University; Brian Zaboski, PhD - Yale University; Juliana Zhang, MD - The University of Texas Health Science Center at Houston; Salih Selek, MD - The University of Texas Health Science Center at Houston; Sarah Yip, MSc, PhD - Yale University; Hua Xu, Ph.D - Yale University;
Poster Number: P66
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Large Language Models (LLMs), Evaluation, Diagnostic Systems, Clinical Decision Support, Natural Language Processing
Primary Track: Applications
Programmatic Theme: Clinical Informatics
We systematically evaluated 8 state-of-the-art LLMs on psychiatric diagnosis using 155 clinician-curated vignettes. We also assessed 3 of the LLMs' diagnostic reasoning on a 30-case subset. Results showed strong accuracy (83.9% at top-5) and high-quality reasoning (3.91/5 at the highest), with a significant correlation between reasoning quality and diagnostic accuracy. Despite promising results, these models struggle with clinical nuance and require oversight; further and deeper evaluation of a wider suite of LLMs is needed.
Speaker:
Kevin Jin, BS
Yale University
Authors:
Kevin Jin, BS - Yale University; Margaret Garrett, MD - The University of Texas Health Science Center at Houston; Ashley Huang, MD - The University of Texas Health Science Center at Houston; Mario Montelongo, MD - The University of Texas Health Science Center at Houston; Caesa Nagpal, MD - The University of Texas Health Science Center at Houston; Jasperina Shei, MD - The University of Texas Health Science Center at Houston; Judah Weathers, MD, DPhil - Yale University; Brian Zaboski, PhD - Yale University; Juliana Zhang, MD - The University of Texas Health Science Center at Houston; Salih Selek, MD - The University of Texas Health Science Center at Houston; Sarah Yip, MSc, PhD - Yale University; Hua Xu, Ph.D - Yale University;
Kevin
Jin,
BS - Yale University
Developing and Evaluating Large Language Model-Powered Spanish-Language Chatbots for Tuberculosis Treatment Support
Poster Number: P67
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Large Language Models (LLMs), Global Health, Infectious Diseases and Epidemiology, Artificial Intelligence, Natural Language Processing, Delivering Health Information and Knowledge to the Public
Primary Track: Applications
Programmatic Theme: Clinical Informatics
Large language model (LLM)-powered chatbots can assist healthcare providers by answering questions, yet no Spanish-language chatbots specifically designed for tuberculosis (TB) treatment support currently exist. We developed 6 chatbot models using different LLM training techniques and conducted internal and external expert evaluations to assess response accuracy, cultural appropriateness, and empathy. Our findings demonstrate the feasibility of LLM-powered chatbots for TB treatment support, with key areas identified for further refinement, including empathy and response accuracy.
Speaker:
Haroon Jakher, MD
Ochsner Healthcare
Authors:
Denise Galdamez, RN - University of Washington; Daniil Filienko, BS in Computer Science and Systems - University of Washington Tacoma; Javier Roberti, PhD - Centre for Research on Epidemiology and Public Health (CIESP); Mahek Nizar, Student - University of Washington Tacoma; Alfonso Aguilar, BS - University of Washington; Charles Kwanin, RN - University of Washington; Yvette Rodriguez, BS - University of Washington; Jennifer Sprecher, BS - University of Washington; Martine De Cock, PhD - University of Washington Tacoma; Weichao Yuwen, PhD, RN - University of Washington Tacoma; Sarah Iribarren, PhD - University of Washington;
Poster Number: P67
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Large Language Models (LLMs), Global Health, Infectious Diseases and Epidemiology, Artificial Intelligence, Natural Language Processing, Delivering Health Information and Knowledge to the Public
Primary Track: Applications
Programmatic Theme: Clinical Informatics
Large language model (LLM)-powered chatbots can assist healthcare providers by answering questions, yet no Spanish-language chatbots specifically designed for tuberculosis (TB) treatment support currently exist. We developed 6 chatbot models using different LLM training techniques and conducted internal and external expert evaluations to assess response accuracy, cultural appropriateness, and empathy. Our findings demonstrate the feasibility of LLM-powered chatbots for TB treatment support, with key areas identified for further refinement, including empathy and response accuracy.
Speaker:
Haroon Jakher, MD
Ochsner Healthcare
Authors:
Denise Galdamez, RN - University of Washington; Daniil Filienko, BS in Computer Science and Systems - University of Washington Tacoma; Javier Roberti, PhD - Centre for Research on Epidemiology and Public Health (CIESP); Mahek Nizar, Student - University of Washington Tacoma; Alfonso Aguilar, BS - University of Washington; Charles Kwanin, RN - University of Washington; Yvette Rodriguez, BS - University of Washington; Jennifer Sprecher, BS - University of Washington; Martine De Cock, PhD - University of Washington Tacoma; Weichao Yuwen, PhD, RN - University of Washington Tacoma; Sarah Iribarren, PhD - University of Washington;
Haroon
Jakher,
MD - Ochsner Healthcare
Human In the Loop Explainable AI Solutions (Hileas)- a Large Language Model-Based Tool to Assess the Systematic Review: A Proof-of-Concept
Poster Number: P68
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Large Language Models (LLMs), Human-computer Interaction, Information Retrieval, Information Extraction, Artificial Intelligence
Primary Track: Foundations
Programmatic Theme: Clinical Research Informatics
Large Language Models (LLMs) have advanced significantly, enhancing research efficiency and reducing manual workload in biomedical fields, including systematic reviews. However, reliable interactive solutions for systematic manuscript evaluation remain underdeveloped. We developed Hileas (Human In the Loop Explainable AI Solutions), a web application leveraging LLMs to assess manuscript quality systematically, integrating human oversight to ensure reliability.
Hileas operates in three stages: Systematic Review, One Paper Review, and Methodology Checklist. Users upload manuscripts or retrieve full-text articles via PubMed Central, applying predefined or custom evaluation criteria. The system employs ChatGPT-4o to generate responses, referencing specific manuscript sections for transparency. Human reviewers assess LLM outputs using a four-point ordinal scale to rate answer quality and identify inaccuracies. Hileas integrates BioC PMC API to ensure compliance with licensing restrictions. The system was developed in Python 3.13 and tested by informatics students, who rated its usability via the System Usability Scale (SUS).
Hileas’ interactive interface supports manuscript evaluation through LLM-based content extraction, direct citation linkage, and structured human feedback. Eight users rated the system with an average SUS score of 85.4 (SD = 13.6), indicating high usability.
Hileas presents a scalable, explainable AI-assisted approach for systematic reviews, with potential to enhance research efficiency and reproducibility. Future work will evaluate LLM accuracy, bias, fairness, and generalizability, align workflows with Cochrane standards, and integrate EBMonFHIR for computable evidence synthesis.
Speaker:
Cody Couperus, MD
University of Maryland Medical Center
Authors:
Chen Dun, MHS - Johns Hopkins University; Harold Lehmann, MD, PhD - Johns Hopkins University;
Poster Number: P68
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Large Language Models (LLMs), Human-computer Interaction, Information Retrieval, Information Extraction, Artificial Intelligence
Primary Track: Foundations
Programmatic Theme: Clinical Research Informatics
Large Language Models (LLMs) have advanced significantly, enhancing research efficiency and reducing manual workload in biomedical fields, including systematic reviews. However, reliable interactive solutions for systematic manuscript evaluation remain underdeveloped. We developed Hileas (Human In the Loop Explainable AI Solutions), a web application leveraging LLMs to assess manuscript quality systematically, integrating human oversight to ensure reliability.
Hileas operates in three stages: Systematic Review, One Paper Review, and Methodology Checklist. Users upload manuscripts or retrieve full-text articles via PubMed Central, applying predefined or custom evaluation criteria. The system employs ChatGPT-4o to generate responses, referencing specific manuscript sections for transparency. Human reviewers assess LLM outputs using a four-point ordinal scale to rate answer quality and identify inaccuracies. Hileas integrates BioC PMC API to ensure compliance with licensing restrictions. The system was developed in Python 3.13 and tested by informatics students, who rated its usability via the System Usability Scale (SUS).
Hileas’ interactive interface supports manuscript evaluation through LLM-based content extraction, direct citation linkage, and structured human feedback. Eight users rated the system with an average SUS score of 85.4 (SD = 13.6), indicating high usability.
Hileas presents a scalable, explainable AI-assisted approach for systematic reviews, with potential to enhance research efficiency and reproducibility. Future work will evaluate LLM accuracy, bias, fairness, and generalizability, align workflows with Cochrane standards, and integrate EBMonFHIR for computable evidence synthesis.
Speaker:
Cody Couperus, MD
University of Maryland Medical Center
Authors:
Chen Dun, MHS - Johns Hopkins University; Harold Lehmann, MD, PhD - Johns Hopkins University;
Cody
Couperus,
MD - University of Maryland Medical Center
Assessing Use of LLM Evaluators for Patient-Facing Conversational Agents
Poster Number: P69
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Large Language Models (LLMs), Human-computer Interaction, Artificial Intelligence, Patient Engagement and Preferences
Primary Track: Applications
Programmatic Theme: Clinical Informatics
Large language models (LLMs) can be used to evaluate and refine AI-generated content. To inform the iterative development of an LLM evaluator for a patient-facing conversational agent, we compared the evaluation processes and outcomes of clinical experts and LLM evaluators. We identified several potential gaps between the criteria of LLM evaluators and the clinical expectations of the conversational agent, proposing strategies for developing LLM evaluators for nuanced patient-facing use cases.
Speaker:
Angela Mastrianni, PhD
NYU Langone Health
Authors:
Katerina Andreadis, MS - NYU Grossman School of Medicine; Ji Chen - NYU Langone Health; Danissa Rodriguez Caraballo, PhD Computer science - NYU Grossman School of Medicine; Fiona McBride, MS - NYU Grossman School of Medicine; Aditya Jain, BA - New York University Grossman School of Medicine; Lisa Groom, PhD, RN - NYU; Devin Mann, MD - NYU Grossman School of Medicine;
Poster Number: P69
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Large Language Models (LLMs), Human-computer Interaction, Artificial Intelligence, Patient Engagement and Preferences
Primary Track: Applications
Programmatic Theme: Clinical Informatics
Large language models (LLMs) can be used to evaluate and refine AI-generated content. To inform the iterative development of an LLM evaluator for a patient-facing conversational agent, we compared the evaluation processes and outcomes of clinical experts and LLM evaluators. We identified several potential gaps between the criteria of LLM evaluators and the clinical expectations of the conversational agent, proposing strategies for developing LLM evaluators for nuanced patient-facing use cases.
Speaker:
Angela Mastrianni, PhD
NYU Langone Health
Authors:
Katerina Andreadis, MS - NYU Grossman School of Medicine; Ji Chen - NYU Langone Health; Danissa Rodriguez Caraballo, PhD Computer science - NYU Grossman School of Medicine; Fiona McBride, MS - NYU Grossman School of Medicine; Aditya Jain, BA - New York University Grossman School of Medicine; Lisa Groom, PhD, RN - NYU; Devin Mann, MD - NYU Grossman School of Medicine;
Angela
Mastrianni,
PhD - NYU Langone Health
Visualizing Multilayer Spatiotemporal Epidemiological Data with CoronaViz
Poster Number: P70
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Information Visualization, Geospatial (GIS) Data/Analysis, Global Health
Primary Track: Foundations
Programmatic Theme: Public Health Informatics
Though COVID-19 spurred many Geographic Information Systems for Visual Analytics, these struggled to encode interactions of variables like cases, deaths, and vaccinations across space and time. As a remedy, we propose an animated, multilayer encoding, and develop CoronaViz, a rich, open-source, browser-based platform for epidemiological exploration. Both task-based user studies and in-depth interviews with epidemiologists emphasize the value of understanding how multiple spatiotemporal variables interact, while demonstrating CoronaViz’s effectiveness for visualizing COVID-19 data and beyond.
Speaker:
Brian Ondov, PhD
Yale School of Medicine
Authors:
Niklas Elmqvist, PhD - Aarhus University; Hanan Samet, PhD - University of Maryland;
Poster Number: P70
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Information Visualization, Geospatial (GIS) Data/Analysis, Global Health
Primary Track: Foundations
Programmatic Theme: Public Health Informatics
Though COVID-19 spurred many Geographic Information Systems for Visual Analytics, these struggled to encode interactions of variables like cases, deaths, and vaccinations across space and time. As a remedy, we propose an animated, multilayer encoding, and develop CoronaViz, a rich, open-source, browser-based platform for epidemiological exploration. Both task-based user studies and in-depth interviews with epidemiologists emphasize the value of understanding how multiple spatiotemporal variables interact, while demonstrating CoronaViz’s effectiveness for visualizing COVID-19 data and beyond.
Speaker:
Brian Ondov, PhD
Yale School of Medicine
Authors:
Niklas Elmqvist, PhD - Aarhus University; Hanan Samet, PhD - University of Maryland;
Brian
Ondov,
PhD - Yale School of Medicine
Developing a Scalable, User-Informed Electronic Patient Reported Outcome (ePRO) Framework to Promote Supportive Care in Ambulatory Oncology
Poster Number: P71
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Informatics Implementation, Chronic Care Management, Clinical Decision Support, User-centered Design Methods, Qualitative Methods
Primary Track: Applications
Programmatic Theme: Clinical Informatics
This project aimed to develop a scalable ePRO framework using existing EHR functionality and validated tools to enhance supportive care in ambulatory oncology. Guided by user design principles, we conducted a developmental formative evaluation, including 46 qualitative interviews with target users (clinicians and patients). The resulting framework, which features automated assessments via patient portals linked to clinician-facing decision support with dynamic referrals, demonstrates a user-informed, sustainable solution without significant infrastructure investment.
Speaker:
Eden Brauer
UCLA
Authors:
Eden Brauer - UCLA; Stephanie Lazaro, BS - UCLA; Amy Chen, PharmD - UCLA; Patricia Ganz, MD - UCLA; Maie St. John, MD, PhD - UCLA; Eric Cheng, MD, MS - UCLA;
Poster Number: P71
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Informatics Implementation, Chronic Care Management, Clinical Decision Support, User-centered Design Methods, Qualitative Methods
Primary Track: Applications
Programmatic Theme: Clinical Informatics
This project aimed to develop a scalable ePRO framework using existing EHR functionality and validated tools to enhance supportive care in ambulatory oncology. Guided by user design principles, we conducted a developmental formative evaluation, including 46 qualitative interviews with target users (clinicians and patients). The resulting framework, which features automated assessments via patient portals linked to clinician-facing decision support with dynamic referrals, demonstrates a user-informed, sustainable solution without significant infrastructure investment.
Speaker:
Eden Brauer
UCLA
Authors:
Eden Brauer - UCLA; Stephanie Lazaro, BS - UCLA; Amy Chen, PharmD - UCLA; Patricia Ganz, MD - UCLA; Maie St. John, MD, PhD - UCLA; Eric Cheng, MD, MS - UCLA;
Eden
Brauer - UCLA
Toward an Evidence-Based Information Technology (IT) Maturity Model for Home Health Agencies: A Qualitative Exploration of IT Maturity Dimensions
Poster Number: P72
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Informatics Implementation, Healthcare Quality, Nursing Informatics, Evaluation
Primary Track: Applications
Home health agencies (HHAs) need to use their resources efficiently to adopt an IT infrastructure that effectively supports their goals toward better care, better health, and reduced costs. Therefore, this study's objective was to systematically develop an evidence-based model that characterizes and grades HHAs' IT processes and infrastructure maturity. The model provides a roadmap for individual HHAs to assess and improve their IT maturity while enabling benchmarking and trend analysis across HHAs.
Speaker:
Güneş Koru, PhD, FAMIA
University of Arkansas for Medical Sciences, Northwest Regional Campus
Authors:
Güneş Koru, PhD, FAMIA - University of Arkansas for Medical Sciences, Northwest Regional Campus; Ali Alsarhan, MS - University of Maryland, Baltimore County;
Poster Number: P72
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Informatics Implementation, Healthcare Quality, Nursing Informatics, Evaluation
Primary Track: Applications
Home health agencies (HHAs) need to use their resources efficiently to adopt an IT infrastructure that effectively supports their goals toward better care, better health, and reduced costs. Therefore, this study's objective was to systematically develop an evidence-based model that characterizes and grades HHAs' IT processes and infrastructure maturity. The model provides a roadmap for individual HHAs to assess and improve their IT maturity while enabling benchmarking and trend analysis across HHAs.
Speaker:
Güneş Koru, PhD, FAMIA
University of Arkansas for Medical Sciences, Northwest Regional Campus
Authors:
Güneş Koru, PhD, FAMIA - University of Arkansas for Medical Sciences, Northwest Regional Campus; Ali Alsarhan, MS - University of Maryland, Baltimore County;
Güneş
Koru,
PhD, FAMIA - University of Arkansas for Medical Sciences, Northwest Regional Campus
A Tool for Assessment and Mitigation of Automation Bias
Poster Number: P73
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Informatics Implementation, Human-computer Interaction, Workflow, Artificial Intelligence
Primary Track: Applications
Programmatic Theme: Clinical Informatics
Automation bias (AB) is the tendency to over-rely on automated systems, which poses risks in healthcare AI applications. Existing literature on AB is fragmented, lacking a comprehensive evaluation framework. We conducted a literature review and collaborated with experts to develop a flexible template of questions and considerations for assessing AB across AI projects. This template provides structured guidance for mitigating AB in AI tool development and deployment. Future work will assess its implementation impact.
Speaker:
Lu Zheng, Ph.D., M.S.
Mayo Clinic
Authors:
Kyle Eickman, Pharm.D. - Mayo Clinic; Lu Zheng, Ph.D., M.S., R.N. - Mayo Clinic; Joshua Ohde, PhD - Mayo Clinic;
Poster Number: P73
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Informatics Implementation, Human-computer Interaction, Workflow, Artificial Intelligence
Primary Track: Applications
Programmatic Theme: Clinical Informatics
Automation bias (AB) is the tendency to over-rely on automated systems, which poses risks in healthcare AI applications. Existing literature on AB is fragmented, lacking a comprehensive evaluation framework. We conducted a literature review and collaborated with experts to develop a flexible template of questions and considerations for assessing AB across AI projects. This template provides structured guidance for mitigating AB in AI tool development and deployment. Future work will assess its implementation impact.
Speaker:
Lu Zheng, Ph.D., M.S.
Mayo Clinic
Authors:
Kyle Eickman, Pharm.D. - Mayo Clinic; Lu Zheng, Ph.D., M.S., R.N. - Mayo Clinic; Joshua Ohde, PhD - Mayo Clinic;
Lu
Zheng,
Ph.D., M.S. - Mayo Clinic
An Informatics-enabled Process for Monitoring Provider Performance
Poster Number: P74
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Informatics Implementation, Information Extraction, Information Visualization
Primary Track: Applications
Programmatic Theme: Clinical Informatics
The Joint Commission’s Ongoing Professional Practice Evaluation (OPPE) attempts to establish an objective methodology for assessing a healthcare provider’s performance and competency. Geisinger historically used a third-party vendor to support our OPPE program. We developed a superior solution that provides better data insights, eases the burden of the review process, and enables better tracking of compliance. This work is part of a systematic effort to increase understanding of our inpatient provider performance and empowering clinical leaders to prioritize and act on high-value opportunities for improving care.
Speaker:
Eric Reich, MSHI
Geisinger
Authors:
David Vawdrey, PhD - Geisinger; Michelle Dempsey, CBPI - Geisinger; Shelly Marek, BSN, RN - Geisinger; Casey Cauthorn, MIE - Geisinger; Michael Ellison, MS - Geisinger; Jason Puckey, MHA - Geisinger;
Poster Number: P74
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Informatics Implementation, Information Extraction, Information Visualization
Primary Track: Applications
Programmatic Theme: Clinical Informatics
The Joint Commission’s Ongoing Professional Practice Evaluation (OPPE) attempts to establish an objective methodology for assessing a healthcare provider’s performance and competency. Geisinger historically used a third-party vendor to support our OPPE program. We developed a superior solution that provides better data insights, eases the burden of the review process, and enables better tracking of compliance. This work is part of a systematic effort to increase understanding of our inpatient provider performance and empowering clinical leaders to prioritize and act on high-value opportunities for improving care.
Speaker:
Eric Reich, MSHI
Geisinger
Authors:
David Vawdrey, PhD - Geisinger; Michelle Dempsey, CBPI - Geisinger; Shelly Marek, BSN, RN - Geisinger; Casey Cauthorn, MIE - Geisinger; Michael Ellison, MS - Geisinger; Jason Puckey, MHA - Geisinger;
Eric
Reich,
MSHI - Geisinger
Life Cycle of a Quality Dashboard: Development, Implementation and Sustainment Strategies
Poster Number: P75
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Informatics Implementation, Information Visualization, Healthcare Quality
Primary Track: Applications
Programmatic Theme: Clinical Research Informatics
We developed quality dashboards to support implementation of three preventive care evidence-based practices for women at 22 VA sites. The dashboards integrate data from multiple sources into a usable format to inform sites about the characteristics of their patient population, provide women-tailored performance measures, and facilitate tracking implementation efforts. Dashboard adaptations based on ongoing communication with implementing sites supported utilization and sustainment of these tools in real time.
Speaker:
Cody Knight, None
Veterans Health Administration
Authors:
Cody Knight, None - Veterans Health Administration; Catherine Chanfreau, PhD - Veterans Health Administration; Bevanne Bean-Mayberry, MD MHS - Veterans Health Administration; Erin Finley, PhD MPH - Veterans Health Administration; Kimberly Clair, PhD - Veterans Health Administration; Rebecca Oberman, MSW MPH - Veteran Health Administration; Rachel Lesser, MPH - Veteran Health Administration; Tannaz Moin, MD MBA MSHS - Veteran Health Administration; Alison Hamilton, PhD MPH - Veteran Health Administration; Melissa Farmer Coste, PhD - Veteran Health Administration;
Poster Number: P75
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Informatics Implementation, Information Visualization, Healthcare Quality
Primary Track: Applications
Programmatic Theme: Clinical Research Informatics
We developed quality dashboards to support implementation of three preventive care evidence-based practices for women at 22 VA sites. The dashboards integrate data from multiple sources into a usable format to inform sites about the characteristics of their patient population, provide women-tailored performance measures, and facilitate tracking implementation efforts. Dashboard adaptations based on ongoing communication with implementing sites supported utilization and sustainment of these tools in real time.
Speaker:
Cody Knight, None
Veterans Health Administration
Authors:
Cody Knight, None - Veterans Health Administration; Catherine Chanfreau, PhD - Veterans Health Administration; Bevanne Bean-Mayberry, MD MHS - Veterans Health Administration; Erin Finley, PhD MPH - Veterans Health Administration; Kimberly Clair, PhD - Veterans Health Administration; Rebecca Oberman, MSW MPH - Veteran Health Administration; Rachel Lesser, MPH - Veteran Health Administration; Tannaz Moin, MD MBA MSHS - Veteran Health Administration; Alison Hamilton, PhD MPH - Veteran Health Administration; Melissa Farmer Coste, PhD - Veteran Health Administration;
Cody
Knight,
None - Veterans Health Administration
Ensuring Standard of Care through Automated Orders: Use Case with Endocrinology Consultations for Hospitalized Patients with Insulin Pumps
Poster Number: P76
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Informatics Implementation, Patient Safety, Natural Language Processing, Artificial Intelligence
Primary Track: Applications
Programmatic Theme: Clinical Informatics
There are a number of systematic factors that hinder adherence to national guidelines on standard of care. The use of automated orders presents a new additional failsafe mechanism to ensure best practices are followed. We developed a system that utilizes natural language processing and data-monitoring strategies to ensure hospitalized patients with an insulin pump are seen by an inpatient diabetes team. This initiative aims to improve patient safety in alignment with national guidelines.
Speaker:
Marc Maldaver, MD
Vanderbilt University Medical Center
Authors:
Sarah Stern, MD - Vanderbilt University Medical Center; Dara Mize, MD, MS - Vanderbilt University Medical Center;
Poster Number: P76
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Informatics Implementation, Patient Safety, Natural Language Processing, Artificial Intelligence
Primary Track: Applications
Programmatic Theme: Clinical Informatics
There are a number of systematic factors that hinder adherence to national guidelines on standard of care. The use of automated orders presents a new additional failsafe mechanism to ensure best practices are followed. We developed a system that utilizes natural language processing and data-monitoring strategies to ensure hospitalized patients with an insulin pump are seen by an inpatient diabetes team. This initiative aims to improve patient safety in alignment with national guidelines.
Speaker:
Marc Maldaver, MD
Vanderbilt University Medical Center
Authors:
Sarah Stern, MD - Vanderbilt University Medical Center; Dara Mize, MD, MS - Vanderbilt University Medical Center;
Marc
Maldaver,
MD - Vanderbilt University Medical Center
Data Source Variability on Comorbidity Indices And Its Impact on 30-Day Readmissions Prediction Performance
Poster Number: P77
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Informatics Implementation, Transitions of Care, Population Health
Primary Track: Applications
Programmatic Theme: Clinical Informatics
This study evaluates the impact of data source variability on the performance of Elixhauser and Charlson Comorbidity Index scores in predicting 30-day hospital readmissions in heart failure patients. Utilizing data from the PCORI-funded INSIGHT Clinical Research Network and OMOP-based clinical data, the analysis examines differences in comorbidity assessments and their effects on model performance across 984 patients from the MIGHTY Heart study. Results indicate consistent model performance, underscoring the indices’ reliability in diverse clinical settings.
Speaker:
Jacky Choi, MPH
Weill Cornell Medicine
Authors:
Jacky Choi, MPH - Weill Cornell Medicine; Ruth Masterson Creber, PhD, MSc, RN - Columbia University; Leah Shafran Topaz - Weill Cornell Medicine; Melani Ellison; Yihong Zhao, PhD - Columbia University School of Nursing; Meghan Reading Turchioe, PhD, MPH, RN - Columbia University School of Nursing; Brock Daniels, MD, MPH - Weill Cornell Medicine;
Poster Number: P77
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Informatics Implementation, Transitions of Care, Population Health
Primary Track: Applications
Programmatic Theme: Clinical Informatics
This study evaluates the impact of data source variability on the performance of Elixhauser and Charlson Comorbidity Index scores in predicting 30-day hospital readmissions in heart failure patients. Utilizing data from the PCORI-funded INSIGHT Clinical Research Network and OMOP-based clinical data, the analysis examines differences in comorbidity assessments and their effects on model performance across 984 patients from the MIGHTY Heart study. Results indicate consistent model performance, underscoring the indices’ reliability in diverse clinical settings.
Speaker:
Jacky Choi, MPH
Weill Cornell Medicine
Authors:
Jacky Choi, MPH - Weill Cornell Medicine; Ruth Masterson Creber, PhD, MSc, RN - Columbia University; Leah Shafran Topaz - Weill Cornell Medicine; Melani Ellison; Yihong Zhao, PhD - Columbia University School of Nursing; Meghan Reading Turchioe, PhD, MPH, RN - Columbia University School of Nursing; Brock Daniels, MD, MPH - Weill Cornell Medicine;
Jacky
Choi,
MPH - Weill Cornell Medicine
The Emerging Role of an Informatics Navigator at a Pediatric Research Institute
Poster Number: P78
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Informatics Implementation, Workforce Development, Education and Training, User-centered Design Methods, Pediatrics, Workflow
Primary Track: Foundations
Programmatic Theme: Clinical Research Informatics
To gain maximum value from investments in complex systems, research organizations must support users of informatics services and applications. Barriers, including limited awareness and training, may inhibit adoption. The "Research Informatics Navigator" role, created in 2022, promotes resource utilization, assists with onboarding, provides consultative services, and facilitates training. In 2024, the navigator identified 51 pain points, 22 suggested ideas, and 21 areas for improvement. The role continues to evolve, addressing these issues and collaborating regionally.
Speaker:
Abigail Kietzman, MS, ACRP-CP
Children's Mercy Kansas City
Author:
Mark Hoffman, PhD - Children's Mercy Kansas City;
Poster Number: P78
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Informatics Implementation, Workforce Development, Education and Training, User-centered Design Methods, Pediatrics, Workflow
Primary Track: Foundations
Programmatic Theme: Clinical Research Informatics
To gain maximum value from investments in complex systems, research organizations must support users of informatics services and applications. Barriers, including limited awareness and training, may inhibit adoption. The "Research Informatics Navigator" role, created in 2022, promotes resource utilization, assists with onboarding, provides consultative services, and facilitates training. In 2024, the navigator identified 51 pain points, 22 suggested ideas, and 21 areas for improvement. The role continues to evolve, addressing these issues and collaborating regionally.
Speaker:
Abigail Kietzman, MS, ACRP-CP
Children's Mercy Kansas City
Author:
Mark Hoffman, PhD - Children's Mercy Kansas City;
Abigail
Kietzman,
MS, ACRP-CP - Children's Mercy Kansas City
Understanding Subjectivity in Clinician's Evaluations of LLM-Based Chatbot Responses for Menopause Care
Poster Number: P79
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Human-computer Interaction, Large Language Models (LLMs), Evaluation, Qualitative Methods, Quantitative Methods
Primary Track: Foundations
Programmatic Theme: Public Health Informatics
As large language models (LLMs) increasingly serve as patient-facing tools for health information-seeking, clinician-based evaluation of their responses is critical. However, clinician-based evaluation of LLMs in itself can be a subjective process. In this mixed-methods study, we assessed clinicians’ ratings of LLM-based chatbot responses to menopause-related questions. We use the S.C.O.R.E. framework (Safety, Consensus, Explainability) to gather clinician feedback. We examined inter-rater variability and evaluator reasoning through qualitative feedback and a post-assessment discussion with the clinicians.
Speaker:
Roshini Deva, MS
Emory University
Authors:
Roshini Deva, MS - Emory University; Nadi Nina Kaonga, MD, MS, MHS - Emory University; Agena Davenport-Nicholson, MD - Emory University; Elizabeth Britton Chahine, MD - Emory University; Selen Bozkurt Watson, PhD, MS - Emory University; Azra Ismail, PhD - Emory University;
Poster Number: P79
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Human-computer Interaction, Large Language Models (LLMs), Evaluation, Qualitative Methods, Quantitative Methods
Primary Track: Foundations
Programmatic Theme: Public Health Informatics
As large language models (LLMs) increasingly serve as patient-facing tools for health information-seeking, clinician-based evaluation of their responses is critical. However, clinician-based evaluation of LLMs in itself can be a subjective process. In this mixed-methods study, we assessed clinicians’ ratings of LLM-based chatbot responses to menopause-related questions. We use the S.C.O.R.E. framework (Safety, Consensus, Explainability) to gather clinician feedback. We examined inter-rater variability and evaluator reasoning through qualitative feedback and a post-assessment discussion with the clinicians.
Speaker:
Roshini Deva, MS
Emory University
Authors:
Roshini Deva, MS - Emory University; Nadi Nina Kaonga, MD, MS, MHS - Emory University; Agena Davenport-Nicholson, MD - Emory University; Elizabeth Britton Chahine, MD - Emory University; Selen Bozkurt Watson, PhD, MS - Emory University; Azra Ismail, PhD - Emory University;
Roshini
Deva,
MS - Emory University
A Systems Engineering Approach to Optimize Pediatric Medication Safety
Poster Number: P80
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Human-computer Interaction, Pediatrics, Clinical Decision Support
Primary Track: Applications
Programmatic Theme: Clinical Informatics
We developed and implemented test cases to assess EHR safeguards against weight-based medication dosing (WBD) in two different EHR systems. Our research identified critical EHR usability issues related to medication dosing calculation and display, and a mismatch between alert timing and clinician workflow that can contribute to WBD errors in pediatric populations.
Speaker:
Garrett Zabala, Master of Science
MedStar Health Research Insititute
Authors:
Garrett Foresman, BS - MedStar Health Research Insititute; Yuuki Unno, MSHS - MedStar Health Research Institute; Sonita Bennett - MedStar Health; Joseph Blumenthal - MedStar Health; Mallory Tidwell, BSN, RN, CCRP - Children's Healthcare of Atlanta; Naveen Muthu, MD - Children's Healthcare of Atlanta; Sadaf Kazi, PhD - National Center for Human Factors in Healthcare, MedStar Health Research Institute;
Poster Number: P80
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Human-computer Interaction, Pediatrics, Clinical Decision Support
Primary Track: Applications
Programmatic Theme: Clinical Informatics
We developed and implemented test cases to assess EHR safeguards against weight-based medication dosing (WBD) in two different EHR systems. Our research identified critical EHR usability issues related to medication dosing calculation and display, and a mismatch between alert timing and clinician workflow that can contribute to WBD errors in pediatric populations.
Speaker:
Garrett Zabala, Master of Science
MedStar Health Research Insititute
Authors:
Garrett Foresman, BS - MedStar Health Research Insititute; Yuuki Unno, MSHS - MedStar Health Research Institute; Sonita Bennett - MedStar Health; Joseph Blumenthal - MedStar Health; Mallory Tidwell, BSN, RN, CCRP - Children's Healthcare of Atlanta; Naveen Muthu, MD - Children's Healthcare of Atlanta; Sadaf Kazi, PhD - National Center for Human Factors in Healthcare, MedStar Health Research Institute;
Garrett
Zabala,
Master of Science - MedStar Health Research Insititute
The Task Performance in a Photorealistic VR Rehabilitation Task Conditioned on Visuospatial and Executive Skills in Young Adults: Insights from a Preliminary Study
Poster Number: P81
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Human-computer Interaction, Quantitative Methods, Real-World Evidence Generation, Usability
Primary Track: Foundations
Programmatic Theme: Clinical Research Informatics
This study explores how cognitive skill levels influence task performance in virtual reality (VR) rehabilitation. Fifteen healthy adults completed a VR-based task, and their visuospatial and executive functioning were assessed using the Block Design Test (BDT). Spearman correlation analyses revealed negative relationships between BDT scores and 64 performance features. Results suggest individuals with higher visuospatial and executive functioning perform VR tasks more efficiently, necessitating cognitively adaptive VR rehabilitation tools for individuals with acquired brain injury.
Speaker:
Fanny D'Souza, MS
Indiana University Indianapolis
Authors:
Fanny D'Souza, MS - Indiana University Indianapolis; Jonathan Liu, BS - Indiana University Indianapolis; Samuel Brunes, MS - Indiana University Indianapolis; Jacob Gibson, MS - Indiana University Indianapolis; Eun Jin Paek, PhD - University of Tennessee Health Science Center; Hee Tae Jung, PhD - Indiana University Indianapolis;
Poster Number: P81
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Human-computer Interaction, Quantitative Methods, Real-World Evidence Generation, Usability
Primary Track: Foundations
Programmatic Theme: Clinical Research Informatics
This study explores how cognitive skill levels influence task performance in virtual reality (VR) rehabilitation. Fifteen healthy adults completed a VR-based task, and their visuospatial and executive functioning were assessed using the Block Design Test (BDT). Spearman correlation analyses revealed negative relationships between BDT scores and 64 performance features. Results suggest individuals with higher visuospatial and executive functioning perform VR tasks more efficiently, necessitating cognitively adaptive VR rehabilitation tools for individuals with acquired brain injury.
Speaker:
Fanny D'Souza, MS
Indiana University Indianapolis
Authors:
Fanny D'Souza, MS - Indiana University Indianapolis; Jonathan Liu, BS - Indiana University Indianapolis; Samuel Brunes, MS - Indiana University Indianapolis; Jacob Gibson, MS - Indiana University Indianapolis; Eun Jin Paek, PhD - University of Tennessee Health Science Center; Hee Tae Jung, PhD - Indiana University Indianapolis;
Fanny
D'Souza,
MS - Indiana University Indianapolis
A Multicenter Distributed Analysis of Routine Healthcare Data on Drug Therapy Safety from German University Hospitals
Poster Number: P82
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Healthcare Quality, Information Extraction, Real-World Evidence Generation, Patient Safety, Interoperability and Health Information Exchange
Primary Track: Applications
Programmatic Theme: Clinical Research Informatics
Drug-related problems remain a leading cause of preventable harm. The Medical Informatics Initiative Germany established data integration centers using HL7® FHIR® to address multicenter analysis barriers. The “POLAR_MI” pipeline across ten German university hospitals (2018–2021) examined upper gastrointestinal bleeding (~1.2%), drug-related hypoglycemia (~2.9%), and potentially inappropriate medications (37.9%) in older adults, detecting significant risk-factor associations. Despite documentation gaps, plausible prevalence estimates emerged, highlighting the feasibility of multi-center, privacy-preserving, large-scale EHR studies to enhance medication safety.
Speaker:
Markues Loeffler, Dr.
Institute for Medical Informatics, Statistics and Epidemiology, Leipzig University, Leipzig, Germany,
Authors:
Daniel Neumann, Medical Informatics and Data Science - University Leipzig, Medical Faculty; Miriam Kesselmeier, Dr. - Institute of Medical Statistics, Computer and Data Sciences, Jena University Hospital, Jena, Germany; Anna Maria Wermund, - - Department of Clinical Pharmacy, Institute of Pharmacy, University of Bonn, Bonn, Germany; Louisa Redeker, - - Department of Clinical Pharmacology, School of Medicine, Faculty of Health, Witten/Herdecke University, Witten, Germany; Florian Schmidt, - - Institute for Medical Informatics, Statistics and Epidemiology, Leipzig University, Leipzig, Germany; Alexander Strübing, - - Institute for Medical Informatics, Statistics and Epidemiology, Leipzig University, Leipzig, Germany; Torsten Thalheim, Dr. - Institute for Medical Informatics, Statistics and Epidemiology, Leipzig University, Leipzig, Germany; Frank Meineke, Dr. - Institute for Medical Informatics, Statistics and Epidemiology, Leipzig University, Leipzig, Germany; Sven Schmiedl, Dr. - Department of Clinical Pharmacology, School of Medicine, Faculty of Health, Witten/Herdecke University; Petra Thürmann, Dr. - Department of Clinical Pharmacology, School of Medicine, Faculty of Health, Witten/Herdecke University; Martin F. Fromm, Dr. - Institute of Experimental and Clinical Pharmacology and Toxicology, Friedrich-Alexander-Universität Erlangen-Nürnberg, Erlangen, Germany (FAU); Renke Maas, Dr. - Institute of Experimental and Clinical Pharmacology and Toxicology, Friedrich-Alexander-Universität Erlangen-Nürnberg, Erlangen, Germany (FAU); André Scherag, Dr. - Institute of Medical Statistics, Computer and Data Sciences, Jena University Hospital, Jena, Germany; Markus Loeffler, Dr. - Institute for Medical Informatics, Statistics and Epidemiology, Leipzig University, Leipzig, Germany;
Poster Number: P82
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Healthcare Quality, Information Extraction, Real-World Evidence Generation, Patient Safety, Interoperability and Health Information Exchange
Primary Track: Applications
Programmatic Theme: Clinical Research Informatics
Drug-related problems remain a leading cause of preventable harm. The Medical Informatics Initiative Germany established data integration centers using HL7® FHIR® to address multicenter analysis barriers. The “POLAR_MI” pipeline across ten German university hospitals (2018–2021) examined upper gastrointestinal bleeding (~1.2%), drug-related hypoglycemia (~2.9%), and potentially inappropriate medications (37.9%) in older adults, detecting significant risk-factor associations. Despite documentation gaps, plausible prevalence estimates emerged, highlighting the feasibility of multi-center, privacy-preserving, large-scale EHR studies to enhance medication safety.
Speaker:
Markues Loeffler, Dr.
Institute for Medical Informatics, Statistics and Epidemiology, Leipzig University, Leipzig, Germany,
Authors:
Daniel Neumann, Medical Informatics and Data Science - University Leipzig, Medical Faculty; Miriam Kesselmeier, Dr. - Institute of Medical Statistics, Computer and Data Sciences, Jena University Hospital, Jena, Germany; Anna Maria Wermund, - - Department of Clinical Pharmacy, Institute of Pharmacy, University of Bonn, Bonn, Germany; Louisa Redeker, - - Department of Clinical Pharmacology, School of Medicine, Faculty of Health, Witten/Herdecke University, Witten, Germany; Florian Schmidt, - - Institute for Medical Informatics, Statistics and Epidemiology, Leipzig University, Leipzig, Germany; Alexander Strübing, - - Institute for Medical Informatics, Statistics and Epidemiology, Leipzig University, Leipzig, Germany; Torsten Thalheim, Dr. - Institute for Medical Informatics, Statistics and Epidemiology, Leipzig University, Leipzig, Germany; Frank Meineke, Dr. - Institute for Medical Informatics, Statistics and Epidemiology, Leipzig University, Leipzig, Germany; Sven Schmiedl, Dr. - Department of Clinical Pharmacology, School of Medicine, Faculty of Health, Witten/Herdecke University; Petra Thürmann, Dr. - Department of Clinical Pharmacology, School of Medicine, Faculty of Health, Witten/Herdecke University; Martin F. Fromm, Dr. - Institute of Experimental and Clinical Pharmacology and Toxicology, Friedrich-Alexander-Universität Erlangen-Nürnberg, Erlangen, Germany (FAU); Renke Maas, Dr. - Institute of Experimental and Clinical Pharmacology and Toxicology, Friedrich-Alexander-Universität Erlangen-Nürnberg, Erlangen, Germany (FAU); André Scherag, Dr. - Institute of Medical Statistics, Computer and Data Sciences, Jena University Hospital, Jena, Germany; Markus Loeffler, Dr. - Institute for Medical Informatics, Statistics and Epidemiology, Leipzig University, Leipzig, Germany;
Markues
Loeffler,
Dr. - Institute for Medical Informatics, Statistics and Epidemiology, Leipzig University, Leipzig, Germany,
Evaluating Large Language Models for Explainable Quality-of-Care Measurement in Treatment of Young Children with ADHD
Poster Number: P83
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Healthcare Quality, Pediatrics, Artificial Intelligence
Primary Track: Applications
Programmatic Theme: Clinical Informatics
We evaluated large language models’ performance for measuring pediatrician adherence to evidence-based guidelines by identifying recommendations for first-line parent training in behavior management (PTBM) treatment in clinical notes of young children with attention-deficit/hyperactivity disorder (ADHD). Using prompting strategies, Claude, GPT-4o, and LLaMA3.3-70B outperformed previous models, achieving high recall (up to 0.91) and generating interpretable explanations for note classifications. This approach enables accurate, explainable, and scalable quality-of-care measurement for ADHD and broader medical conditions.
Speaker:
Malvika Pillai, PhD
Stanford University & VA Palo Alto
Authors:
Fatma Gunturkun, PhD - Stanford University; Ingrid Luo, MS - Stanford University; Tracy Huang, MSPH - Stanford University; Yair Bannett, MD, MS - Stanford University;
Poster Number: P83
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Healthcare Quality, Pediatrics, Artificial Intelligence
Primary Track: Applications
Programmatic Theme: Clinical Informatics
We evaluated large language models’ performance for measuring pediatrician adherence to evidence-based guidelines by identifying recommendations for first-line parent training in behavior management (PTBM) treatment in clinical notes of young children with attention-deficit/hyperactivity disorder (ADHD). Using prompting strategies, Claude, GPT-4o, and LLaMA3.3-70B outperformed previous models, achieving high recall (up to 0.91) and generating interpretable explanations for note classifications. This approach enables accurate, explainable, and scalable quality-of-care measurement for ADHD and broader medical conditions.
Speaker:
Malvika Pillai, PhD
Stanford University & VA Palo Alto
Authors:
Fatma Gunturkun, PhD - Stanford University; Ingrid Luo, MS - Stanford University; Tracy Huang, MSPH - Stanford University; Yair Bannett, MD, MS - Stanford University;
Malvika
Pillai,
PhD - Stanford University & VA Palo Alto
Digital Inclusion Screening Activities in Healthcare Organizations: A Cross-Sectional Survey Study
Poster Number: P84
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Health Equity and Social Determinants of Health (SDoH), Diversity, Equity, Inclusion, and Accessibility, Telemedicine
Primary Track: Policy
Programmatic Theme: Clinical Informatics
In this cross-sectional survey of 144 U.S healthcare organizations, 64 (44.4%) organizations screened for barriers to DI. Organizations that accepted uninsured patients had lower odds (OR=0.32; 95% CI 0.14-0.72) of screening compared to those that did not care for uninsured patients. Increased familiarity with the Affordable Connectivity Program, the Digital Health Equity Act, or the Medicare Advantage digital literacy screening requirement were significantly associated with higher odds of screening.
Speaker:
Robert Ellis, PhD, MHA
University of California Davis (Center for Healthcare Policy and Research)
Authors:
Elaine Khoong, MD, MS - University of California San Francisco; Jonathan J. Shih, BS - UCSF; Vivian E. Kwok, MPH - UCSF; Andersen Yang, MPH - UCSF; Robert Ellis, PhD, MHA - University of California Davis (Center for Healthcare Policy and Research); Jorge Rodriguez, MD - Brigham and Women's Hospital/Harvard Medical School; Lisa Diamond, MD, MPH; Sarah Rahman, MD - Alameda Health System; Emilia De Marchis, MD, MAS; Courtney Lyles, PhD - UC Davis Center for Healthcare Policy and Research; Amy Sheon, PhD, MPH - Public Health Innovators, LLC;
Poster Number: P84
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Health Equity and Social Determinants of Health (SDoH), Diversity, Equity, Inclusion, and Accessibility, Telemedicine
Primary Track: Policy
Programmatic Theme: Clinical Informatics
In this cross-sectional survey of 144 U.S healthcare organizations, 64 (44.4%) organizations screened for barriers to DI. Organizations that accepted uninsured patients had lower odds (OR=0.32; 95% CI 0.14-0.72) of screening compared to those that did not care for uninsured patients. Increased familiarity with the Affordable Connectivity Program, the Digital Health Equity Act, or the Medicare Advantage digital literacy screening requirement were significantly associated with higher odds of screening.
Speaker:
Robert Ellis, PhD, MHA
University of California Davis (Center for Healthcare Policy and Research)
Authors:
Elaine Khoong, MD, MS - University of California San Francisco; Jonathan J. Shih, BS - UCSF; Vivian E. Kwok, MPH - UCSF; Andersen Yang, MPH - UCSF; Robert Ellis, PhD, MHA - University of California Davis (Center for Healthcare Policy and Research); Jorge Rodriguez, MD - Brigham and Women's Hospital/Harvard Medical School; Lisa Diamond, MD, MPH; Sarah Rahman, MD - Alameda Health System; Emilia De Marchis, MD, MAS; Courtney Lyles, PhD - UC Davis Center for Healthcare Policy and Research; Amy Sheon, PhD, MPH - Public Health Innovators, LLC;
Robert
Ellis,
PhD, MHA - University of California Davis (Center for Healthcare Policy and Research)
The Impact of Socioeconomic Status on ICU Admission from the Emergency Department
Poster Number: P85
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Health Equity and Social Determinants of Health (SDoH), Fairness and elimination of bias, Machine Learning, Critical Care, Data Mining, Healthcare Quality, Population Health, Racial disparities
Primary Track: Applications
Programmatic Theme: Clinical Informatics
This study investigates the influence of socioeconomic status on intensive care unit(ICU) admissions coming from the emergency department. Using MIMIC-IV data (n=43,471), we identified five socioeconomic status(SES) clusters through unsupervised clustering analysis. Logit model revealed significant associations between socioeconomic status and ICU admissions that persisted after adjustment for clinical acuity and vital signs. These associations were more pronounced in patients with lower clinical severity, suggesting that non-clinical factors may influence critical care decision-making.
Speaker:
Prathamesh Nitin Bapat, Doctor of philosophy
University of Illinois Chicago
Authors:
Masayuki Teramoto, PhD - Northwestern University; Amy E Krefman, PhD - Northwestern University; Jenny Y Ding, PhD - Northwestern University; Yuan Luo, PhD - Northwestern University;
Poster Number: P85
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Health Equity and Social Determinants of Health (SDoH), Fairness and elimination of bias, Machine Learning, Critical Care, Data Mining, Healthcare Quality, Population Health, Racial disparities
Primary Track: Applications
Programmatic Theme: Clinical Informatics
This study investigates the influence of socioeconomic status on intensive care unit(ICU) admissions coming from the emergency department. Using MIMIC-IV data (n=43,471), we identified five socioeconomic status(SES) clusters through unsupervised clustering analysis. Logit model revealed significant associations between socioeconomic status and ICU admissions that persisted after adjustment for clinical acuity and vital signs. These associations were more pronounced in patients with lower clinical severity, suggesting that non-clinical factors may influence critical care decision-making.
Speaker:
Prathamesh Nitin Bapat, Doctor of philosophy
University of Illinois Chicago
Authors:
Masayuki Teramoto, PhD - Northwestern University; Amy E Krefman, PhD - Northwestern University; Jenny Y Ding, PhD - Northwestern University; Yuan Luo, PhD - Northwestern University;
Prathamesh Nitin
Bapat,
Doctor of philosophy - University of Illinois Chicago
Impact of Community-Level Social Determinants of Health on Predicting 30-Day Readmissions in Spinal Patients using Machine Learning
Poster Number: P86
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Health Equity and Social Determinants of Health (SDoH), Machine Learning, Healthcare Quality
Primary Track: Applications
Programmatic Theme: Clinical Informatics
This study evaluates machine learning (ML) models for predicting 30-day hospital readmissions in spinal surgery patients, incorporating both clinical and community-level social determinants of health (SDoH). Using data from 2,182 admissions, five ML models were trained on pre-surgical and discharge data with and without SDoH features. The inclusion of SDoH improved model performance, particularly in the pre-surgical phase, with an average AUROC increase of 1.84% and AUPRC increase of 19.46%. While discharge models outperformed pre-surgical models overall, the inclusion of SDoH enhanced early risk stratification, offering opportunities for proactive intervention and equitable postoperative care.
Speaker:
Advika Sumit, Undergraduate Student
University of Cincinnati
Authors:
Tzu-Chun Wu, PhD - University of Cincinnati; Advika Sumit, Undergraduate Student - University of Cincinnati; Joseph Cheng, MD - University of Cincinnati College of Medicine; Owoicho Adogwa, MD - University of Cincinnati College of Medicine; Danny Wu, PhD - University of North Carolina at Chapel Hill;
Poster Number: P86
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Health Equity and Social Determinants of Health (SDoH), Machine Learning, Healthcare Quality
Primary Track: Applications
Programmatic Theme: Clinical Informatics
This study evaluates machine learning (ML) models for predicting 30-day hospital readmissions in spinal surgery patients, incorporating both clinical and community-level social determinants of health (SDoH). Using data from 2,182 admissions, five ML models were trained on pre-surgical and discharge data with and without SDoH features. The inclusion of SDoH improved model performance, particularly in the pre-surgical phase, with an average AUROC increase of 1.84% and AUPRC increase of 19.46%. While discharge models outperformed pre-surgical models overall, the inclusion of SDoH enhanced early risk stratification, offering opportunities for proactive intervention and equitable postoperative care.
Speaker:
Advika Sumit, Undergraduate Student
University of Cincinnati
Authors:
Tzu-Chun Wu, PhD - University of Cincinnati; Advika Sumit, Undergraduate Student - University of Cincinnati; Joseph Cheng, MD - University of Cincinnati College of Medicine; Owoicho Adogwa, MD - University of Cincinnati College of Medicine; Danny Wu, PhD - University of North Carolina at Chapel Hill;
Advika
Sumit,
Undergraduate Student - University of Cincinnati
Improving Machine Learning-based Readmission Prediction in a Neurological Intensive Care Unit using Social Determinants of Health
Poster Number: P87
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Health Equity and Social Determinants of Health (SDoH), Machine Learning, Clinical Decision Support
Primary Track: Applications
Programmatic Theme: Clinical Informatics
Unplanned hospital readmissions are costly to patients and hospitals and are important outcome measures in high-risk environments like the neurological intensive care unit (NSICU). While machine learning models have been used to predict neurosurgical outcomes, the impact of social determinants of health (SDoH) data on readmission predictions remains unexplored. This study assessed the addition of SDoH data to structured electronic health record data for readmission prediction, finding that SDoH variables improved model readmission prediction performance.
Speaker:
Mallika Desai, Student
University of Cincinnati College of Medicine
Authors:
Tzu-Chun Wu, PhD - University of Cincinnati; Andy Gao; Mallika Desai, Student - University of Cincinnati College of Medicine; Joseph Cheng, MD, MS - University of Cincinnati College of Medicine; Laura Ngwenya, MD, PhD - University of Cincinnati College of Medicine; Brandon Foreman, MD - University of Cincinnati College of Medicine; Danny Wu, PhD - University of North Carolina at Chapel Hill;
Poster Number: P87
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Health Equity and Social Determinants of Health (SDoH), Machine Learning, Clinical Decision Support
Primary Track: Applications
Programmatic Theme: Clinical Informatics
Unplanned hospital readmissions are costly to patients and hospitals and are important outcome measures in high-risk environments like the neurological intensive care unit (NSICU). While machine learning models have been used to predict neurosurgical outcomes, the impact of social determinants of health (SDoH) data on readmission predictions remains unexplored. This study assessed the addition of SDoH data to structured electronic health record data for readmission prediction, finding that SDoH variables improved model readmission prediction performance.
Speaker:
Mallika Desai, Student
University of Cincinnati College of Medicine
Authors:
Tzu-Chun Wu, PhD - University of Cincinnati; Andy Gao; Mallika Desai, Student - University of Cincinnati College of Medicine; Joseph Cheng, MD, MS - University of Cincinnati College of Medicine; Laura Ngwenya, MD, PhD - University of Cincinnati College of Medicine; Brandon Foreman, MD - University of Cincinnati College of Medicine; Danny Wu, PhD - University of North Carolina at Chapel Hill;
Mallika
Desai,
Student - University of Cincinnati College of Medicine
Comparing Dimensionality Reduction Techniques for Housing Determinants of Health
Poster Number: P88
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Health Equity and Social Determinants of Health (SDoH), Quantitative Methods, Population Health, Machine Learning, Geospatial (GIS) Data/Analysis
Primary Track: Foundations
Programmatic Theme: Public Health Informatics
In this study, we evaluated different dimensionality reduction techniques for summarizing housing conditions as social determinants of health across US geographic levels. We analyzed 15 housing characteristics using American Community Survey data and compared each method performance. PCA demonstrated superior stability, acceptable explainability, and high correlation with underlying variables, suggesting a reliable approach for quantifying housing conditions to inform health disparities research and policy interventions.
Speaker:
Hadi Kharrazi, MD, PhD, FAMIA, FACMI
Johns Hopkins University
Authors:
Xingyu Chen, Master of Science - Johns Hopkins University School of Medicine; Hadi Kharrazi, MD, PhD, FAMIA, FACMI - Johns Hopkins University;
Poster Number: P88
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Health Equity and Social Determinants of Health (SDoH), Quantitative Methods, Population Health, Machine Learning, Geospatial (GIS) Data/Analysis
Primary Track: Foundations
Programmatic Theme: Public Health Informatics
In this study, we evaluated different dimensionality reduction techniques for summarizing housing conditions as social determinants of health across US geographic levels. We analyzed 15 housing characteristics using American Community Survey data and compared each method performance. PCA demonstrated superior stability, acceptable explainability, and high correlation with underlying variables, suggesting a reliable approach for quantifying housing conditions to inform health disparities research and policy interventions.
Speaker:
Hadi Kharrazi, MD, PhD, FAMIA, FACMI
Johns Hopkins University
Authors:
Xingyu Chen, Master of Science - Johns Hopkins University School of Medicine; Hadi Kharrazi, MD, PhD, FAMIA, FACMI - Johns Hopkins University;
Hadi
Kharrazi,
MD, PhD, FAMIA, FACMI - Johns Hopkins University
Exploring the Association Between Social Determinants of Health and Telehealth Utilization for ADHD Among Adults Using Machine Learning: A Cross-Sectional Study
Poster Number: P89
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Health Equity and Social Determinants of Health (SDoH), Telemedicine, Machine Learning, Health Equity, Quantitative Methods, Population Health
Primary Track: Applications
Programmatic Theme: Public Health Informatics
Using 2023 cross-sectional data from NCHS Rapid Surveys System, we examined how social determinants of health affect telehealth utilization among U.S. adults with ADHD. Key predictors of telehealth use varied across ML models, with internet use for doctor communication, difficulty paying medical bills, usual place for care, homeownership, metropolitan status, and gender consistently ranking among top factors. Addressing these challenges can enhance care accessibility and efficiency, guiding policymakers to implement targeted interventions and reduce health disparities.
Speaker:
Yunshu Yang, Ph.D.
University of Minnesota
Authors:
Yunshu Yang, Ph.D. - University of Minnesota; Weijian Qin, Master - Weill Cornell Medicine; Shiqin Tong, Master - Cornell; Hang Liu, Ph.D. - University of Southern California; Zongbo Li, Ph.D. - University of Minnesota; Hawking Yam, Ph.D. - University of Minnesota; Dongze Li, MS. - University of Columbia; Mohan Wang, MS. - Duke University; Xianghan Tan, MS in Health Informatics - Weill Cornell Medicine; Jin Huang, BS. - School of Public Health, Stanford University; Jose Florez-Arango, MD MS PhD - Weill Cornell Medicine;
Poster Number: P89
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Health Equity and Social Determinants of Health (SDoH), Telemedicine, Machine Learning, Health Equity, Quantitative Methods, Population Health
Primary Track: Applications
Programmatic Theme: Public Health Informatics
Using 2023 cross-sectional data from NCHS Rapid Surveys System, we examined how social determinants of health affect telehealth utilization among U.S. adults with ADHD. Key predictors of telehealth use varied across ML models, with internet use for doctor communication, difficulty paying medical bills, usual place for care, homeownership, metropolitan status, and gender consistently ranking among top factors. Addressing these challenges can enhance care accessibility and efficiency, guiding policymakers to implement targeted interventions and reduce health disparities.
Speaker:
Yunshu Yang, Ph.D.
University of Minnesota
Authors:
Yunshu Yang, Ph.D. - University of Minnesota; Weijian Qin, Master - Weill Cornell Medicine; Shiqin Tong, Master - Cornell; Hang Liu, Ph.D. - University of Southern California; Zongbo Li, Ph.D. - University of Minnesota; Hawking Yam, Ph.D. - University of Minnesota; Dongze Li, MS. - University of Columbia; Mohan Wang, MS. - Duke University; Xianghan Tan, MS in Health Informatics - Weill Cornell Medicine; Jin Huang, BS. - School of Public Health, Stanford University; Jose Florez-Arango, MD MS PhD - Weill Cornell Medicine;
Yunshu
Yang,
Ph.D. - University of Minnesota
Assessing the Representation of Disaster Hazards in Standardized Clinical Terminologies: A Study of ICD-10, ICD-11, and LOINC
Poster Number: P90
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Environmental Health and Climate Informatics, Data Standards, Clinical Decision Support
Working Group: Climate, Health and Informatics Working Group
Primary Track: Applications
Programmatic Theme: Public Health Informatics
Climate-driven disasters pose growing health risks, yet it remains unclear how well clinical terminologies capture related hazard concepts. This study assessed 78 disaster hazards from the UNDRR-ISC HIP across ICD-10, ICD-11, and LOINC. Results showed major gaps, especially in LOINC and environmental categories. Many hazards lacked consistent or specific representation. Expanding clinical terminologies to include disaster-related terms is essential to improve data interoperability, support public health response, and address climate-related health challenges.
Speaker:
Beenish Chaudhry, PhD
University of Louisiana at Lafayette
Authors:
Mohammad Shafi, MSc - University of Louisiana at Lafayette; Manijeh Berenji, MD MPH - UC Irvine School of Medicine/Joe C Wen School of Population and Public Health (at UC Irvine); VA Long Beach Healthcare System; Stefan Wheat, MD - University of Washington; Chethan Sarabu, MD - Cornell Tech; Zerina Lokmic-Tomkins, PhD - Monash University; Titus Schleyer, DMD, PhD, MBA - Regenstrief Institute;
Poster Number: P90
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Environmental Health and Climate Informatics, Data Standards, Clinical Decision Support
Working Group: Climate, Health and Informatics Working Group
Primary Track: Applications
Programmatic Theme: Public Health Informatics
Climate-driven disasters pose growing health risks, yet it remains unclear how well clinical terminologies capture related hazard concepts. This study assessed 78 disaster hazards from the UNDRR-ISC HIP across ICD-10, ICD-11, and LOINC. Results showed major gaps, especially in LOINC and environmental categories. Many hazards lacked consistent or specific representation. Expanding clinical terminologies to include disaster-related terms is essential to improve data interoperability, support public health response, and address climate-related health challenges.
Speaker:
Beenish Chaudhry, PhD
University of Louisiana at Lafayette
Authors:
Mohammad Shafi, MSc - University of Louisiana at Lafayette; Manijeh Berenji, MD MPH - UC Irvine School of Medicine/Joe C Wen School of Population and Public Health (at UC Irvine); VA Long Beach Healthcare System; Stefan Wheat, MD - University of Washington; Chethan Sarabu, MD - Cornell Tech; Zerina Lokmic-Tomkins, PhD - Monash University; Titus Schleyer, DMD, PhD, MBA - Regenstrief Institute;
Beenish
Chaudhry,
PhD - University of Louisiana at Lafayette
Data to Action: Using Environmental Drivers of Health Data and Messaging Standards to Improve Decision-Making in Population Health
Poster Number: P91
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Environmental Health and Climate Informatics, Data Standards, Interoperability and Health Information Exchange, Public Health, Machine Learning
Working Group: Climate, Health and Informatics Working Group
Primary Track: Applications
Programmatic Theme: Public Health Informatics
Environmental Drivers of Health (EDOH) are recognized as factors influencing individual and population health outcomes. Integrating EDOH data into electronic health records (EHRs) and public health information systems can improve clinical decision making and real-time response, but there is limited adoption. A multidisciplinary group of practitioners convened stakeholders and conducted a study to identify applications of EDOH data, including data-sharing to identify heat risk and using machine learning to predict air pollution’s impact on health.
Speaker:
Priyanka Surio, MPH, PMP, CHES
EMI Advisors
Authors:
Beenish Chaudhry, PhD - University of Louisiana at Lafayette; Titus Schleyer, DMD, PhD, MBA - Regenstrief Institute; Manijeh Berenji, MD MPH - UC Irvine School of Medicine/Joe C Wen School of Population and Public Health (at UC Irvine); VA Long Beach Healthcare System; Sarah DeSilvey, DNP, FNP-C - The Gravity Project; Chethan Sarabu, MD - Cornell Tech;
Poster Number: P91
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Environmental Health and Climate Informatics, Data Standards, Interoperability and Health Information Exchange, Public Health, Machine Learning
Working Group: Climate, Health and Informatics Working Group
Primary Track: Applications
Programmatic Theme: Public Health Informatics
Environmental Drivers of Health (EDOH) are recognized as factors influencing individual and population health outcomes. Integrating EDOH data into electronic health records (EHRs) and public health information systems can improve clinical decision making and real-time response, but there is limited adoption. A multidisciplinary group of practitioners convened stakeholders and conducted a study to identify applications of EDOH data, including data-sharing to identify heat risk and using machine learning to predict air pollution’s impact on health.
Speaker:
Priyanka Surio, MPH, PMP, CHES
EMI Advisors
Authors:
Beenish Chaudhry, PhD - University of Louisiana at Lafayette; Titus Schleyer, DMD, PhD, MBA - Regenstrief Institute; Manijeh Berenji, MD MPH - UC Irvine School of Medicine/Joe C Wen School of Population and Public Health (at UC Irvine); VA Long Beach Healthcare System; Sarah DeSilvey, DNP, FNP-C - The Gravity Project; Chethan Sarabu, MD - Cornell Tech;
Priyanka
Surio,
MPH, PMP, CHES - EMI Advisors
Harnessing Geographic Information System Technology to Support Colleagues and Consumers during the 2025 California Wildfires
Poster Number: P92
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Environmental Health and Climate Informatics, Geospatial (GIS) Data/Analysis, Population Health, Real-World Evidence Generation
Working Group: Climate, Health and Informatics Working Group
Primary Track: Applications
Programmatic Theme: Public Health Informatics
We describe real-world outcomes from the an informatics-enabled, geospatially enriched framework to support operational resiliency and preserve continuity of care for a large, integrated healthcare enterprise. Retrospective, mixed-methods analysis of a centralized data warehouse characterized mitigation actions in response to the Southern California wildfires (FM-5549-CA, FM-55510-CA, DR-4856-CA; January 7 - 31, 2025); the first major disaster event of the 2025 year. Results highlight accelerated hyperlocal response enabled by GIS technology.
Speaker:
Amanda Zaleski, PhD, MS
CVS Health
Authors:
Amanda Zaleski, PhD, MS - CVS Health; Joshua Wright, Bachelor of Science - CVS Health; Sean Horman, MPA - CVS Health; Kelly Jean Craig, PhD - CVS Health; Patrick Getler, MS - CVS Health; Travis Andrews, BSBA - CVS Health; Eleanor Beltz, PhD, ATC - Aetna Medical Affairs, CVS Health; Samson Williams, MA - CVS Health; Eric Simoni, MBA - CVS Health; Paige Wickner, MD, MPH - CVS Health; Sreekanth Chaguturu, MD - CVS Health;
Poster Number: P92
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Environmental Health and Climate Informatics, Geospatial (GIS) Data/Analysis, Population Health, Real-World Evidence Generation
Working Group: Climate, Health and Informatics Working Group
Primary Track: Applications
Programmatic Theme: Public Health Informatics
We describe real-world outcomes from the an informatics-enabled, geospatially enriched framework to support operational resiliency and preserve continuity of care for a large, integrated healthcare enterprise. Retrospective, mixed-methods analysis of a centralized data warehouse characterized mitigation actions in response to the Southern California wildfires (FM-5549-CA, FM-55510-CA, DR-4856-CA; January 7 - 31, 2025); the first major disaster event of the 2025 year. Results highlight accelerated hyperlocal response enabled by GIS technology.
Speaker:
Amanda Zaleski, PhD, MS
CVS Health
Authors:
Amanda Zaleski, PhD, MS - CVS Health; Joshua Wright, Bachelor of Science - CVS Health; Sean Horman, MPA - CVS Health; Kelly Jean Craig, PhD - CVS Health; Patrick Getler, MS - CVS Health; Travis Andrews, BSBA - CVS Health; Eleanor Beltz, PhD, ATC - Aetna Medical Affairs, CVS Health; Samson Williams, MA - CVS Health; Eric Simoni, MBA - CVS Health; Paige Wickner, MD, MPH - CVS Health; Sreekanth Chaguturu, MD - CVS Health;
Amanda
Zaleski,
PhD, MS - CVS Health
Rapid Review of Models Assessing Suicide Risk from Patient Portal and Crisis Text Line Messages
Poster Number: P93
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Diagnostic Systems, Information Retrieval, Clinical Decision Support, Natural Language Processing, Patient / Person Generated Health Data (Patient Reported Outcomes)
Primary Track: Applications
Programmatic Theme: Clinical Informatics
Despite significant amounts of research on the widespread adoption of patient portals, research has been limited with respect to how people use patient portals to communicate suicidality. This project looks at previous studies characterizing the communication of suicidality in patient portals and crisis text lines. We identified 11 papers describing a total of 80 models. None of the models were implemented and tested for clinical utility. None of the models utilized large-language models.
Speaker:
Hannah Slater, MS
Vanderbilt University Department of Biomedical Informatics
Authors:
Hannah Slater, MS - Vanderbilt University Department of Biomedical Informatics; Colin Walsh, MD MA - Department of Biomedical Informatics, Vanderbilt University;
Poster Number: P93
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Diagnostic Systems, Information Retrieval, Clinical Decision Support, Natural Language Processing, Patient / Person Generated Health Data (Patient Reported Outcomes)
Primary Track: Applications
Programmatic Theme: Clinical Informatics
Despite significant amounts of research on the widespread adoption of patient portals, research has been limited with respect to how people use patient portals to communicate suicidality. This project looks at previous studies characterizing the communication of suicidality in patient portals and crisis text lines. We identified 11 papers describing a total of 80 models. None of the models were implemented and tested for clinical utility. None of the models utilized large-language models.
Speaker:
Hannah Slater, MS
Vanderbilt University Department of Biomedical Informatics
Authors:
Hannah Slater, MS - Vanderbilt University Department of Biomedical Informatics; Colin Walsh, MD MA - Department of Biomedical Informatics, Vanderbilt University;
Hannah
Slater,
MS - Vanderbilt University Department of Biomedical Informatics
MADS: Multi-agent Dynamic Synergy for Interpretable Medical Diagnosis
Poster Number: P94
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Diagnostic Systems, Large Language Models (LLMs), Bioinformatics
Primary Track: Foundations
Programmatic Theme: Clinical Informatics
We propose MADS: a Multi-agent Dynamic Synergy framework designed to integrate specialized AI agents for context-aware, collaborative reasoning in complex diagnostic scenarios. MADS enhances diagnostic precision through dynamic agent interactions. This approach improves accuracy without the need for fine-tuning on medical datasets, effectively bridging the gap between standalone AI models and clinical needs through adaptive, explainable decision-making.
Speaker:
Fan Ma, Phd
yale
Authors:
Fan Ma, Phd - yale; Qianqian Xie, PhD - Yale University; Lingfei Qian, PHD - Yale University; Hua Xu, Ph.D - Yale University;
Poster Number: P94
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Diagnostic Systems, Large Language Models (LLMs), Bioinformatics
Primary Track: Foundations
Programmatic Theme: Clinical Informatics
We propose MADS: a Multi-agent Dynamic Synergy framework designed to integrate specialized AI agents for context-aware, collaborative reasoning in complex diagnostic scenarios. MADS enhances diagnostic precision through dynamic agent interactions. This approach improves accuracy without the need for fine-tuning on medical datasets, effectively bridging the gap between standalone AI models and clinical needs through adaptive, explainable decision-making.
Speaker:
Fan Ma, Phd
yale
Authors:
Fan Ma, Phd - yale; Qianqian Xie, PhD - Yale University; Lingfei Qian, PHD - Yale University; Hua Xu, Ph.D - Yale University;
Fan
Ma,
Phd - yale
MeCaMIL: Causal Multiple Instance Learning for Medical Whole Slide Image Diagnosis
Poster Number: P95
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Diagnostic Systems, Patient / Person Generated Health Data (Patient Reported Outcomes), Causal Inference
Primary Track: Applications
Programmatic Theme: Academic Informatics / LIEAF
Multiple Instance Learning (MIL) is widely used in whole slide image (WSI) analysis. Despite strong classification performance, these methods lack interpretability and struggle to integrate additional diagnostic data, such as demographics, limiting their robustness and fairness. We propose a novel causality-based MIL approach that models relationships between image patches via a causal graph and incorporates non-image data (e.g., race, age) as exogenous variables. This enables more effective multi-modal integration beyond simple feature concatenation.
Speaker:
Yiran Song, doctor
University of Minnesota
Authors:
Rui Zhang, PhD, FAMIA, FACMI - University of Minnesota, Twin Cities; Mingquan Lin, PhD - University of Minnesota; Yiran Song, doctor - University of Minnesota; Yikai Zhang, BA - University of Minnesota;
Poster Number: P95
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Diagnostic Systems, Patient / Person Generated Health Data (Patient Reported Outcomes), Causal Inference
Primary Track: Applications
Programmatic Theme: Academic Informatics / LIEAF
Multiple Instance Learning (MIL) is widely used in whole slide image (WSI) analysis. Despite strong classification performance, these methods lack interpretability and struggle to integrate additional diagnostic data, such as demographics, limiting their robustness and fairness. We propose a novel causality-based MIL approach that models relationships between image patches via a causal graph and incorporates non-image data (e.g., race, age) as exogenous variables. This enables more effective multi-modal integration beyond simple feature concatenation.
Speaker:
Yiran Song, doctor
University of Minnesota
Authors:
Rui Zhang, PhD, FAMIA, FACMI - University of Minnesota, Twin Cities; Mingquan Lin, PhD - University of Minnesota; Yiran Song, doctor - University of Minnesota; Yikai Zhang, BA - University of Minnesota;
Yiran
Song,
doctor - University of Minnesota
Leveraging Publicly Available Home Health Agency Performance Data for Quality Improvement Purposes: Development of an Information Resource for Increased Quality Awareness and Decision Support
Poster Number: P96
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Delivering Health Information and Knowledge to the Public, Delivering Health Information and Knowledge to the Public, Delivering Health Information and Knowledge to the Public
Primary Track: Applications
Home healthcare quality varies across the U.S., yet public data remains hard hard to access and understand. We developed an online portal to visualize nationwide quality indicators using CMS and CAHPS data. Built through agile methods and user feedback, the portal features filters, maps, and scorecards for exploring key performance data. The portal’s usability (ease of use) was rated highly (SUS score: 92.02). It improves data access, informed decisions, and transparency in home healthcare quality.
Speaker:
Ali Alsarhan
University of Maryland Baltimore County
Authors:
Güneş Koru, PhD, FAMIA - University of Arkansas for Medical Sciences, Northwest Regional Campus; Ali Alsarhan - University of Maryland Baltimore County;
Poster Number: P96
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Delivering Health Information and Knowledge to the Public, Delivering Health Information and Knowledge to the Public, Delivering Health Information and Knowledge to the Public
Primary Track: Applications
Home healthcare quality varies across the U.S., yet public data remains hard hard to access and understand. We developed an online portal to visualize nationwide quality indicators using CMS and CAHPS data. Built through agile methods and user feedback, the portal features filters, maps, and scorecards for exploring key performance data. The portal’s usability (ease of use) was rated highly (SUS score: 92.02). It improves data access, informed decisions, and transparency in home healthcare quality.
Speaker:
Ali Alsarhan
University of Maryland Baltimore County
Authors:
Güneş Koru, PhD, FAMIA - University of Arkansas for Medical Sciences, Northwest Regional Campus; Ali Alsarhan - University of Maryland Baltimore County;
Ali
Alsarhan - University of Maryland Baltimore County
Enhancing Digital Health Literacy in Caregivers Through a Workshop: Insights from a Pilot Eye-Tracking Study
Poster Number: P97
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Delivering Health Information and Knowledge to the Public, Education and Training, Human-computer Interaction
Primary Track: Applications
Programmatic Theme: Consumer Health Informatics
This pilot study evaluated the impact of a digital health literacy workshop on caregivers' ability to assess online health information and identify challenges in improving caregivers' digital health literacy using eye-tracking technology. Post-workshop, caregivers showed increased attention to scam signs, reduced cognitive effort, and improved search efficiency. Findings suggest the workshop enhances digital skills, though challenges remain. Future workshop should provide diverse educational materials and offer personalized feedback to participants.
Speaker:
Fei Yu, PhD
UNC at Chapel Hill
Authors:
Xiaomeng Wang, Master of Science - University of Texas Health Science Center at San Antonio; Kayla Morales, MA - University of Texas Health Science Center at San Antonio; Fei Yu, PhD - UNC at Chapel Hill; Lixin Song, PhD; Jia Liu, PhD - UT Health San Antonio; Tian Wang, PhD - University of Texas Health Science Center at San Antonio; Francisco Hernandez, BA - University of Texas Health Science Center at San Antonio; Roxana Delgado, PhD - University of Texas Health Science Center at San Antonio;
Poster Number: P97
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Delivering Health Information and Knowledge to the Public, Education and Training, Human-computer Interaction
Primary Track: Applications
Programmatic Theme: Consumer Health Informatics
This pilot study evaluated the impact of a digital health literacy workshop on caregivers' ability to assess online health information and identify challenges in improving caregivers' digital health literacy using eye-tracking technology. Post-workshop, caregivers showed increased attention to scam signs, reduced cognitive effort, and improved search efficiency. Findings suggest the workshop enhances digital skills, though challenges remain. Future workshop should provide diverse educational materials and offer personalized feedback to participants.
Speaker:
Fei Yu, PhD
UNC at Chapel Hill
Authors:
Xiaomeng Wang, Master of Science - University of Texas Health Science Center at San Antonio; Kayla Morales, MA - University of Texas Health Science Center at San Antonio; Fei Yu, PhD - UNC at Chapel Hill; Lixin Song, PhD; Jia Liu, PhD - UT Health San Antonio; Tian Wang, PhD - University of Texas Health Science Center at San Antonio; Francisco Hernandez, BA - University of Texas Health Science Center at San Antonio; Roxana Delgado, PhD - University of Texas Health Science Center at San Antonio;
Fei
Yu,
PhD - UNC at Chapel Hill
Membership Inference Attacks against Multi-Institutional Medical Image Classification
Poster Number: P98
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Data Sharing, Deep Learning, Privacy and Security
Working Group: Ethical, Legal, and Social Issues Working Group
Primary Track: Applications
Programmatic Theme: Clinical Research Informatics
Advancements in deep learning made medical image classification for disease diagnosis increasingly viable. Multiple institutions may share models or train a model together for improved generalization ability. However, sharing models and anonymized data are vulnerable to membership inference attacks (MIA). This study investigates MIA on tuberculosis diagnosis using chest X-rays, employing six CNN architectures and compares two attack models. Our results demonstrate that both attack models are effective, highlighting the need for privacy-preserving approaches in multi-institutional collaborations.
Speaker:
Hongzhu Jiang, N/A
ShanghaiTech University
Authors:
Hongzhu Jiang, N/A - ShanghaiTech University; Jiayue Hou, N/A - ShanghaiTech University; Sihan Xie, N/A - ShanghaiTech University; Bradley Malin, PhD - Vanderbilt University Medical Center; Zhiyu Wan, PhD - ShanghaiTech University;
Poster Number: P98
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Data Sharing, Deep Learning, Privacy and Security
Working Group: Ethical, Legal, and Social Issues Working Group
Primary Track: Applications
Programmatic Theme: Clinical Research Informatics
Advancements in deep learning made medical image classification for disease diagnosis increasingly viable. Multiple institutions may share models or train a model together for improved generalization ability. However, sharing models and anonymized data are vulnerable to membership inference attacks (MIA). This study investigates MIA on tuberculosis diagnosis using chest X-rays, employing six CNN architectures and compares two attack models. Our results demonstrate that both attack models are effective, highlighting the need for privacy-preserving approaches in multi-institutional collaborations.
Speaker:
Hongzhu Jiang, N/A
ShanghaiTech University
Authors:
Hongzhu Jiang, N/A - ShanghaiTech University; Jiayue Hou, N/A - ShanghaiTech University; Sihan Xie, N/A - ShanghaiTech University; Bradley Malin, PhD - Vanderbilt University Medical Center; Zhiyu Wan, PhD - ShanghaiTech University;
Hongzhu
Jiang,
N/A - ShanghaiTech University
Identifying Barriers in Implementing EMERSE Network Functionality: Findings from a Multi-Site Collaboration
Poster Number: P99
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Data Sharing, Governance, Legal, Ethical, Social and Regulatory Issues
Primary Track: Applications
Programmatic Theme: Clinical Research Informatics
This study aimed to develop a framework for broadening the adoption of the Electronic Medical Record Search Engine (EMERSE) network by examining regulatory, security, and compliance challenges across five institutions. Key barriers included diverse consultative requirements, slow approval processes, lack of structured pathways for authorization, excessive data-sharing caution, and limited demand/awareness. Despite robust technical documentation and architecture, socio-technical and bureaucratic obstacles limited the activation of EMERSE’s network feature, highlighting the need for addressing institutional resistance.
Speaker:
David Hanauer, MD
University of Michigan
Authors:
David Hanauer, MD - University of Michigan; Donald Brown, Ph.D. - University of Virginia; Lisa Ferguson, MS - University of Michigan; Daniel Harris, PhD - University of Kentucky; Jong Jeong, PhD - University of Kentucky; Jason Keller, MS - University of Cincinnati Center for Health Informatics; Karthik Natarajan, PhD - Columbia University Dept of Biomedical Informatics; Janie Weiss, BS - Columbia University;
Poster Number: P99
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Data Sharing, Governance, Legal, Ethical, Social and Regulatory Issues
Primary Track: Applications
Programmatic Theme: Clinical Research Informatics
This study aimed to develop a framework for broadening the adoption of the Electronic Medical Record Search Engine (EMERSE) network by examining regulatory, security, and compliance challenges across five institutions. Key barriers included diverse consultative requirements, slow approval processes, lack of structured pathways for authorization, excessive data-sharing caution, and limited demand/awareness. Despite robust technical documentation and architecture, socio-technical and bureaucratic obstacles limited the activation of EMERSE’s network feature, highlighting the need for addressing institutional resistance.
Speaker:
David Hanauer, MD
University of Michigan
Authors:
David Hanauer, MD - University of Michigan; Donald Brown, Ph.D. - University of Virginia; Lisa Ferguson, MS - University of Michigan; Daniel Harris, PhD - University of Kentucky; Jong Jeong, PhD - University of Kentucky; Jason Keller, MS - University of Cincinnati Center for Health Informatics; Karthik Natarajan, PhD - Columbia University Dept of Biomedical Informatics; Janie Weiss, BS - Columbia University;
David
Hanauer,
MD - University of Michigan
Evaluating Differences in EHR Audit Log Cognitive Load Metrics for Patients Appropriately and Not Appropriately Prescribed Statins
Poster Number: P100
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Data Mining, Clinical Guidelines, Workflow, Evaluation, Quantitative Methods, Information Extraction
Primary Track: Applications
Programmatic Theme: Clinical Informatics
High cognitive load tasks in clinical environments contribute to provider burnout and impact patient care. This study examines EHR audit log data to assess cognitive load during primary care encounters and its relationship to statin initiation for at-risk patients. Results suggest that higher cognitive load metrics are associated with increased statin prescribing. Findings highlight opportunities to enhance EHR usability and workflow efficiency to support quality-of-care measures and provider decision-making.
Speaker:
Ratnalekha Viswanadham, PhD, PhD
NYU Grossman School of Medicine
Authors:
Yuhan Cui, MS - NYU Langone; Nicole Redfern, MPH - NYU Langone Health; Angela Mastrianni, PhD - NYU Langone Health; Priyanka Solanki, MD - NYU; Amelia Shunk, MMCi - NYU Grossman School of Medicine; Safiya Richardson, MD, MPH - New York University;
Poster Number: P100
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Data Mining, Clinical Guidelines, Workflow, Evaluation, Quantitative Methods, Information Extraction
Primary Track: Applications
Programmatic Theme: Clinical Informatics
High cognitive load tasks in clinical environments contribute to provider burnout and impact patient care. This study examines EHR audit log data to assess cognitive load during primary care encounters and its relationship to statin initiation for at-risk patients. Results suggest that higher cognitive load metrics are associated with increased statin prescribing. Findings highlight opportunities to enhance EHR usability and workflow efficiency to support quality-of-care measures and provider decision-making.
Speaker:
Ratnalekha Viswanadham, PhD, PhD
NYU Grossman School of Medicine
Authors:
Yuhan Cui, MS - NYU Langone; Nicole Redfern, MPH - NYU Langone Health; Angela Mastrianni, PhD - NYU Langone Health; Priyanka Solanki, MD - NYU; Amelia Shunk, MMCi - NYU Grossman School of Medicine; Safiya Richardson, MD, MPH - New York University;
Ratnalekha
Viswanadham,
PhD, PhD - NYU Grossman School of Medicine
How Researchers Claim Novelty in Biomedical Science: A Taxonomy for Understanding Innovation
Poster Number: P101
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Data Mining, Controlled Terminologies, Ontologies, and Vocabularies, Natural Language Processing, Large Language Models (LLMs)
Primary Track: Foundations
Scientific novelty drives biomedical progress, yet its forms are poorly defined. We present NovelTax, a hierarchical taxonomy, Concept, Method and/or Material Novelty, and Finding Novelty, developed using LLM-assisted annotation of a self-built corpus. Evaluated for semantic coherence and coverage, NovelTax reveals that Finding Novelty dominates (48%), while Concept Novelty is rare (19%). This framework enables structured innovation assessment, aiding research analysis, funding decisions, and strategic planning across biomedicine and beyond.
Speaker:
Xueqing Peng, PhD
Yale University
Authors:
Xueqing Peng, PhD - Yale University; Huan He, Ph.D. - Yale University; Rui Shi, Bachelor - Yale University; Vipina K. Keloth, PhD - Yale University; Lingfei Qian, PHD - Yale University; Yan Hu, MS - UTHealth Science Center Houston; Jimin Huang, MS - Yale University; Qianqian Xie, PhD - Yale University; Na Hong, PhD - Yale University; Hua Xu, Ph.D - Yale University;
Poster Number: P101
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Data Mining, Controlled Terminologies, Ontologies, and Vocabularies, Natural Language Processing, Large Language Models (LLMs)
Primary Track: Foundations
Scientific novelty drives biomedical progress, yet its forms are poorly defined. We present NovelTax, a hierarchical taxonomy, Concept, Method and/or Material Novelty, and Finding Novelty, developed using LLM-assisted annotation of a self-built corpus. Evaluated for semantic coherence and coverage, NovelTax reveals that Finding Novelty dominates (48%), while Concept Novelty is rare (19%). This framework enables structured innovation assessment, aiding research analysis, funding decisions, and strategic planning across biomedicine and beyond.
Speaker:
Xueqing Peng, PhD
Yale University
Authors:
Xueqing Peng, PhD - Yale University; Huan He, Ph.D. - Yale University; Rui Shi, Bachelor - Yale University; Vipina K. Keloth, PhD - Yale University; Lingfei Qian, PHD - Yale University; Yan Hu, MS - UTHealth Science Center Houston; Jimin Huang, MS - Yale University; Qianqian Xie, PhD - Yale University; Na Hong, PhD - Yale University; Hua Xu, Ph.D - Yale University;
Xueqing
Peng,
PhD - Yale University
Deeper Insights with Structured and Unstructured Data
Poster Number: P102
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Data Mining, Information Retrieval, Natural Language Processing
Primary Track: Applications
Programmatic Theme: Clinical Research Informatics
Eighty percent of clinical data is reportedly unstructured, yet secondary use research primarily relies on structured data. Using a multimodal approach, we combined querying structured data (ICD-10, LOINC) with searching unstructured text from 300 million documents across 18 U.S. sites. Keyword search expanded cohort identification by a median of 23.9%. This method provides a scalable, low-resource alternative for researchers to retrieve relevant clinical data without requiring NLP-based tools.
Speaker:
John Doole, Pharm. D., MFA
TriNetX, LLC
Author:
Matvey Palchuk, MD, MS, FAMIA - TriNetX, LLC.;
Poster Number: P102
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Data Mining, Information Retrieval, Natural Language Processing
Primary Track: Applications
Programmatic Theme: Clinical Research Informatics
Eighty percent of clinical data is reportedly unstructured, yet secondary use research primarily relies on structured data. Using a multimodal approach, we combined querying structured data (ICD-10, LOINC) with searching unstructured text from 300 million documents across 18 U.S. sites. Keyword search expanded cohort identification by a median of 23.9%. This method provides a scalable, low-resource alternative for researchers to retrieve relevant clinical data without requiring NLP-based tools.
Speaker:
John Doole, Pharm. D., MFA
TriNetX, LLC
Author:
Matvey Palchuk, MD, MS, FAMIA - TriNetX, LLC.;
John
Doole,
Pharm. D., MFA - TriNetX, LLC
TAMER: A Test-Time Adaptive MoE-Driven Framework for EHR Representation Learning
Poster Number: P103
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Data Mining, Knowledge Representation and Information Modeling, Deep Learning, Personal Health Informatics
Primary Track: Applications
Programmatic Theme: Clinical Informatics
We propose TAMER, a Test-time Adaptive MoE-driven framework for Electronic Health Record (EHR) Representation learning. TAMER introduces a framework where a Mixture-of-Experts (MoE) architecture is co-designed with Test-Time Adaptation (TTA) to jointly mitigate the intertwined challenges of patient heterogeneity and distribution shifts in EHR modeling. The MoE focuses on latent patient subgroups through domain-aware expert specialization, while TTA enables real-time adaptation to evolving health status distributions when new patient samples are introduced. Extensive experiments across four real-world EHR datasets demonstrate that TAMER consistently improves predictive performance for both mortality and readmission risk tasks when combined with diverse EHR modeling backbones. TAMER offers a promising approach for dynamic and personalized EHR-based predictions in practical clinical settings.
Speaker:
Xiaochen Zheng, Master of Science
University of Zurich
Authors:
Yinghao Zhu, Master of Science - University of Zurich; Xiaochen Zheng, Master of Science - University of Zurich; Ahmed Allam, PhD - University of Zurich; Michael Krauthammer, PhD - University of Zurich;
Poster Number: P103
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Data Mining, Knowledge Representation and Information Modeling, Deep Learning, Personal Health Informatics
Primary Track: Applications
Programmatic Theme: Clinical Informatics
We propose TAMER, a Test-time Adaptive MoE-driven framework for Electronic Health Record (EHR) Representation learning. TAMER introduces a framework where a Mixture-of-Experts (MoE) architecture is co-designed with Test-Time Adaptation (TTA) to jointly mitigate the intertwined challenges of patient heterogeneity and distribution shifts in EHR modeling. The MoE focuses on latent patient subgroups through domain-aware expert specialization, while TTA enables real-time adaptation to evolving health status distributions when new patient samples are introduced. Extensive experiments across four real-world EHR datasets demonstrate that TAMER consistently improves predictive performance for both mortality and readmission risk tasks when combined with diverse EHR modeling backbones. TAMER offers a promising approach for dynamic and personalized EHR-based predictions in practical clinical settings.
Speaker:
Xiaochen Zheng, Master of Science
University of Zurich
Authors:
Yinghao Zhu, Master of Science - University of Zurich; Xiaochen Zheng, Master of Science - University of Zurich; Ahmed Allam, PhD - University of Zurich; Michael Krauthammer, PhD - University of Zurich;
Xiaochen
Zheng,
Master of Science - University of Zurich
Associations between trajectories of serum phosphate and the in-hospital mortality for patients with AKI in ICU
Poster Number: P104
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Critical Care, Clinical Decision Support, Population Health
Primary Track: Applications
Programmatic Theme: Clinical Informatics
This study employed the MIMIC-IV database to explore associations between serum phosphate trajectories and in-hospital mortality for patients with AKI acquired in ICU. Three serum phosphate trajectories were identified: continuously normal (Trajectory 1, 82.75%), hyperphosphatemia with decreasing trend (Trajectory 2,4.84%), and hyperphosphatemia with gradually increasing trend (Trajectory 3,12.41%). Logistic regression analysis revealed that patients with Trajectory 3 were associated with a higher risk of in-hospital mortality (OR=1.38, p=0.034), but not in patients with Trajectory 2. Subgroup analysis stratified by trajectories showed that the usage of diuretics was associated with a reduced in-hospital mortality in patients with Trajectory 1 (OR=0.58, p<0.001) and Trajectory 3 (OR=0.42, p=0.001), whereas renal replacement therapy was associated with an increased in-hospital mortality in patients with Trajectory 1 (OR=3.46, p<0.001) and Trajectory 2 (OR=11.5, p=0.002). These findings highlight the prognostic value of phosphate trajectories for AKI patients in ICU.
Speaker:
Guilan Kong, PhD
National Institute of Health Data Science, Peking University
Authors:
Haoran Su, MS - Peking University; Tongyue Shi, MS - Peking University; Guilan Kong, PhD - National Institute of Health Data Science, Peking University;
Poster Number: P104
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Critical Care, Clinical Decision Support, Population Health
Primary Track: Applications
Programmatic Theme: Clinical Informatics
This study employed the MIMIC-IV database to explore associations between serum phosphate trajectories and in-hospital mortality for patients with AKI acquired in ICU. Three serum phosphate trajectories were identified: continuously normal (Trajectory 1, 82.75%), hyperphosphatemia with decreasing trend (Trajectory 2,4.84%), and hyperphosphatemia with gradually increasing trend (Trajectory 3,12.41%). Logistic regression analysis revealed that patients with Trajectory 3 were associated with a higher risk of in-hospital mortality (OR=1.38, p=0.034), but not in patients with Trajectory 2. Subgroup analysis stratified by trajectories showed that the usage of diuretics was associated with a reduced in-hospital mortality in patients with Trajectory 1 (OR=0.58, p<0.001) and Trajectory 3 (OR=0.42, p=0.001), whereas renal replacement therapy was associated with an increased in-hospital mortality in patients with Trajectory 1 (OR=3.46, p<0.001) and Trajectory 2 (OR=11.5, p=0.002). These findings highlight the prognostic value of phosphate trajectories for AKI patients in ICU.
Speaker:
Guilan Kong, PhD
National Institute of Health Data Science, Peking University
Authors:
Haoran Su, MS - Peking University; Tongyue Shi, MS - Peking University; Guilan Kong, PhD - National Institute of Health Data Science, Peking University;
Guilan
Kong,
PhD - National Institute of Health Data Science, Peking University
Click, Calculate, Control: EHR-Integrated Insulin Calculator Boosts Glycemic Management in Cardiac Surgery ICU
Poster Number: P105
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Critical Care, Informatics Implementation, Patient Safety, Workflow, Evaluation
Primary Track: Applications
Programmatic Theme: Clinical Informatics
This study assessed the impact of an integrated intravenous insulin calculator within the hospital’s EHR system on glycemic control in cardiac surgery ICU patients. A pre- and post-implementation analysis (Sep-Dec 2023 vs. Sep-Dec 2024) showed a slight decrease in average glucose levels, improved blood glucose control, and a reduction in severe hypoglycemia and hyperglycemia. Further evaluations will explore additional factors influencing glucose regulation.
Speaker:
Shayma Alzaidi, PharnD
Brigham and Women's Hospital
Authors:
Diane Seger, RPh - Mass General Brigham; Andrew Hwang, PharmD - Massachusetts College of Pharmacy and Health Sciences; David Bates, MD - Mass General Brigham; Harvard University;
Poster Number: P105
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Critical Care, Informatics Implementation, Patient Safety, Workflow, Evaluation
Primary Track: Applications
Programmatic Theme: Clinical Informatics
This study assessed the impact of an integrated intravenous insulin calculator within the hospital’s EHR system on glycemic control in cardiac surgery ICU patients. A pre- and post-implementation analysis (Sep-Dec 2023 vs. Sep-Dec 2024) showed a slight decrease in average glucose levels, improved blood glucose control, and a reduction in severe hypoglycemia and hyperglycemia. Further evaluations will explore additional factors influencing glucose regulation.
Speaker:
Shayma Alzaidi, PharnD
Brigham and Women's Hospital
Authors:
Diane Seger, RPh - Mass General Brigham; Andrew Hwang, PharmD - Massachusetts College of Pharmacy and Health Sciences; David Bates, MD - Mass General Brigham; Harvard University;
Shayma
Alzaidi,
PharnD - Brigham and Women's Hospital
Ventriculitis Computable Phenotyping Algorithm: Pilot Study
Poster Number: P106
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Critical Care, Interoperability and Health Information Exchange, Informatics Implementation
Primary Track: Applications
Programmatic Theme: Clinical Informatics
Early detection and treatment of healthcare-associated ventriculitis can improve patient outcomes. However, due to the adoption of antibiotic coated catheters, the most commonly used criteria for ventriculitis definition, positive cerebrospinal fluid culture, would have low sensitivity. Diagnosis of ventriculitis is thus solely clinical, creating challenges for consistent cohort generation and multi-center studies. Therefore, we developed a computable phenotyping algorithm to efficiently and accurately label ventriculitis-positive patients from retrospective data, achieving 0.70 precision and 0.87 sensitivity.
Speaker:
Yanwei Li, MA
Columbia University
Authors:
Soojin Park, MD - Columbia University Medical Center; Murad Megjhani, PhD - Columbia University Medical Center; Tammam Alalqum, BS - Columbia University Medical Center; Giselle Grassi, BA - Columbia University Medical Center; Zachary Levin, BS - Columbia University Medical Center; Shalmali Joshi, PhD - Columbia University; Jiayu Yao, PhD - Columbia University;
Poster Number: P106
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Critical Care, Interoperability and Health Information Exchange, Informatics Implementation
Primary Track: Applications
Programmatic Theme: Clinical Informatics
Early detection and treatment of healthcare-associated ventriculitis can improve patient outcomes. However, due to the adoption of antibiotic coated catheters, the most commonly used criteria for ventriculitis definition, positive cerebrospinal fluid culture, would have low sensitivity. Diagnosis of ventriculitis is thus solely clinical, creating challenges for consistent cohort generation and multi-center studies. Therefore, we developed a computable phenotyping algorithm to efficiently and accurately label ventriculitis-positive patients from retrospective data, achieving 0.70 precision and 0.87 sensitivity.
Speaker:
Yanwei Li, MA
Columbia University
Authors:
Soojin Park, MD - Columbia University Medical Center; Murad Megjhani, PhD - Columbia University Medical Center; Tammam Alalqum, BS - Columbia University Medical Center; Giselle Grassi, BA - Columbia University Medical Center; Zachary Levin, BS - Columbia University Medical Center; Shalmali Joshi, PhD - Columbia University; Jiayu Yao, PhD - Columbia University;
Yanwei
Li,
MA - Columbia University
Semantic Supercharging for VTE Detection: Smart Phenptyping with SemMedDB and Real-Time NLP
Poster Number: P107
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Controlled Terminologies, Ontologies, and Vocabularies, Information Extraction, Knowledge Representation and Information Modeling
Primary Track: Applications
Programmatic Theme: Clinical Informatics
Venous thromboembolism (VTE) is a preventable hospital-acquired condition, but identifying at-risk patients in EHRs is challenging due to unstructured data. Vanderbilt University Medical Center developed an NLP pipeline knowledge base using SemMedDB to expand VTE-related concepts, improving phenotyping accuracy. Validated on 200 records, the method achieved 97.9% specificity and 91.3% sensitivity. Semantic expansion increased case identification by 46%, significantly broadening patient cohort capture
Speaker:
Sina Madani, MD, PhD,FAMIA
Vanderbilt University Medical Center
Author:
Asli Ozdas Weitkamp, PhD - Vanderbilt University Medical Center;
Poster Number: P107
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Controlled Terminologies, Ontologies, and Vocabularies, Information Extraction, Knowledge Representation and Information Modeling
Primary Track: Applications
Programmatic Theme: Clinical Informatics
Venous thromboembolism (VTE) is a preventable hospital-acquired condition, but identifying at-risk patients in EHRs is challenging due to unstructured data. Vanderbilt University Medical Center developed an NLP pipeline knowledge base using SemMedDB to expand VTE-related concepts, improving phenotyping accuracy. Validated on 200 records, the method achieved 97.9% specificity and 91.3% sensitivity. Semantic expansion increased case identification by 46%, significantly broadening patient cohort capture
Speaker:
Sina Madani, MD, PhD,FAMIA
Vanderbilt University Medical Center
Author:
Asli Ozdas Weitkamp, PhD - Vanderbilt University Medical Center;
Sina
Madani,
MD, PhD,FAMIA - Vanderbilt University Medical Center
Facilitating Concept Mapping Process: AI-Enhanced, Expert-Validated
Poster Number: P108
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Controlled Terminologies, Ontologies, and Vocabularies, Interoperability and Health Information Exchange, Large Language Models (LLMs)
Primary Track: Applications
Programmatic Theme: Clinical Informatics
Concept mapping supports data harmonization and enhances interoperability. We developed a web-based concept mapping tool that assists in mapping electronic health records data fields to standard terminologies using large language models (LLM). Domain experts can review and verify LLM-suggested concepts or manually select concepts. The tool enables team collaboration through progress tracking, discrepancy monitoring, and agreement assessment. It facilitates and supports the concept mapping process for multi-site harmonization and standardization.
Speaker:
Hao Fan, MBBS
Washington University School of Medicine in St Louis
Authors:
Hao Fan, MBBS - Washington University School of Medicine in St Louis; Joseph Lim, BS Candidate - Washington University in St. Louis; Rosie Mugoya, Bsn - Goldfarb School of Nursing and Washington University of St. Louis; Jennifer Thate, PhD, RN, CNE - Siena College; Amy Finnegan, PhD - Columbia University Medical Center; Sarah Rossetti, RN, PhD - Columbia University Department of Biomedical Informatics; Po-Yin Yen, PhD, RN - Washington University in St. Louis;
Poster Number: P108
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Controlled Terminologies, Ontologies, and Vocabularies, Interoperability and Health Information Exchange, Large Language Models (LLMs)
Primary Track: Applications
Programmatic Theme: Clinical Informatics
Concept mapping supports data harmonization and enhances interoperability. We developed a web-based concept mapping tool that assists in mapping electronic health records data fields to standard terminologies using large language models (LLM). Domain experts can review and verify LLM-suggested concepts or manually select concepts. The tool enables team collaboration through progress tracking, discrepancy monitoring, and agreement assessment. It facilitates and supports the concept mapping process for multi-site harmonization and standardization.
Speaker:
Hao Fan, MBBS
Washington University School of Medicine in St Louis
Authors:
Hao Fan, MBBS - Washington University School of Medicine in St Louis; Joseph Lim, BS Candidate - Washington University in St. Louis; Rosie Mugoya, Bsn - Goldfarb School of Nursing and Washington University of St. Louis; Jennifer Thate, PhD, RN, CNE - Siena College; Amy Finnegan, PhD - Columbia University Medical Center; Sarah Rossetti, RN, PhD - Columbia University Department of Biomedical Informatics; Po-Yin Yen, PhD, RN - Washington University in St. Louis;
Hao
Fan,
MBBS - Washington University School of Medicine in St Louis
Leveraging Informatics Tools to Reduce Lab Overutilization
Poster Number: P109
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Clinical Decision Support, Change Management, Healthcare Quality, Information Visualization, Laboratory Systems and Reporting
Primary Track: Applications
Programmatic Theme: Clinical Informatics
Up to two-thirds of lab tests in hospitalized patients may be unnecessary. Informatics tools can support the multimodal interventions necessary to create an effective and sustainable lab stewardship program, including education, clinical decision support (CDS), and clinician feedback. In addition to deploying multiple CDS interventions, we deployed health system-wide educational screensavers and built a customizable dashboard to monitor performance and enable service-specific feedback to support a new lab stewardship program in our health system.
Speaker:
Anoop Muniyappa, MD, MS
UCSF
Authors:
Theodore Peng, MD, MBA - UCSF; Brandon Scott, MD, MBA - UCSF; Armond Esmaili, MD - UCSF; Sajan Patel, MD - UCSF; Caitlin Richards, RD - UCSF; Andrew Auerbach, MD, MPH - UCSF; Amy Lu, MD, MPH - UCSF; Parul Bhargava, MD - UCSF;
Poster Number: P109
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Clinical Decision Support, Change Management, Healthcare Quality, Information Visualization, Laboratory Systems and Reporting
Primary Track: Applications
Programmatic Theme: Clinical Informatics
Up to two-thirds of lab tests in hospitalized patients may be unnecessary. Informatics tools can support the multimodal interventions necessary to create an effective and sustainable lab stewardship program, including education, clinical decision support (CDS), and clinician feedback. In addition to deploying multiple CDS interventions, we deployed health system-wide educational screensavers and built a customizable dashboard to monitor performance and enable service-specific feedback to support a new lab stewardship program in our health system.
Speaker:
Anoop Muniyappa, MD, MS
UCSF
Authors:
Theodore Peng, MD, MBA - UCSF; Brandon Scott, MD, MBA - UCSF; Armond Esmaili, MD - UCSF; Sajan Patel, MD - UCSF; Caitlin Richards, RD - UCSF; Andrew Auerbach, MD, MPH - UCSF; Amy Lu, MD, MPH - UCSF; Parul Bhargava, MD - UCSF;
Anoop
Muniyappa,
MD, MS - UCSF
Active Choice Clinical Decision Support Tool: A Novel Bot-Based Approach to Improving Hepatocellular Carcinoma Screening in Patients with Cirrhosis
Poster Number: P110
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Clinical Decision Support, Chronic Care Management, Informatics Implementation, Population Health, Workflow
Primary Track: Applications
Programmatic Theme: Clinical Informatics
At NYU Langone Health, we developed a novel clinical decision support bot tool to improve hepatocellular carcinoma screening in cirrhosis patients by automatically pending liver ultrasound orders for clinicians to review. A pilot study (Nov 2024–Jan 2025) with 36 patients showed that the intervention arm had a higher ultrasound order rate (67%) compared to the control arm (6%). Preliminary results suggest the tool may increase ultrasound screening rates, with ongoing trials assessing its broader impact.
Speaker:
Sarah Tsuruo, BA
NYU Langone Health
Authors:
Tamara Brodsky, MD, MBA - NYU Langone Health; Feng Su, MD - NYU Langone Health; Adam Szerencsy, DO - NYU Langone Health; Steven Lim; Dinesha Prabhu, NA - NYU Langone Health; Vigneshwaran Velayudham, NA - NYU Langone Health; Ajay Mansukhani, NA - NYU Langone Health; Nathalia Ladino, MS - NYU Langone Health; William King, MS Biostatistics; Leora Horwitz, MD, MHS - NYU Langone Health; Arielle Nagler, MD - NYU Langone Health; Amrita Mukhopadhyay, MD - NYU Langone Health; Saul Blecker, MD - NYU School of Medicine, Population Health;
Poster Number: P110
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Clinical Decision Support, Chronic Care Management, Informatics Implementation, Population Health, Workflow
Primary Track: Applications
Programmatic Theme: Clinical Informatics
At NYU Langone Health, we developed a novel clinical decision support bot tool to improve hepatocellular carcinoma screening in cirrhosis patients by automatically pending liver ultrasound orders for clinicians to review. A pilot study (Nov 2024–Jan 2025) with 36 patients showed that the intervention arm had a higher ultrasound order rate (67%) compared to the control arm (6%). Preliminary results suggest the tool may increase ultrasound screening rates, with ongoing trials assessing its broader impact.
Speaker:
Sarah Tsuruo, BA
NYU Langone Health
Authors:
Tamara Brodsky, MD, MBA - NYU Langone Health; Feng Su, MD - NYU Langone Health; Adam Szerencsy, DO - NYU Langone Health; Steven Lim; Dinesha Prabhu, NA - NYU Langone Health; Vigneshwaran Velayudham, NA - NYU Langone Health; Ajay Mansukhani, NA - NYU Langone Health; Nathalia Ladino, MS - NYU Langone Health; William King, MS Biostatistics; Leora Horwitz, MD, MHS - NYU Langone Health; Arielle Nagler, MD - NYU Langone Health; Amrita Mukhopadhyay, MD - NYU Langone Health; Saul Blecker, MD - NYU School of Medicine, Population Health;
Sarah
Tsuruo,
BA - NYU Langone Health
Decision support for low dose aspirin recommendation in high-risk pregnancies
Poster Number: P111
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Clinical Decision Support, Clinical Guidelines, Education and Training
Primary Track: Applications
Programmatic Theme: Clinical Informatics
Low dose aspirin use decreases the risk of preeclampsia in pregnant patients with risk factors. The study team hypothesized that the use of a clinical decision support alert in the electronic health record would increase obstetric clinician recommendation for low dose aspirin use in the appropriate patient population. This was evaluated prospectively via a randomized controlled trial. Preliminary findings suggest that a behavioral science-informed decision support alert increases clinician recommendation aiming to improve patient outcomes.
Speaker:
Maranda Sullivan, DO
Geisinger
Authors:
Maranda Sullivan, DO - Geisinger; Amir Goren, PhD - Geisinger; Jignaben Chaudhari, DO - Geisinger; Kyle Marshall, MD, MBI, FACEP, FAAEM, FAMIA - Geisinger; Celia Gray, MS - Geisinger; Henri Santos, PhD - Geisinger; Christopher Chabris, PhD - Geisinger; Michelle Meyer, PhD - Geisinger; A. Dhanya Mackeen, MD, MPH - Geisinger;
Poster Number: P111
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Clinical Decision Support, Clinical Guidelines, Education and Training
Primary Track: Applications
Programmatic Theme: Clinical Informatics
Low dose aspirin use decreases the risk of preeclampsia in pregnant patients with risk factors. The study team hypothesized that the use of a clinical decision support alert in the electronic health record would increase obstetric clinician recommendation for low dose aspirin use in the appropriate patient population. This was evaluated prospectively via a randomized controlled trial. Preliminary findings suggest that a behavioral science-informed decision support alert increases clinician recommendation aiming to improve patient outcomes.
Speaker:
Maranda Sullivan, DO
Geisinger
Authors:
Maranda Sullivan, DO - Geisinger; Amir Goren, PhD - Geisinger; Jignaben Chaudhari, DO - Geisinger; Kyle Marshall, MD, MBI, FACEP, FAAEM, FAMIA - Geisinger; Celia Gray, MS - Geisinger; Henri Santos, PhD - Geisinger; Christopher Chabris, PhD - Geisinger; Michelle Meyer, PhD - Geisinger; A. Dhanya Mackeen, MD, MPH - Geisinger;
Maranda
Sullivan,
DO - Geisinger
Novel CDS Tools to Improve Pediatric Lead Follow-Up in Primary Care
Poster Number: P112
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Clinical Decision Support, Clinical Guidelines, Pediatrics, Population Health, Healthcare Quality
Primary Track: Applications
Programmatic Theme: Clinical Informatics
Childhood lead exposure harms growth and development. Our institution identified a need for electronic health record-based tools to support follow-up for elevated blood lead levels. We developed a standardized pathway incorporating regional guidelines and implemented novel clinical decision support tools to automate orders and documentation using rule-based logic. Additionally, we implemented a non-interruptive, documentation-based nudge during telephone encounters to identify overdue patients. Early nudge data are promising, and real-time dashboards enable continuous monitoring and improvement.
Speaker:
Peter Zhang, MD, MS
Children's Hospital of Philadelphia
Authors:
Jeremy Michel, MD, MHS - The Children's Hospital of Philadelphia, Center for Biomedical Informatics; Lauren Coogle, MD - Children's Hospital of Philadelphia;
Poster Number: P112
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Clinical Decision Support, Clinical Guidelines, Pediatrics, Population Health, Healthcare Quality
Primary Track: Applications
Programmatic Theme: Clinical Informatics
Childhood lead exposure harms growth and development. Our institution identified a need for electronic health record-based tools to support follow-up for elevated blood lead levels. We developed a standardized pathway incorporating regional guidelines and implemented novel clinical decision support tools to automate orders and documentation using rule-based logic. Additionally, we implemented a non-interruptive, documentation-based nudge during telephone encounters to identify overdue patients. Early nudge data are promising, and real-time dashboards enable continuous monitoring and improvement.
Speaker:
Peter Zhang, MD, MS
Children's Hospital of Philadelphia
Authors:
Jeremy Michel, MD, MHS - The Children's Hospital of Philadelphia, Center for Biomedical Informatics; Lauren Coogle, MD - Children's Hospital of Philadelphia;
Peter
Zhang,
MD, MS - Children's Hospital of Philadelphia
Nudging Away Bad Clinical Decision Support
Poster Number: P113
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Clinical Decision Support, Clinical Guidelines, User-centered Design Methods
Primary Track: Applications
Programmatic Theme: Clinical Informatics
A ‘nudge’ encourages certain courses of action without taking away the freedom of choice. Using nudge-type strategies to replace ineffective, interruptive, and frustrating reminders, we aimed to improve adherence to guideline liver re-imaging intervals for patients with cystic fibrosis hepatobiliary involvement (CFHBI) while enhancing user experience. Our interventions improved biennial ultrasound adherence from 53.6% to 62.5%, annual Fibroscans from 53.6% to 90.9%, and user perception of effectiveness, usability, and efficiency.
Speaker:
Derek Ngai, MD
UT Southwestern
Author:
Philip Bernard, M.D. - Children's Health of Texas;
Poster Number: P113
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Clinical Decision Support, Clinical Guidelines, User-centered Design Methods
Primary Track: Applications
Programmatic Theme: Clinical Informatics
A ‘nudge’ encourages certain courses of action without taking away the freedom of choice. Using nudge-type strategies to replace ineffective, interruptive, and frustrating reminders, we aimed to improve adherence to guideline liver re-imaging intervals for patients with cystic fibrosis hepatobiliary involvement (CFHBI) while enhancing user experience. Our interventions improved biennial ultrasound adherence from 53.6% to 62.5%, annual Fibroscans from 53.6% to 90.9%, and user perception of effectiveness, usability, and efficiency.
Speaker:
Derek Ngai, MD
UT Southwestern
Author:
Philip Bernard, M.D. - Children's Health of Texas;
Derek
Ngai,
MD - UT Southwestern
SPARK-3: A Real-Time Machine Learning Approach to Early Sepsis Detection at Emory Healthcare
Poster Number: P114
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Clinical Decision Support, Critical Care, Machine Learning, Evaluation, Artificial Intelligence, Healthcare Quality
Primary Track: Applications
Programmatic Theme: Clinical Informatics
The SPARK-3, developed by Emory, and Epic Sepsis Model (ESM) 2.5 were evaluated for early sepsis detection at Emory Healthcare using retrospective data from 96,000 hospitalized patients, including 5,100 who developed sepsis. Both models exhibited similar predictive patterns, but SPARK-3 showed a higher threshold score. SPARK-3's higher threshold offers an actionable approach, with potential for earlier intervention, highlighting the need for further refinement in clinical adoption to optimize sepsis management.
Speaker:
Megan Schwinne, MPH
Emory University
Authors:
Megan Schwinne, MPH - Emory University; Barney Chan, MS CIS - Emory University; Ryan Birmingham, MS - Emory University; Alasdair Gent, PhD - Duke University; Matthew Pagel, BS - Emory University; Chad Robichaux, MPH - Emory University; Dileep Gunda, MS - Emory University; Sheida Habibi, MS - Emory University; Sivasubramanium Bhavani, MD - Emory University; Raymund Dantes, MD, MPH - Emory University; Reza Sameni, PhD - Emory Unviersity; Timothy Buchman, MD, PhD - Emory University; Tony Pan, PhD - Emory University; Rishikesan Kamaleswaran, PhD - Duke University;
Poster Number: P114
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Clinical Decision Support, Critical Care, Machine Learning, Evaluation, Artificial Intelligence, Healthcare Quality
Primary Track: Applications
Programmatic Theme: Clinical Informatics
The SPARK-3, developed by Emory, and Epic Sepsis Model (ESM) 2.5 were evaluated for early sepsis detection at Emory Healthcare using retrospective data from 96,000 hospitalized patients, including 5,100 who developed sepsis. Both models exhibited similar predictive patterns, but SPARK-3 showed a higher threshold score. SPARK-3's higher threshold offers an actionable approach, with potential for earlier intervention, highlighting the need for further refinement in clinical adoption to optimize sepsis management.
Speaker:
Megan Schwinne, MPH
Emory University
Authors:
Megan Schwinne, MPH - Emory University; Barney Chan, MS CIS - Emory University; Ryan Birmingham, MS - Emory University; Alasdair Gent, PhD - Duke University; Matthew Pagel, BS - Emory University; Chad Robichaux, MPH - Emory University; Dileep Gunda, MS - Emory University; Sheida Habibi, MS - Emory University; Sivasubramanium Bhavani, MD - Emory University; Raymund Dantes, MD, MPH - Emory University; Reza Sameni, PhD - Emory Unviersity; Timothy Buchman, MD, PhD - Emory University; Tony Pan, PhD - Emory University; Rishikesan Kamaleswaran, PhD - Duke University;
Megan
Schwinne,
MPH - Emory University
Privacy-Preserving, Asynchronous, Federated Training and Validation of a Clinical Decision Support System for Ventilator Management
Poster Number: P115
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Clinical Decision Support, Critical Care, Machine Learning, Artificial Intelligence, Privacy and Security, Informatics Implementation, Evaluation, Data Sharing
Primary Track: Applications
Programmatic Theme: Clinical Informatics
Managing patients receiving invasive mechanical ventilation settings is complex and no effective clinical decision support systems exist that are generalizable and comprehensive to support such management. We trained the AI Vent Assistant (AVA) using an asynchronous, privacy-preserving, federated approach across seven geographically diverse health systems to make simultaneous recommendations for twelve interrelated ventilator settings. The resultant, aggregated model required no data sharing or additional training and outperformed most local models in held-out, external validation.
Speaker:
Antonia Angeli Gazola, MD
University of Pennsylvania
Authors:
Antonia Angeli Gazola, MD - University of Pennsylvania; Benjamin Schmid, MS - University of Pennsylvania; Nicholas Bishop, BA - University of Pennsylvania; Alexander Ortiz, MS, MS - University of Pennsylvania; Nicholas Ingraham, MD, MS - University of Minnesota Medical School; Patrick Lyons, MD, MSc - Oregon Health & Science University; Brenna Park-Egan, M.S. - Oregon Health & Science University; Kaveri Chhikara, Senior Data Scientist - University of Chicago; Catherine Gao, MD - Northwestern; Wang-Ting Liao, MS - Northwestern University; Anna Barker, MD PhD - University of Michigan; Juan Rojas, MD MS - Rush University; Vaishvik Chaudhari, Masters in Data Science - Rush University Medical College; GARY WEISSMAN, MD, MSHP - University of Pennsylvania;
Poster Number: P115
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Clinical Decision Support, Critical Care, Machine Learning, Artificial Intelligence, Privacy and Security, Informatics Implementation, Evaluation, Data Sharing
Primary Track: Applications
Programmatic Theme: Clinical Informatics
Managing patients receiving invasive mechanical ventilation settings is complex and no effective clinical decision support systems exist that are generalizable and comprehensive to support such management. We trained the AI Vent Assistant (AVA) using an asynchronous, privacy-preserving, federated approach across seven geographically diverse health systems to make simultaneous recommendations for twelve interrelated ventilator settings. The resultant, aggregated model required no data sharing or additional training and outperformed most local models in held-out, external validation.
Speaker:
Antonia Angeli Gazola, MD
University of Pennsylvania
Authors:
Antonia Angeli Gazola, MD - University of Pennsylvania; Benjamin Schmid, MS - University of Pennsylvania; Nicholas Bishop, BA - University of Pennsylvania; Alexander Ortiz, MS, MS - University of Pennsylvania; Nicholas Ingraham, MD, MS - University of Minnesota Medical School; Patrick Lyons, MD, MSc - Oregon Health & Science University; Brenna Park-Egan, M.S. - Oregon Health & Science University; Kaveri Chhikara, Senior Data Scientist - University of Chicago; Catherine Gao, MD - Northwestern; Wang-Ting Liao, MS - Northwestern University; Anna Barker, MD PhD - University of Michigan; Juan Rojas, MD MS - Rush University; Vaishvik Chaudhari, Masters in Data Science - Rush University Medical College; GARY WEISSMAN, MD, MSHP - University of Pennsylvania;
Antonia
Angeli Gazola,
MD - University of Pennsylvania
Bundle Compliance: Challenges of Transition from Paper to Electronic
Poster Number: P116
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Clinical Decision Support, Critical Care, Pediatrics, Informatics Implementation
Primary Track: Foundations
Programmatic Theme: Clinical Informatics
The transition of care bundles from paper to electronic health records (EHRs) aims to improve compliance and patient outcomes as part of quality improvement initiatives in healthcare settings. While electronic integration can improve documentation accessibility, there is limits to its effectiveness without staff engagement to the adoption of new practices.
Speaker:
Brittany Brennan, MSN, PNP-AC
CHOA
Authors:
Sarah Thompson, MSHIMI, BSN, RN - Children's Healthcare of Atlanta; Brittany Brennan, MSN, PNP-AC - CHOA;
Poster Number: P116
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Clinical Decision Support, Critical Care, Pediatrics, Informatics Implementation
Primary Track: Foundations
Programmatic Theme: Clinical Informatics
The transition of care bundles from paper to electronic health records (EHRs) aims to improve compliance and patient outcomes as part of quality improvement initiatives in healthcare settings. While electronic integration can improve documentation accessibility, there is limits to its effectiveness without staff engagement to the adoption of new practices.
Speaker:
Brittany Brennan, MSN, PNP-AC
CHOA
Authors:
Sarah Thompson, MSHIMI, BSN, RN - Children's Healthcare of Atlanta; Brittany Brennan, MSN, PNP-AC - CHOA;
Brittany
Brennan,
MSN, PNP-AC - CHOA
Time-Sensitive Prediction of 30-Day Hospital Readmissions Using the MIMIC-IV Dataset
Poster Number: P117
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Clinical Decision Support, Data Mining, Machine Learning, Chronic Care Management
Primary Track: Applications
Programmatic Theme: Clinical Informatics
Our study develops and validates a machine learning model to predict 30-day hospital readmissions by utilizing lab test results from multiple timeframes and integrating both structured and unstructured data sources, such as radiology reports and discharge notes. Employing various predictive models, the research demonstrates improved accuracy, particularly when historical lab data is included, highlighting the value of comprehensive data in enhancing predictive performance for hospital readmissions.
Speaker:
MANAL ALHUSSEIN, PhD Student
George Mason University
Authors:
MANAL ALHUSSEIN, PhD Student - George Mason University; Janusz Wojtusiak, PhD - George Mason University;
Poster Number: P117
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Clinical Decision Support, Data Mining, Machine Learning, Chronic Care Management
Primary Track: Applications
Programmatic Theme: Clinical Informatics
Our study develops and validates a machine learning model to predict 30-day hospital readmissions by utilizing lab test results from multiple timeframes and integrating both structured and unstructured data sources, such as radiology reports and discharge notes. Employing various predictive models, the research demonstrates improved accuracy, particularly when historical lab data is included, highlighting the value of comprehensive data in enhancing predictive performance for hospital readmissions.
Speaker:
MANAL ALHUSSEIN, PhD Student
George Mason University
Authors:
MANAL ALHUSSEIN, PhD Student - George Mason University; Janusz Wojtusiak, PhD - George Mason University;
MANAL
ALHUSSEIN,
PhD Student - George Mason University
Missed and Misclassified: Evaluating Emergent Transfer Identification Across Data Sources
Poster Number: P118
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Clinical Decision Support, Data Mining, Evaluation, Information Extraction, Information Retrieval, Pediatrics
Primary Track: Applications
Programmatic Theme: Clinical Informatics
This study compares emergent transfer (ET) identification methods using PAC3/PC4 registry data and Epic Clarity EHR data for pediatric cardiology patients. Of 90 ETs identified by Clarity and 84 by the registry, 66 were identified in both methods, Clarity identified 23 not identified in the registry, and the registry identified 4 not identified in Clarity. Both methods misidentified ETs that actually did not meet criteria: 1 in Clarity and 14 in the registry.
Speaker:
David Kulp, MSc
Emory University School of Medicine
Authors:
Zachary West, MD - Children's Healthcare of Atlanta; Naveen Muthu, MD - Children's Healthcare of Atlanta;
Poster Number: P118
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Clinical Decision Support, Data Mining, Evaluation, Information Extraction, Information Retrieval, Pediatrics
Primary Track: Applications
Programmatic Theme: Clinical Informatics
This study compares emergent transfer (ET) identification methods using PAC3/PC4 registry data and Epic Clarity EHR data for pediatric cardiology patients. Of 90 ETs identified by Clarity and 84 by the registry, 66 were identified in both methods, Clarity identified 23 not identified in the registry, and the registry identified 4 not identified in Clarity. Both methods misidentified ETs that actually did not meet criteria: 1 in Clarity and 14 in the registry.
Speaker:
David Kulp, MSc
Emory University School of Medicine
Authors:
Zachary West, MD - Children's Healthcare of Atlanta; Naveen Muthu, MD - Children's Healthcare of Atlanta;
David
Kulp,
MSc - Emory University School of Medicine
Development of a Mobile Contraception Decision Aid for Transgender and Gender-Nonconforming Individuals Assigned Female at Birth
Poster Number: P119
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Clinical Decision Support, Diversity, Equity, Inclusion, and Accessibility, Patient Engagement and Preferences
Primary Track: Applications
Programmatic Theme: Clinical Informatics
We report a study to develop a mobile contraception decision aid for transgender and gender-nonconforming (TGNC) assigned female at birth (AFAB) individuals to address their unique challenges. Utilizing design principles, ethnographic data, and pilot tool feedback, we developed a working mobile prototype and scalable desktop version. The new version has additional community-relevant information and features, which is informing an ongoing study on the usability, trust, and decision-making factors in contraceptive decision aids for this community.
Speaker:
Isabel Glass, MA
Arizona State University
Authors:
I-Wen Weng, BA - Arizona State University; Pei-Yu Tsai, BA - Arizona State University; Rushabh Jaiswal, BT - Arizona State University; Kanishk Tanotra, MS - Arizona State University; Erin Chiou, PhD - Arizona State University; Jenny Brian, PhD - Arizona State University; Dongwen Wang, PhD - Arizona State University;
Poster Number: P119
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Clinical Decision Support, Diversity, Equity, Inclusion, and Accessibility, Patient Engagement and Preferences
Primary Track: Applications
Programmatic Theme: Clinical Informatics
We report a study to develop a mobile contraception decision aid for transgender and gender-nonconforming (TGNC) assigned female at birth (AFAB) individuals to address their unique challenges. Utilizing design principles, ethnographic data, and pilot tool feedback, we developed a working mobile prototype and scalable desktop version. The new version has additional community-relevant information and features, which is informing an ongoing study on the usability, trust, and decision-making factors in contraceptive decision aids for this community.
Speaker:
Isabel Glass, MA
Arizona State University
Authors:
I-Wen Weng, BA - Arizona State University; Pei-Yu Tsai, BA - Arizona State University; Rushabh Jaiswal, BT - Arizona State University; Kanishk Tanotra, MS - Arizona State University; Erin Chiou, PhD - Arizona State University; Jenny Brian, PhD - Arizona State University; Dongwen Wang, PhD - Arizona State University;
Isabel
Glass,
MA - Arizona State University
Evaluating CDS Alert Overrides: Patterns, Influences, and Implications for Drug Allergy Alerts
Poster Number: P120
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Clinical Decision Support, Documentation Burden, Evaluation
Primary Track: Applications
Programmatic Theme: Clinical Informatics
This study analyzed 224,066 drug allergy alerts (DAAs) from 9,545 patients at UAB Hospital in 2023, revealing a 96.5% override rate. Overrides varied by drug class, reaction type, and provider role, with residents overriding the most. While gender differences were statistically significant, clinical relevance was limited. Findings highlight the need to refine CDS design and investigate additional factors to reduce unnecessary alerts and improve decision-making effectiveness.
Speaker:
Jakir Hossain Bhuiyan Masud, PhD
University of Alabama at Birmingham
Author:
Tiago Colicchio, PhD., MBA - University of Alabama at Birmingham;
Poster Number: P120
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Clinical Decision Support, Documentation Burden, Evaluation
Primary Track: Applications
Programmatic Theme: Clinical Informatics
This study analyzed 224,066 drug allergy alerts (DAAs) from 9,545 patients at UAB Hospital in 2023, revealing a 96.5% override rate. Overrides varied by drug class, reaction type, and provider role, with residents overriding the most. While gender differences were statistically significant, clinical relevance was limited. Findings highlight the need to refine CDS design and investigate additional factors to reduce unnecessary alerts and improve decision-making effectiveness.
Speaker:
Jakir Hossain Bhuiyan Masud, PhD
University of Alabama at Birmingham
Author:
Tiago Colicchio, PhD., MBA - University of Alabama at Birmingham;
Jakir Hossain Bhuiyan
Masud,
PhD - University of Alabama at Birmingham
Machine Learning-based Clinical Decision Support (ML-CDS) in the Emergency Department: A Toolkit for Effective Implementation
Poster Number: P121
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Clinical Decision Support, Informatics Implementation, Machine Learning, Artificial Intelligence
Primary Track: Applications
Programmatic Theme: Clinical Informatics
In this abstract, we present an implementation toolkit developed to support the effective implementation of machine learning-based clinical decision support (ML-CDS) tools in emergency departments (EDs). Grounded in Implementation Science and Human Factors Engineering, the toolkit was developed through iterative, interdisciplinary design and testing across three ED sites when implementing an ML-CDS to prevent future falls. It includes methods for stakeholder engagement, workflow mapping, model validation, and ongoing monitoring to ensure successful implementation of ML-CDS.
Speaker:
Hanna Barton, PhD
University of Wisconsin-Madison
Authors:
Hanna Barton, PhD - University of Wisconsin-Madison; Erkin Ötleş, MD, PhD - University of Wisconsin-Madison; Apoorva Maru, BS - University of Wisconsin-Madison; Margaret Leaf, MS - UW Health; Daniel Hekman, MS; Douglas Wiegmann, PhD - University of Wisconsin-Madison; Brian Patterson, MD MPH - University of Wisconsin-Madison;
Poster Number: P121
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Clinical Decision Support, Informatics Implementation, Machine Learning, Artificial Intelligence
Primary Track: Applications
Programmatic Theme: Clinical Informatics
In this abstract, we present an implementation toolkit developed to support the effective implementation of machine learning-based clinical decision support (ML-CDS) tools in emergency departments (EDs). Grounded in Implementation Science and Human Factors Engineering, the toolkit was developed through iterative, interdisciplinary design and testing across three ED sites when implementing an ML-CDS to prevent future falls. It includes methods for stakeholder engagement, workflow mapping, model validation, and ongoing monitoring to ensure successful implementation of ML-CDS.
Speaker:
Hanna Barton, PhD
University of Wisconsin-Madison
Authors:
Hanna Barton, PhD - University of Wisconsin-Madison; Erkin Ötleş, MD, PhD - University of Wisconsin-Madison; Apoorva Maru, BS - University of Wisconsin-Madison; Margaret Leaf, MS - UW Health; Daniel Hekman, MS; Douglas Wiegmann, PhD - University of Wisconsin-Madison; Brian Patterson, MD MPH - University of Wisconsin-Madison;
Hanna
Barton,
PhD - University of Wisconsin-Madison
Bilingual Access to a Patient Portal Intervention: A Multisite Feasibility Study of My Diabetes Care in English and Spanish
Poster Number: P122
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Chronic Care Management, Human-computer Interaction, Usability, Patient Engagement and Preferences
Primary Track: Applications
Programmatic Theme: Consumer Health Informatics
Background: My Diabetes Care (MDC) is a patient portal-integrated intervention for diabetes self-management. Using Design Sprint methodology, we enhanced MDC by expanding its user interface to display additional diabetes data and developing a Spanish-language version. This study examined the feasibility (engagement) and acceptability (usability and perceived effectiveness) of the enhanced MDC.
Methods: In a pre-post, single-group study, participants with type 2 diabetes from two academic medical centers accessed MDC for one month. Feasibility was assessed through engagement, while acceptability was evaluated via usability (System Usability Scale, SUS), perceived effectiveness (changes in diabetes knowledge, self-care activities, medication adherence, diabetes distress), and qualitative interviews (n=9).
Results: Among 63 participants (mean age 57.2 years), 78% used MDC, with 90% using the English and 10% the Spanish version. The median SUS score was 75.0, indicating “good” usability. Scores were slightly lower for participants with limited health literacy (median 71.2 vs. 75.0, p=0.12), those aged 65+ (median 72.5 vs. 78.8, p=0.05), and Spanish-speaking users (median 57.5 vs. 75.0, p=0.19). Post-intervention, diabetes self-care activities improved, notably in diet adherence (median 4.0 pre vs. 4.5 post, p=0.07), though not statistically significant. Participants valued consolidated data display, color coding, and educational resources but some, particularly Spanish speakers, reported navigation challenges.
Conclusions: MDC demonstrated high usability but requires targeted support for Spanish-speaking and older patients. Future research should focus on long-term outcomes, clinical integration, and improving usability for diverse populations.
Speaker:
William Martinez, MD, MS
Vanderbilt University Medical Center
Authors:
Jorge Rodriguez, MD - Brigham and Women's Hospital/Harvard Medical School; Lipika Samal, MD - Brigham and Women's Hospital; Tom Elasy, MD, MPH - Vanderbilt University Medical Center; Isaac Gorgy, MD - University of Southern California/Los Angeles General Medical Center; Amber Hackstadt, PhD - Vanderbilt University Medical Center; Lindsay Mayberry, PhD - Vanderbilt University Medical Center; Lyndsay Nelson, PhD - Vanderbilt University Medical Center; Audriana Audriana, BA - Vanderbilt University Medical Center; S. Trent Rosenbloom, MD, MPH, FACMI, FAMIA - Vanderbilt University Medical Center Dept of Biomedical Informatics; Adam Wright, PhD - Vanderbilt University Medical Center; Zhihong Yu, PhD - Vanderbilt University Medical Center; William Martinez, MD, MS - Vanderbilt University Medical Center;
Poster Number: P122
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Chronic Care Management, Human-computer Interaction, Usability, Patient Engagement and Preferences
Primary Track: Applications
Programmatic Theme: Consumer Health Informatics
Background: My Diabetes Care (MDC) is a patient portal-integrated intervention for diabetes self-management. Using Design Sprint methodology, we enhanced MDC by expanding its user interface to display additional diabetes data and developing a Spanish-language version. This study examined the feasibility (engagement) and acceptability (usability and perceived effectiveness) of the enhanced MDC.
Methods: In a pre-post, single-group study, participants with type 2 diabetes from two academic medical centers accessed MDC for one month. Feasibility was assessed through engagement, while acceptability was evaluated via usability (System Usability Scale, SUS), perceived effectiveness (changes in diabetes knowledge, self-care activities, medication adherence, diabetes distress), and qualitative interviews (n=9).
Results: Among 63 participants (mean age 57.2 years), 78% used MDC, with 90% using the English and 10% the Spanish version. The median SUS score was 75.0, indicating “good” usability. Scores were slightly lower for participants with limited health literacy (median 71.2 vs. 75.0, p=0.12), those aged 65+ (median 72.5 vs. 78.8, p=0.05), and Spanish-speaking users (median 57.5 vs. 75.0, p=0.19). Post-intervention, diabetes self-care activities improved, notably in diet adherence (median 4.0 pre vs. 4.5 post, p=0.07), though not statistically significant. Participants valued consolidated data display, color coding, and educational resources but some, particularly Spanish speakers, reported navigation challenges.
Conclusions: MDC demonstrated high usability but requires targeted support for Spanish-speaking and older patients. Future research should focus on long-term outcomes, clinical integration, and improving usability for diverse populations.
Speaker:
William Martinez, MD, MS
Vanderbilt University Medical Center
Authors:
Jorge Rodriguez, MD - Brigham and Women's Hospital/Harvard Medical School; Lipika Samal, MD - Brigham and Women's Hospital; Tom Elasy, MD, MPH - Vanderbilt University Medical Center; Isaac Gorgy, MD - University of Southern California/Los Angeles General Medical Center; Amber Hackstadt, PhD - Vanderbilt University Medical Center; Lindsay Mayberry, PhD - Vanderbilt University Medical Center; Lyndsay Nelson, PhD - Vanderbilt University Medical Center; Audriana Audriana, BA - Vanderbilt University Medical Center; S. Trent Rosenbloom, MD, MPH, FACMI, FAMIA - Vanderbilt University Medical Center Dept of Biomedical Informatics; Adam Wright, PhD - Vanderbilt University Medical Center; Zhihong Yu, PhD - Vanderbilt University Medical Center; William Martinez, MD, MS - Vanderbilt University Medical Center;
William
Martinez,
MD, MS - Vanderbilt University Medical Center
Exploring Blood Pressure Trajectories: A Focus on Essential Hypotension
Poster Number: P123
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Chronic Care Management, Information Visualization, Real-World Evidence Generation
Primary Track: Applications
This study analyzes blood pressure trajectories in essential hypotension using data from the ESSENTIAL registry. Ambulatory blood pressure monitoring identified significantly more masked hypotension than clinic-based methods. While transition rates remained stable, half of untreated hypotensive patients remained hypotensive at follow-up. Adherence declined sharply, with 72.84% lost by the fifth visit. These findings highlight the need for improved long-term monitoring and personalized care strategies to enhance essential hypotension management and patient outcomes.
Speaker:
Febin Aby Simon, Master's in health informatics
Weill Cornell Medicine
Authors:
Haoxin Chen, MS - Weill Cornell Medical College; Ke Yu, MS in Health Informatics - Weill Cornell Medicine; Febin Aby Simon, Master's in health informatics - Weill Cornell Medicine; Jose Florez-Arango, MD MS PhD - Weill Cornell Medicine; Luis Eduardo Medina, MD - CES Clinic (Clínica CES);
Poster Number: P123
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Chronic Care Management, Information Visualization, Real-World Evidence Generation
Primary Track: Applications
This study analyzes blood pressure trajectories in essential hypotension using data from the ESSENTIAL registry. Ambulatory blood pressure monitoring identified significantly more masked hypotension than clinic-based methods. While transition rates remained stable, half of untreated hypotensive patients remained hypotensive at follow-up. Adherence declined sharply, with 72.84% lost by the fifth visit. These findings highlight the need for improved long-term monitoring and personalized care strategies to enhance essential hypotension management and patient outcomes.
Speaker:
Febin Aby Simon, Master's in health informatics
Weill Cornell Medicine
Authors:
Haoxin Chen, MS - Weill Cornell Medical College; Ke Yu, MS in Health Informatics - Weill Cornell Medicine; Febin Aby Simon, Master's in health informatics - Weill Cornell Medicine; Jose Florez-Arango, MD MS PhD - Weill Cornell Medicine; Luis Eduardo Medina, MD - CES Clinic (Clínica CES);
Febin
Aby Simon,
Master's in health informatics - Weill Cornell Medicine
Identifying Co-Occurrence Patterns of Chronic Conditions in Acute Myeloid Leukemia: An EHR-based Analysis
Poster Number: P124
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Chronic Care Management, Information Visualization, Nursing Informatics, Knowledge Representation and Information Modeling
Primary Track: Applications
Programmatic Theme: Clinical Research Informatics
Acute myeloid leukemia (AML) patients often experience co-occurring chronic conditions that exacerbate symptom burden and healthcare utilization. This study analyzed electronic health record data from 446 AML patients to identify co-occurrence patterns. Hypertension (56.73%), metabolic diseases (46.86%), and colitis (31.84%) were most prevalent. Key associations included cataracts & eye diseases and pancreatic diseases & atrial fibrillation. Findings highlight the complexity of AML-related comorbidities and emphasize the need for personalized care strategies.
Speaker:
Sena Chae, PhD, RN
University of Iowa
Authors:
Jude Shelton, MS - University of Iowa College of Public Health; Alaa Harb, MSN, RN, PMHN - University of Iowa College of Nursing;
Poster Number: P124
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Chronic Care Management, Information Visualization, Nursing Informatics, Knowledge Representation and Information Modeling
Primary Track: Applications
Programmatic Theme: Clinical Research Informatics
Acute myeloid leukemia (AML) patients often experience co-occurring chronic conditions that exacerbate symptom burden and healthcare utilization. This study analyzed electronic health record data from 446 AML patients to identify co-occurrence patterns. Hypertension (56.73%), metabolic diseases (46.86%), and colitis (31.84%) were most prevalent. Key associations included cataracts & eye diseases and pancreatic diseases & atrial fibrillation. Findings highlight the complexity of AML-related comorbidities and emphasize the need for personalized care strategies.
Speaker:
Sena Chae, PhD, RN
University of Iowa
Authors:
Jude Shelton, MS - University of Iowa College of Public Health; Alaa Harb, MSN, RN, PMHN - University of Iowa College of Nursing;
Sena
Chae,
PhD, RN - University of Iowa
Development and Implementation of the Virtual Headache Hospital
Poster Number: P125
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Chronic Care Management, Population Health, Informatics Implementation, Transitions of Care, User-centered Design Methods
Primary Track: Applications
Programmatic Theme: Clinical Informatics
A quality improvement initiative to enhance the care of undiagnosed headache patients and patients whose headache severity or frequency suggested a need to escalate care was developed within an integrated health system by leveraging EHR-based tools and data. Preliminary findings from the pilot implementation demonstrate the utility of such workflows in bridging existing care gaps. Feedback from providers highlights acceptability within health systems that could improve the sustainability of such initiatives.
Speaker:
Apoorva Pradhan, BAMS, MPH
Geisinger
Authors:
Apoorva Pradhan, BAMS, MPH - Geisinger; Scott Friedenberg, MD - Geisinger; Payton Whary, BS - Geisinger; Peyton Latorre, BS, PMP - Geisinger; Adriene Zook, PharmD - Geisinger; Leonard Learn, PharmD - Geisinger; Rachel Dragano, DNP, CRNP - Geisinger; Sandra Herr, BSN - Geisinger Health Plan; Henry Aftewicz, PharmD - Geisinger; Jamie Kerestes, PharmD - Geisinger; Malory Sponenberg, MSN, RN - Geisinger; Michael Stoppie, MS - Geisinger; Kimberly Mackes, - - Geisinger; Nathaniel Stark, BS - Geisinger; Eric Wright, PharmD, MPH - Geisinger;
Poster Number: P125
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Chronic Care Management, Population Health, Informatics Implementation, Transitions of Care, User-centered Design Methods
Primary Track: Applications
Programmatic Theme: Clinical Informatics
A quality improvement initiative to enhance the care of undiagnosed headache patients and patients whose headache severity or frequency suggested a need to escalate care was developed within an integrated health system by leveraging EHR-based tools and data. Preliminary findings from the pilot implementation demonstrate the utility of such workflows in bridging existing care gaps. Feedback from providers highlights acceptability within health systems that could improve the sustainability of such initiatives.
Speaker:
Apoorva Pradhan, BAMS, MPH
Geisinger
Authors:
Apoorva Pradhan, BAMS, MPH - Geisinger; Scott Friedenberg, MD - Geisinger; Payton Whary, BS - Geisinger; Peyton Latorre, BS, PMP - Geisinger; Adriene Zook, PharmD - Geisinger; Leonard Learn, PharmD - Geisinger; Rachel Dragano, DNP, CRNP - Geisinger; Sandra Herr, BSN - Geisinger Health Plan; Henry Aftewicz, PharmD - Geisinger; Jamie Kerestes, PharmD - Geisinger; Malory Sponenberg, MSN, RN - Geisinger; Michael Stoppie, MS - Geisinger; Kimberly Mackes, - - Geisinger; Nathaniel Stark, BS - Geisinger; Eric Wright, PharmD, MPH - Geisinger;
Apoorva
Pradhan,
BAMS, MPH - Geisinger
Tracking Unmet Needs and Episodes of Care to Review Population Health Analytics for Behavioral Health Patients
Poster Number: P126
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Change Management, Data Mining, Healthcare Economics/Cost of Care, Informatics Implementation, Usability, User-centered Design Methods, Policy, Population Health
Primary Track: Applications
Programmatic Theme: Clinical Informatics
Children’s Healthcare of Atlanta’s Behavioral and Mental Health Center identified a gap in the ability to collect discrete data. Development of an Epic SmartBlock allowed for a better understanding of the evidenced-based treatments providers are recommending and whether patient’s needs are being met or why not. Data analysis will be used for informed decisions of investments, personnel hiring, and legislative lobbying against mental health parity violations in Georgia.
Speaker:
Kayla Mays, DNP, APRN, PMHNP-BC
Children's Healthcare of Atlanta
Authors:
Naveen Muthu, MD - Children's Healthcare of Atlanta; Katherine Daniel, DNP, PMHNP-BC - Children's Healthcare of Atlanta;
Poster Number: P126
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Change Management, Data Mining, Healthcare Economics/Cost of Care, Informatics Implementation, Usability, User-centered Design Methods, Policy, Population Health
Primary Track: Applications
Programmatic Theme: Clinical Informatics
Children’s Healthcare of Atlanta’s Behavioral and Mental Health Center identified a gap in the ability to collect discrete data. Development of an Epic SmartBlock allowed for a better understanding of the evidenced-based treatments providers are recommending and whether patient’s needs are being met or why not. Data analysis will be used for informed decisions of investments, personnel hiring, and legislative lobbying against mental health parity violations in Georgia.
Speaker:
Kayla Mays, DNP, APRN, PMHNP-BC
Children's Healthcare of Atlanta
Authors:
Naveen Muthu, MD - Children's Healthcare of Atlanta; Katherine Daniel, DNP, PMHNP-BC - Children's Healthcare of Atlanta;
Kayla
Mays,
DNP, APRN, PMHNP-BC - Children's Healthcare of Atlanta
Federated Target Trial Emulation with Distributed Electronic Health Records Data across Multiple Sites for Estimating the Real-world Effectiveness of Treatments
Poster Number: P127
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Causal Inference, Real-World Evidence Generation, Machine Learning
Primary Track: Applications
We propose a Federated Learning-based Target Trial Emulation (FL-TTE) framework to estimate real-world treatment effects using electronic health records (EHRs) from distributed clinical institutions in a privacy-preserved way. Experiments on two different clinical research network datasets validated the effectiveness of FL-TTE and its potential of enabling privacy-preserving multi-institutional collaborations on generating robust real-world evidence for treatments.
Speaker:
Haoyang Li, PhD
Weill Cornell Medicine
Authors:
Haoyang Li, PhD - Weill Cornell Medicine; Chengxi Zang, PHD - Weill Cornell Medicine; Zhenxing Xu, Ph.D. - Weill Cornell Medical College; Weishen Pan, PhD - Weill Cornell Medicine; Suraj Rajendran, PhD - Weill Cornell Medicine; Yong Chen, PhD - University of Pennsylvania; Fei Wang, PhD - Weill Cornell Medicine;
Poster Number: P127
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Causal Inference, Real-World Evidence Generation, Machine Learning
Primary Track: Applications
We propose a Federated Learning-based Target Trial Emulation (FL-TTE) framework to estimate real-world treatment effects using electronic health records (EHRs) from distributed clinical institutions in a privacy-preserved way. Experiments on two different clinical research network datasets validated the effectiveness of FL-TTE and its potential of enabling privacy-preserving multi-institutional collaborations on generating robust real-world evidence for treatments.
Speaker:
Haoyang Li, PhD
Weill Cornell Medicine
Authors:
Haoyang Li, PhD - Weill Cornell Medicine; Chengxi Zang, PHD - Weill Cornell Medicine; Zhenxing Xu, Ph.D. - Weill Cornell Medical College; Weishen Pan, PhD - Weill Cornell Medicine; Suraj Rajendran, PhD - Weill Cornell Medicine; Yong Chen, PhD - University of Pennsylvania; Fei Wang, PhD - Weill Cornell Medicine;
Haoyang
Li,
PhD - Weill Cornell Medicine
mCodeGPT for GARDE: An LLM‐Based Pipeline for Extracting Family History from Clinical Notes
Poster Number: P128
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Cancer Prevention, Large Language Models (LLMs), Information Retrieval, Clinical Decision Support, Precision Medicine
Primary Track: Applications
Programmatic Theme: Clinical Informatics
GARDE is a population clinical decision support (CDS) platform designed to identify individuals with hereditary risk of developing breast, ovarian, and colorectal cancers. We developed mCodeGPT, an innovative large language model (LLM)–pipeline that extracts and reconciles family health history (FHx) from both structured (SD) and unstructured (UD) clinical notes. An algorithm using SD augmented with UD identified 29.0% patients meeting GARDE genetic testing criteria, compared with 37.8% of patients with and integrated SD+UD algorithm.
Speaker:
Uday Singh, MS
UTHealth McWilliams School of Biomedical Informatics
Authors:
Richard Bradshaw, MS - University of Utah Health Sciences; Emerson Borsato, PhD - University of Utah; Lishan Yu, PhD - UThealth Houston McWilliams School of Biomedical Informatics; Jiantao Bian, PhD - Biomedical Informatics Department, University of Utah; Fitia Rakoto, BS - UT Health Houston; Dulin Wang, MS - The University of Texas Health Science Center at Houston; Leyang Sun, Doctorate - UTHealth; Yuning Xie, PhD - UThealth Houston; Kensaku Kawamoto, MD, PhD, MHS - University of Utah; Xiaoqian Jiang, PhD - University of Texas Health Science Center at Houston; Guilherme Del Fiol, MD, PhD - University of Utah;
Poster Number: P128
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Cancer Prevention, Large Language Models (LLMs), Information Retrieval, Clinical Decision Support, Precision Medicine
Primary Track: Applications
Programmatic Theme: Clinical Informatics
GARDE is a population clinical decision support (CDS) platform designed to identify individuals with hereditary risk of developing breast, ovarian, and colorectal cancers. We developed mCodeGPT, an innovative large language model (LLM)–pipeline that extracts and reconciles family health history (FHx) from both structured (SD) and unstructured (UD) clinical notes. An algorithm using SD augmented with UD identified 29.0% patients meeting GARDE genetic testing criteria, compared with 37.8% of patients with and integrated SD+UD algorithm.
Speaker:
Uday Singh, MS
UTHealth McWilliams School of Biomedical Informatics
Authors:
Richard Bradshaw, MS - University of Utah Health Sciences; Emerson Borsato, PhD - University of Utah; Lishan Yu, PhD - UThealth Houston McWilliams School of Biomedical Informatics; Jiantao Bian, PhD - Biomedical Informatics Department, University of Utah; Fitia Rakoto, BS - UT Health Houston; Dulin Wang, MS - The University of Texas Health Science Center at Houston; Leyang Sun, Doctorate - UTHealth; Yuning Xie, PhD - UThealth Houston; Kensaku Kawamoto, MD, PhD, MHS - University of Utah; Xiaoqian Jiang, PhD - University of Texas Health Science Center at Houston; Guilherme Del Fiol, MD, PhD - University of Utah;
Uday
Singh,
MS - UTHealth McWilliams School of Biomedical Informatics
CATT (ClinGen AI data Transformation Tool): A Toolkit for Mitigating Harmful Hallucinations in Genetic Variant Summarization by Large Language Models
Poster Number: P129
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Bioinformatics, Artificial Intelligence, Large Language Models (LLMs), Computational Biology, Omics (genomics, metabolomics, proteomics, transcriptomics, etc.) and Integrative Analyses
Primary Track: Applications
Programmatic Theme: Translational Bioinformatics
Large language models (LLMs) are prone to hallucinations, especially when prompts lack evidence-based context. In this study, we developed an open-source tool - ClinGen AI Data Transformation Tool (CATT) - that mitigates harmful hallucinations in LLM-generated genetic variant summaries. By providing expert-curated knowledge via processing ClinGen, ClinVar, and GenCC databases into LLM-compatible files, CATT effectively mitigates harmful hallucinations made by LLMs, thus improving the clinical utility of LLM-generated variant summaries.
Speaker:
Xinsong Du, Ph.D.
Brigham and Women's Hospital/Harvard Medical School
Authors:
Xinsong Du, Ph.D. - Brigham and Women's Hospital/Harvard Medical School; Anna Angy, M.S. - Mass General Brigham; Michael Oats, M.S. - Mass General Brigham; Yifei Wang, Ph.D. - Brandeis University; Xinyi Wang, M.S. - Harvard Medical School; Joseph Plasek, PhD - Mass General Brigham; Samuel Aronson - Partners HealthCare Personalized Medicine; Matthew Lebo, Ph.D. - Brigham and Women's Hospital/Harvard Medical School; Li Zhou, MD, PhD, FACMI, FIAHSI, FAMIA - Brigham and Women's Hospital, Harvard Medical School;
Poster Number: P129
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Bioinformatics, Artificial Intelligence, Large Language Models (LLMs), Computational Biology, Omics (genomics, metabolomics, proteomics, transcriptomics, etc.) and Integrative Analyses
Primary Track: Applications
Programmatic Theme: Translational Bioinformatics
Large language models (LLMs) are prone to hallucinations, especially when prompts lack evidence-based context. In this study, we developed an open-source tool - ClinGen AI Data Transformation Tool (CATT) - that mitigates harmful hallucinations in LLM-generated genetic variant summaries. By providing expert-curated knowledge via processing ClinGen, ClinVar, and GenCC databases into LLM-compatible files, CATT effectively mitigates harmful hallucinations made by LLMs, thus improving the clinical utility of LLM-generated variant summaries.
Speaker:
Xinsong Du, Ph.D.
Brigham and Women's Hospital/Harvard Medical School
Authors:
Xinsong Du, Ph.D. - Brigham and Women's Hospital/Harvard Medical School; Anna Angy, M.S. - Mass General Brigham; Michael Oats, M.S. - Mass General Brigham; Yifei Wang, Ph.D. - Brandeis University; Xinyi Wang, M.S. - Harvard Medical School; Joseph Plasek, PhD - Mass General Brigham; Samuel Aronson - Partners HealthCare Personalized Medicine; Matthew Lebo, Ph.D. - Brigham and Women's Hospital/Harvard Medical School; Li Zhou, MD, PhD, FACMI, FIAHSI, FAMIA - Brigham and Women's Hospital, Harvard Medical School;
Xinsong
Du,
Ph.D. - Brigham and Women's Hospital/Harvard Medical School
Leveraging SHAP-based clustering to improve cluster quality, interpretation and visualization
Poster Number: P130
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Bioinformatics, Artificial Intelligence, Machine Learning, Data Mining
Primary Track: Foundations
Programmatic Theme: Clinical Informatics
Machine learning methods have long been considered a ‘black box:’ experts cannot get a clear understanding of how predictions are calculated. SHAP values are usually used in model interpretation to help solve this problem. Here we explore how we can use SHAP value-based clustering to help explain predictions. In this paper, we provide a pipeline to perform SHAP-based clustering. We evaluate SHAP-based clustering quality on simulated data and real world urinary tract infection data in different SHAP generated models. We then examine cluster interpretation through Skope-Rules, and visualization via tSNE. We found that SHAP-based clustering can lead to better cluster quality when compared with unsupervised clustering. In addition, among all supervised models we examined to generate SHAP values (logistic regression, random forest, XGBoost), random forest outperformed the other two models. Results suggest SHAP-based clustering may have a larger role in model interpretation in the future than it does now.
Speaker:
Xueting Wang, Master of Public Health
Yale University
Authors:
Jihoon Kim, PhD - Yale University; Mark Iscoe, MD, MHS; Lucila Ohno-Machado, M.D., PhD, MBA - Yale University;
Poster Number: P130
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Bioinformatics, Artificial Intelligence, Machine Learning, Data Mining
Primary Track: Foundations
Programmatic Theme: Clinical Informatics
Machine learning methods have long been considered a ‘black box:’ experts cannot get a clear understanding of how predictions are calculated. SHAP values are usually used in model interpretation to help solve this problem. Here we explore how we can use SHAP value-based clustering to help explain predictions. In this paper, we provide a pipeline to perform SHAP-based clustering. We evaluate SHAP-based clustering quality on simulated data and real world urinary tract infection data in different SHAP generated models. We then examine cluster interpretation through Skope-Rules, and visualization via tSNE. We found that SHAP-based clustering can lead to better cluster quality when compared with unsupervised clustering. In addition, among all supervised models we examined to generate SHAP values (logistic regression, random forest, XGBoost), random forest outperformed the other two models. Results suggest SHAP-based clustering may have a larger role in model interpretation in the future than it does now.
Speaker:
Xueting Wang, Master of Public Health
Yale University
Authors:
Jihoon Kim, PhD - Yale University; Mark Iscoe, MD, MHS; Lucila Ohno-Machado, M.D., PhD, MBA - Yale University;
Xueting
Wang,
Master of Public Health - Yale University
Multi-Omics Analysis and Treg-Targeted Therapy Prediction for Breast Cancer
Poster Number: P131
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Bioinformatics, Computational Biology, Precision Medicine, Systems Biology
Primary Track: Foundations
Programmatic Theme: Translational Bioinformatics
Regulatory T cells (Tregs) play a critical role in immunosuppression within the tumor microenvironment (TME) and contribute to immune evasion in breast cancer. While immune checkpoint inhibitors (ICIs) have shown promise, their efficacy remains limited, potentially due to Treg-mediated suppression. In this study, we applied multi-omics integration and machine learning-based clustering to identify a Treg-enriched breast cancer subtype and assess its molecular and genomic characteristics.
To further evaluate the therapeutic potential of Treg modulation, we leveraged an existing quantitative systems pharmacology (QSP) model simulating the TNBC TME. By systematically adjusting Treg influx and density, we examined their impact on CD8+ T cell activation and cancer stem cell death, with and without ICI therapy. The simulation revealed that Treg-targeted therapy alone had a limited effect, whereas its combination with ICIs significantly enhanced anti-tumor responses.
Our findings highlight the potential of Treg modulation as a complementary approach to ICI therapy and underscore the need for biomarker-driven patient stratification in breast cancer immunotherapy.
Speaker:
Nari Kim, Ph.D.
Asan Medical Center
Authors:
Kyungwon Kim, M.D, Ph.D. - Asan Medical Center; Seongwon Na, Ph.D. - Asan Medical Center;
Poster Number: P131
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Bioinformatics, Computational Biology, Precision Medicine, Systems Biology
Primary Track: Foundations
Programmatic Theme: Translational Bioinformatics
Regulatory T cells (Tregs) play a critical role in immunosuppression within the tumor microenvironment (TME) and contribute to immune evasion in breast cancer. While immune checkpoint inhibitors (ICIs) have shown promise, their efficacy remains limited, potentially due to Treg-mediated suppression. In this study, we applied multi-omics integration and machine learning-based clustering to identify a Treg-enriched breast cancer subtype and assess its molecular and genomic characteristics.
To further evaluate the therapeutic potential of Treg modulation, we leveraged an existing quantitative systems pharmacology (QSP) model simulating the TNBC TME. By systematically adjusting Treg influx and density, we examined their impact on CD8+ T cell activation and cancer stem cell death, with and without ICI therapy. The simulation revealed that Treg-targeted therapy alone had a limited effect, whereas its combination with ICIs significantly enhanced anti-tumor responses.
Our findings highlight the potential of Treg modulation as a complementary approach to ICI therapy and underscore the need for biomarker-driven patient stratification in breast cancer immunotherapy.
Speaker:
Nari Kim, Ph.D.
Asan Medical Center
Authors:
Kyungwon Kim, M.D, Ph.D. - Asan Medical Center; Seongwon Na, Ph.D. - Asan Medical Center;
Nari
Kim,
Ph.D. - Asan Medical Center
Association of Serious Mental Illnesses with PASC Development: Evidence from the RECOVER PCORnet EHR Program
Poster Number: P132
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Bioinformatics, Data Sharing, Racial disparities, Diversity, Equity, Inclusion, and Accessibility, Healthcare Quality, Infectious Diseases and Epidemiology, Quantitative Methods, Public Health
Primary Track: Applications
Programmatic Theme: Public Health Informatics
This study investigated whether individuals with serious mental illness (SMI) were more likely to develop Post-acute Sequelae of SARS-CoV-2 infection (PASC) than those without SMI. Using data from over 1.6 million COVID-19 patients (March 2020–October 2022), we found that 28% of patients with SMI developed PASC, with significantly increased adjusted odds (OR: 1.1). Predictors included older age, racial/ethnic minorities, chronic disease, and severe initial COVID-19 infection. Our findings underscored the need for targeted approaches to manage PASC in SMI patients, particularly among underserved populations.
Speaker:
Veer Vekaria, BS
Weill Cornell Medicine / NewYork-Presbyterian Hospital
Authors:
Rohith Kumar Thiruvalluru - Weill Cornell Medicine; Zoe Verzani, MS - Weill Cornell Medicine; Sajjad Abedian, MS - Weill Cornell Medicine; Braja Gopal Patra, PhD - Weill Cornell Medicine; Dept of Population Health Sciences; Div of Health Informatics; Mark Olfson, MD, MPH - Columbia University; Jyotishman Pathak, PhD - Weill Cornell Medical College, Cornell University;
Poster Number: P132
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Bioinformatics, Data Sharing, Racial disparities, Diversity, Equity, Inclusion, and Accessibility, Healthcare Quality, Infectious Diseases and Epidemiology, Quantitative Methods, Public Health
Primary Track: Applications
Programmatic Theme: Public Health Informatics
This study investigated whether individuals with serious mental illness (SMI) were more likely to develop Post-acute Sequelae of SARS-CoV-2 infection (PASC) than those without SMI. Using data from over 1.6 million COVID-19 patients (March 2020–October 2022), we found that 28% of patients with SMI developed PASC, with significantly increased adjusted odds (OR: 1.1). Predictors included older age, racial/ethnic minorities, chronic disease, and severe initial COVID-19 infection. Our findings underscored the need for targeted approaches to manage PASC in SMI patients, particularly among underserved populations.
Speaker:
Veer Vekaria, BS
Weill Cornell Medicine / NewYork-Presbyterian Hospital
Authors:
Rohith Kumar Thiruvalluru - Weill Cornell Medicine; Zoe Verzani, MS - Weill Cornell Medicine; Sajjad Abedian, MS - Weill Cornell Medicine; Braja Gopal Patra, PhD - Weill Cornell Medicine; Dept of Population Health Sciences; Div of Health Informatics; Mark Olfson, MD, MPH - Columbia University; Jyotishman Pathak, PhD - Weill Cornell Medical College, Cornell University;
Veer
Vekaria,
BS - Weill Cornell Medicine / NewYork-Presbyterian Hospital
Multimodal Retrospective Identification of Cancer Recurrence Severity
Poster Number: P133
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Artificial Intelligence, Bioinformatics, Data Mining
Primary Track: Applications
Cancer recurrence severity (local, regional or distant spread) is an important measure of treatment effectiveness but is not captured by cancer registries. We assess the ability of Large Lange Models (LLM) to retrospectively identify cancer recurrence severity from Electronic Health Records (EHR) including clinic text. Our preliminary results on 403 reportable cancer cases only limited ability for either classical machine learning classification methods or LLMs to identify recurrence severity.
Speaker:
Chris Coffee, BA
UAB
Author:
Chris COffee, BA - UAB;
Poster Number: P133
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Artificial Intelligence, Bioinformatics, Data Mining
Primary Track: Applications
Cancer recurrence severity (local, regional or distant spread) is an important measure of treatment effectiveness but is not captured by cancer registries. We assess the ability of Large Lange Models (LLM) to retrospectively identify cancer recurrence severity from Electronic Health Records (EHR) including clinic text. Our preliminary results on 403 reportable cancer cases only limited ability for either classical machine learning classification methods or LLMs to identify recurrence severity.
Speaker:
Chris Coffee, BA
UAB
Author:
Chris COffee, BA - UAB;
Chris
Coffee,
BA - UAB
Active Learning for Forecasting Severity among Patients with Post Acute Sequelae of SARS-CoV-2
Poster Number: P134
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Artificial Intelligence, Causal Inference, Natural Language Processing
Primary Track: Applications
The long-term effects of Postacute Sequelae of SARS-CoV-2, known as PASC, pose a significant challenge to healthcare systems worldwide. Accurate identification of progression events—such as hospitalization and reinfection—is essential for effective patient management and resource allocation. However, traditional models trained on structured data struggle to capture the nuanced progression of PASC. In this study, we introduce the first publicly available cohort of 18 PASC patients, with text time series features based on Large Language Model Llama-3.1-70B-Instruct and clinical risk annotated by clinical expert \cite{touvron2023llama}. We propose an Active Attention Network to predict the clinical risk and identify progression events related to the risk. By integrating human expertise with active learning, we aim to enhance clinical risk prediction accuracy and enable progression events identification with fewer number of annotation. The ultimate goal is to improves patient care and decision-making for SARS-CoV-2 patient.
Speaker:
Jeremy Weiss, MD PhD
National Library of Medicine
Authors:
Amar Sra, MD, MS - The George Washington University; Jeremy Weiss, MD PhD - National Library of Medicine;
Poster Number: P134
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Artificial Intelligence, Causal Inference, Natural Language Processing
Primary Track: Applications
The long-term effects of Postacute Sequelae of SARS-CoV-2, known as PASC, pose a significant challenge to healthcare systems worldwide. Accurate identification of progression events—such as hospitalization and reinfection—is essential for effective patient management and resource allocation. However, traditional models trained on structured data struggle to capture the nuanced progression of PASC. In this study, we introduce the first publicly available cohort of 18 PASC patients, with text time series features based on Large Language Model Llama-3.1-70B-Instruct and clinical risk annotated by clinical expert \cite{touvron2023llama}. We propose an Active Attention Network to predict the clinical risk and identify progression events related to the risk. By integrating human expertise with active learning, we aim to enhance clinical risk prediction accuracy and enable progression events identification with fewer number of annotation. The ultimate goal is to improves patient care and decision-making for SARS-CoV-2 patient.
Speaker:
Jeremy Weiss, MD PhD
National Library of Medicine
Authors:
Amar Sra, MD, MS - The George Washington University; Jeremy Weiss, MD PhD - National Library of Medicine;
Jeremy
Weiss,
MD PhD - National Library of Medicine
GPT Model Based on Speech and Text Conversation in Mild Cognitive Impairment Screening
Poster Number: P135
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Artificial Intelligence, Chronic Care Management, Evaluation
Primary Track: Applications
Programmatic Theme: Clinical Informatics
This study explored the use of GPT-4, a large language model, in the screening of mild cognitive impairment (MCI) using speech-based interactions. A total of 66 individuals with MCI and 108 cognitively normal controls were randomly divided into training (70%) and test (30%) groups. Using standardized prompt-based methods derived from open-access DementiaBank datasets, GPT-4 effectively discriminated MCI cases from normal cognitive function. The text-based model achieved a sensitivity of 0.77, specificity of 0.83, and an AUC of 0.80 on the test set. Additionally, three linguistic indicators—tip-of-the-tongue phenomena, difficulty expressing complex ideas, and memory problems—significantly enhanced model accuracy (P < .001). However, performance degradation occurred due to limitations in speech-to-text conversion accuracy, underscoring the need for improved transcription and feature extraction. These findings demonstrate the potential of intelligent, scalable LLM-based tools in clinical MCI screening and emphasize the necessity of refining speech-to-text technology and relevant feature extraction for clinical applications.
Speaker:
Jialin Liu, MD
West China Hospital
Authors:
Xinxin Zhang, BS - West China School of Basic Medicine and Forensic Medicine, Sichuan University; Yueling Liu, BS - School of Software, Sichuan University; Ziyou Wang, BS - School of Clinical Medicine, Sichuan University; Yujie Tian, BS - School of Computer Science, Sichuan University; Weizhen Li, BS - School of Life Sciences, Sichuan University; Jiaxi Li, BS - School of Software, Sichuan University; Siru Liu, PhD - Department of Biomedical Informatics, Vanderbilt University Medical Center; Jialin Liu, MD - Department of Medical Informatics, West China Medical School, Sichuan University;
Poster Number: P135
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Artificial Intelligence, Chronic Care Management, Evaluation
Primary Track: Applications
Programmatic Theme: Clinical Informatics
This study explored the use of GPT-4, a large language model, in the screening of mild cognitive impairment (MCI) using speech-based interactions. A total of 66 individuals with MCI and 108 cognitively normal controls were randomly divided into training (70%) and test (30%) groups. Using standardized prompt-based methods derived from open-access DementiaBank datasets, GPT-4 effectively discriminated MCI cases from normal cognitive function. The text-based model achieved a sensitivity of 0.77, specificity of 0.83, and an AUC of 0.80 on the test set. Additionally, three linguistic indicators—tip-of-the-tongue phenomena, difficulty expressing complex ideas, and memory problems—significantly enhanced model accuracy (P < .001). However, performance degradation occurred due to limitations in speech-to-text conversion accuracy, underscoring the need for improved transcription and feature extraction. These findings demonstrate the potential of intelligent, scalable LLM-based tools in clinical MCI screening and emphasize the necessity of refining speech-to-text technology and relevant feature extraction for clinical applications.
Speaker:
Jialin Liu, MD
West China Hospital
Authors:
Xinxin Zhang, BS - West China School of Basic Medicine and Forensic Medicine, Sichuan University; Yueling Liu, BS - School of Software, Sichuan University; Ziyou Wang, BS - School of Clinical Medicine, Sichuan University; Yujie Tian, BS - School of Computer Science, Sichuan University; Weizhen Li, BS - School of Life Sciences, Sichuan University; Jiaxi Li, BS - School of Software, Sichuan University; Siru Liu, PhD - Department of Biomedical Informatics, Vanderbilt University Medical Center; Jialin Liu, MD - Department of Medical Informatics, West China Medical School, Sichuan University;
Jialin
Liu,
MD - West China Hospital
Transforming Obesity Management Through Generative AI: Evaluating An AI Coach for Enhanced Patient Outcomes
Poster Number: P136
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Artificial Intelligence, Chronic Care Management, Health Equity
Primary Track: Applications
Programmatic Theme: Consumer Health Informatics
Obesity is a significant public health challenge. FusionCare AI's generative obesity coach improves patient adherence by delivering personalized, evidence-based interventions, leading to better glycemic control, fewer hospital readmissions, and increased satisfaction. Evaluated using the HumanELY framework, the system outperformed GPT-4o and Gemini by over 20% on critical clinical metrics. This scalable model holds promise for transforming obesity care within metabolic centers of excellence.
Speaker:
Aarit Atreja, Student
BrainX
Authors:
Aarit Atreja, Student - BrainX; Dharan Sankar Jaisankar, MS - GenServe.AI;
Poster Number: P136
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Artificial Intelligence, Chronic Care Management, Health Equity
Primary Track: Applications
Programmatic Theme: Consumer Health Informatics
Obesity is a significant public health challenge. FusionCare AI's generative obesity coach improves patient adherence by delivering personalized, evidence-based interventions, leading to better glycemic control, fewer hospital readmissions, and increased satisfaction. Evaluated using the HumanELY framework, the system outperformed GPT-4o and Gemini by over 20% on critical clinical metrics. This scalable model holds promise for transforming obesity care within metabolic centers of excellence.
Speaker:
Aarit Atreja, Student
BrainX
Authors:
Aarit Atreja, Student - BrainX; Dharan Sankar Jaisankar, MS - GenServe.AI;
Aarit
Atreja,
Student - BrainX
Transfer Learning for Electronic Health Records
Poster Number: P137
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Artificial Intelligence, Clinical Decision Support, Fairness and elimination of bias, Health Equity, Machine Learning
Primary Track: Applications
Programmatic Theme: Clinical Informatics
Transfer learning (TL) is a machine learning technique that utilizes pre-trained models to update new models, enabling researchers to leverage information from external tasks. While TL has found applications in healthcare research, it primarily addresses unstructured data and predominantly employs engineering-based TL methods, overlooking statistics-based TL techniques. This study presents a comprehensive review that addresses the gap in structured clinical and biomedical data applications of TL, providing detailed clinical perspectives alongside technical discussions.
Speaker:
Siqi Li, Bachelor of Science
Duke-NUS Medical School
Authors:
Siqi Li, Bachelor of Science - Duke-NUS Medical School; Xin Li, Master of science - Duke-NUS medical school; Qiming Wu, MSc - Duke NUS Medical School; Kunyu Yu, M.S. - Duke-NUS Medical School; Nan Liu, PhD - National University of Singapore;
Poster Number: P137
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Artificial Intelligence, Clinical Decision Support, Fairness and elimination of bias, Health Equity, Machine Learning
Primary Track: Applications
Programmatic Theme: Clinical Informatics
Transfer learning (TL) is a machine learning technique that utilizes pre-trained models to update new models, enabling researchers to leverage information from external tasks. While TL has found applications in healthcare research, it primarily addresses unstructured data and predominantly employs engineering-based TL methods, overlooking statistics-based TL techniques. This study presents a comprehensive review that addresses the gap in structured clinical and biomedical data applications of TL, providing detailed clinical perspectives alongside technical discussions.
Speaker:
Siqi Li, Bachelor of Science
Duke-NUS Medical School
Authors:
Siqi Li, Bachelor of Science - Duke-NUS Medical School; Xin Li, Master of science - Duke-NUS medical school; Qiming Wu, MSc - Duke NUS Medical School; Kunyu Yu, M.S. - Duke-NUS Medical School; Nan Liu, PhD - National University of Singapore;
Siqi
Li,
Bachelor of Science - Duke-NUS Medical School
Evaluating GPT-4 Versus Neurologist Assessments for Detecting Mild Cognitive Impairment in the Elderly
Poster Number: P138
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Artificial Intelligence, Clinical Decision Support, Diagnostic Systems
Primary Track: Applications
Programmatic Theme: Clinical Informatics
This study assesses the effectiveness of the GPT-4 model in screening for mild cognitive impairment (MCI) in the elderly and compares its performance with that of neurologists. Utilizing de-identified DementiaBank data from 174 participants (66 with MCI and 108 cognitively normal), GPT-4 evaluated test texts based on language analysis indicators, while neurologists assessed both text and speech data in a randomized and blinded manner. The results revealed that GPT-4 achieved a significantly higher accuracy (0.81) compared to the neurologists (ranging from 0.41 to 0.49, p < 0.001). Furthermore, GPT-4 outperformed the neurologists in all diagnostic metrics, including true and false positives and negatives. Additionally, the study introduced a clinical risk assessment nomogram derived from the top ten features weighted by GPT-4, enhancing the diagnostic process for MCI. The findings suggest that GPT-4 holds promise as a diagnostic aid for MCI, potentially improving patient outcomes and alleviating healthcare burdens; however, further clinical validation in diverse real-world settings is warranted.
Speaker:
Jialin Liu, MD
West China Hospital Sichuan University
Authors:
Hao Yang, ME - Department of Medical Informatics, West China Hospital, Sichuan University; Ruihan Wang, MD - Department of Neurology, West China Hospital, Sichuan University; Changyu Wang, BS - Department of Medical Informatics, West China Medical School, Sichuan University; Hui Gao, MD - Department of Neurology, West China Hospital, Sichuan University; Hanlin Cai, MD - Department of Neurology, West China Hospital, Sichuan University; Siru Liu, PhD - Department of Biomedical Informatics, Vanderbilt University Medical Center;
Poster Number: P138
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Artificial Intelligence, Clinical Decision Support, Diagnostic Systems
Primary Track: Applications
Programmatic Theme: Clinical Informatics
This study assesses the effectiveness of the GPT-4 model in screening for mild cognitive impairment (MCI) in the elderly and compares its performance with that of neurologists. Utilizing de-identified DementiaBank data from 174 participants (66 with MCI and 108 cognitively normal), GPT-4 evaluated test texts based on language analysis indicators, while neurologists assessed both text and speech data in a randomized and blinded manner. The results revealed that GPT-4 achieved a significantly higher accuracy (0.81) compared to the neurologists (ranging from 0.41 to 0.49, p < 0.001). Furthermore, GPT-4 outperformed the neurologists in all diagnostic metrics, including true and false positives and negatives. Additionally, the study introduced a clinical risk assessment nomogram derived from the top ten features weighted by GPT-4, enhancing the diagnostic process for MCI. The findings suggest that GPT-4 holds promise as a diagnostic aid for MCI, potentially improving patient outcomes and alleviating healthcare burdens; however, further clinical validation in diverse real-world settings is warranted.
Speaker:
Jialin Liu, MD
West China Hospital Sichuan University
Authors:
Hao Yang, ME - Department of Medical Informatics, West China Hospital, Sichuan University; Ruihan Wang, MD - Department of Neurology, West China Hospital, Sichuan University; Changyu Wang, BS - Department of Medical Informatics, West China Medical School, Sichuan University; Hui Gao, MD - Department of Neurology, West China Hospital, Sichuan University; Hanlin Cai, MD - Department of Neurology, West China Hospital, Sichuan University; Siru Liu, PhD - Department of Biomedical Informatics, Vanderbilt University Medical Center;
Jialin
Liu,
MD - West China Hospital Sichuan University
Large Language Models in Breast Cancer Management: A Scope Review
Poster Number: P139
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Artificial Intelligence, Clinical Decision Support, Large Language Models (LLMs)
Primary Track: Applications
Programmatic Theme: Clinical Informatics
This scope review systematically assesses recent literature on the applications of Large Language Models (LLMs) in breast cancer management. A comprehensive search of SCI, MEDLINE, and INSPEC databases identified 13 studies from 2022 to 2024 focusing on clinical decision support and medical report analysis. Findings suggest significant potential for LLMs to improve diagnostic accuracy, treatment decision-making, patient outcome prediction, and automated extraction of clinical data from medical reports. However, key challenges remain regarding model accuracy, interpretability, clinician acceptance, ethical transparency, and patient privacy. Future research should prioritize rigorous validation, enhanced usability assessments, transparent methodologies, and broader exploration of LLM applications in imaging and personalized medicine.
Speaker:
Jialin Liu, MD
West China Hosiptal Sichuan University
Authors:
Zeng Wang, MS - West China Hospital Sichuna University; Siru Liu, PhD - Department of Biomedical Informatics, Vanderbilt University Medical Center;
Poster Number: P139
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Artificial Intelligence, Clinical Decision Support, Large Language Models (LLMs)
Primary Track: Applications
Programmatic Theme: Clinical Informatics
This scope review systematically assesses recent literature on the applications of Large Language Models (LLMs) in breast cancer management. A comprehensive search of SCI, MEDLINE, and INSPEC databases identified 13 studies from 2022 to 2024 focusing on clinical decision support and medical report analysis. Findings suggest significant potential for LLMs to improve diagnostic accuracy, treatment decision-making, patient outcome prediction, and automated extraction of clinical data from medical reports. However, key challenges remain regarding model accuracy, interpretability, clinician acceptance, ethical transparency, and patient privacy. Future research should prioritize rigorous validation, enhanced usability assessments, transparent methodologies, and broader exploration of LLM applications in imaging and personalized medicine.
Speaker:
Jialin Liu, MD
West China Hosiptal Sichuan University
Authors:
Zeng Wang, MS - West China Hospital Sichuna University; Siru Liu, PhD - Department of Biomedical Informatics, Vanderbilt University Medical Center;
Jialin
Liu,
MD - West China Hosiptal Sichuan University
Artificial Intelligence Model for Anemia Treatment in Dialysis Patients
Poster Number: P140
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Artificial Intelligence, Clinical Decision Support, Deep Learning
Primary Track: Foundations
Programmatic Theme: Clinical Informatics
The GRU+Attention model exhibited strong predictive performance for both hemoglobin levels and ESA dosages in dialysis patients. For hemoglobin prediction, the model achieved a Mean Squared Error (MSE) of 0.3422, Root Mean Squared Error (RMSE) of 0.5850, Mean Absolute Error (MAE) of 0.3437, Mean Error (ME) of 0.0087, and an R-squared value of 0.7491, indicating good accuracy and consistency with observed values. In ESA dosage prediction, the model significantly outperformed all baseline methods, with MSE: 16.3620, RMSE: 4.0450, MAE: 1.3326, and R-squared: 0.9639. These results highlight the model's ability to accurately predict individual patient responses and recommend clinically relevant ESA doses. The findings demonstrate the potential utility of this AI-based approach for improving personalized anemia management in chronic kidney disease patients undergoing dialysis.
Speaker:
Dong Young Park, BS
College of Pharmacy, The Catholic University of Korea
Authors:
Se Hun Oh, BS - College of Pharmacy, The Catholic University of Korea; Yun-Kyoung Song, Ph.D - 1College of Pharmacy, The Catholic University of Korea; Seung Yun Chae, MD, Ph.D - 2Incheon St. Mary`s Hospital, Division of Nephrology, Department of Internal Medicine, College of Medicine, The Catholic University of Korea, Seoul, Republic of Korea; Byung Ha Chung, MD, Ph.D - Seoul St. Mary’s Hospital, Division of Nephrology, Department of Internal Medicine, College of Medicine, The Catholic University of Korea, Seoul, Republic of Korea; Hye Eun Yoon, MD, Ph.D - Seoul St. Mary’s Hospital, Division of Nephrology, Department of Internal Medicine, College of Medicine, The Catholic University of Korea, Seoul, Republic of Korea;
Poster Number: P140
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Artificial Intelligence, Clinical Decision Support, Deep Learning
Primary Track: Foundations
Programmatic Theme: Clinical Informatics
The GRU+Attention model exhibited strong predictive performance for both hemoglobin levels and ESA dosages in dialysis patients. For hemoglobin prediction, the model achieved a Mean Squared Error (MSE) of 0.3422, Root Mean Squared Error (RMSE) of 0.5850, Mean Absolute Error (MAE) of 0.3437, Mean Error (ME) of 0.0087, and an R-squared value of 0.7491, indicating good accuracy and consistency with observed values. In ESA dosage prediction, the model significantly outperformed all baseline methods, with MSE: 16.3620, RMSE: 4.0450, MAE: 1.3326, and R-squared: 0.9639. These results highlight the model's ability to accurately predict individual patient responses and recommend clinically relevant ESA doses. The findings demonstrate the potential utility of this AI-based approach for improving personalized anemia management in chronic kidney disease patients undergoing dialysis.
Speaker:
Dong Young Park, BS
College of Pharmacy, The Catholic University of Korea
Authors:
Se Hun Oh, BS - College of Pharmacy, The Catholic University of Korea; Yun-Kyoung Song, Ph.D - 1College of Pharmacy, The Catholic University of Korea; Seung Yun Chae, MD, Ph.D - 2Incheon St. Mary`s Hospital, Division of Nephrology, Department of Internal Medicine, College of Medicine, The Catholic University of Korea, Seoul, Republic of Korea; Byung Ha Chung, MD, Ph.D - Seoul St. Mary’s Hospital, Division of Nephrology, Department of Internal Medicine, College of Medicine, The Catholic University of Korea, Seoul, Republic of Korea; Hye Eun Yoon, MD, Ph.D - Seoul St. Mary’s Hospital, Division of Nephrology, Department of Internal Medicine, College of Medicine, The Catholic University of Korea, Seoul, Republic of Korea;
Dong Young
Park,
BS - College of Pharmacy, The Catholic University of Korea
Machine Learning-Based Model to Predict Response to Induction in Acute Myeloid Leukemia
Poster Number: P141
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Artificial Intelligence, Computational Biology, Machine Learning, Omics (genomics, metabolomics, proteomics, transcriptomics, etc.) and Integrative Analyses
Primary Track: Applications
This study evaluates machine learning (ML) models predicting response to induction therapy in acute myeloid leukemia (AML) patients, integrating clinical and gene expression data. After preprocessing data from the Beat AML trial, scaling, feature selection, and hyperparameter tuning were conducted. Six ML models were tested, with the XGB Classifier achieving the highest performance (AUROC = 0.86, AUPRC = 0.92). Findings highlight the potential of ML models in predicting AML treatment response and identifying relevant features.
Speaker:
Judy Bai, High school student
Greenhills School
Authors:
Jessica Patricoski, PhD Candidate, Computational Biology - Brown University Center for Computational Molecular Biology; Ece Uzun, PhD - Brown University Health/Brown University;
Poster Number: P141
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Artificial Intelligence, Computational Biology, Machine Learning, Omics (genomics, metabolomics, proteomics, transcriptomics, etc.) and Integrative Analyses
Primary Track: Applications
This study evaluates machine learning (ML) models predicting response to induction therapy in acute myeloid leukemia (AML) patients, integrating clinical and gene expression data. After preprocessing data from the Beat AML trial, scaling, feature selection, and hyperparameter tuning were conducted. Six ML models were tested, with the XGB Classifier achieving the highest performance (AUROC = 0.86, AUPRC = 0.92). Findings highlight the potential of ML models in predicting AML treatment response and identifying relevant features.
Speaker:
Judy Bai, High school student
Greenhills School
Authors:
Jessica Patricoski, PhD Candidate, Computational Biology - Brown University Center for Computational Molecular Biology; Ece Uzun, PhD - Brown University Health/Brown University;
Judy
Bai,
High school student - Greenhills School
Integrating GenAI Tools into Nursing Education: Enhancing Digital Health Competencies, Equity-Centered Innovation, and User-Centered Design
Poster Number: P142
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Artificial Intelligence, Curriculum Development, Mobile Health, Teaching Innovation, Nursing Informatics, Education and Training, User-centered Design Methods
Primary Track: Applications
Programmatic Theme: Academic Informatics / LIEAF
This poster describes an informatics education model that integrates Generative Artificial Intelligence (GenAI) tools into curriculum to enhance digital health competencies. By incorporating GenAI tools, students design and assess consumer health apps in fostering digital literacy, equity-centered innovation, and user-centered design. The model combines theory, practical application, and peer critique to strengthen ethical and usability principles. Future efforts should focus on evaluating AI-driven tools to enhance their effectiveness in health informatics education.
Speaker:
Grace Gao, PhD, DNP
St Catherine University
Author:
Poster Number: P142
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Artificial Intelligence, Curriculum Development, Mobile Health, Teaching Innovation, Nursing Informatics, Education and Training, User-centered Design Methods
Primary Track: Applications
Programmatic Theme: Academic Informatics / LIEAF
This poster describes an informatics education model that integrates Generative Artificial Intelligence (GenAI) tools into curriculum to enhance digital health competencies. By incorporating GenAI tools, students design and assess consumer health apps in fostering digital literacy, equity-centered innovation, and user-centered design. The model combines theory, practical application, and peer critique to strengthen ethical and usability principles. Future efforts should focus on evaluating AI-driven tools to enhance their effectiveness in health informatics education.
Speaker:
Grace Gao, PhD, DNP
St Catherine University
Author:
Grace
Gao,
PhD, DNP - St Catherine University
EnECG: Efficient Ensemble Learning for Electrocardiogram Multi-task Foundation Models
Poster Number: P143
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Artificial Intelligence, Data Mining, Large Language Models (LLMs), Machine Learning
Working Group: Student Working Group
Primary Track: Foundations
Electrocardiogram (ECG) analysis plays a vital role in the early detection, monitoring, and management of various cardiovascular conditions. While single-task models have achieved notable success in ECG interpretation, they fail to leverage the interrelated nature of various cardiac abnormalities. Conversely, developing a single model capable of extracting all relevant features for multiple tasks remains a significant challenge. Large-scale foundation models, though powerful, are not typically pretrained on ECG data, making full re-training or fine-tuning computationally expensive. To address these challenges, we propose EnECG, an ensemble-based framework that integrates multiple specialized foundation models, each excelling in different aspects of ECG interpretation. Instead of relying on a single model or single task, EnECG leverages the strengths of multiple specialized models to tackle a variety of ECG-based tasks. To mitigate the high computational cost of full re-training or fine-tuning, we introduce a lightweight adaptation strategy: attaching dedicated output layers to each foundation model and applying Low-Rank Adaptation (LoRA) only to these newly added parameters. Our experimental results demonstrate that by minimizing the scope of fine-tuning, EnECG significantly reduces computational and memory costs while maintaining the strong representational power of foundation models. This framework not only enhances feature extraction and predictive performance but also ensures practical efficiency for real-world clinical applications. The code is available at https://github.com/yuhaoxu99/EnECG.git.
Speaker:
Yuhao Xu, Master
Emory University
Yuhao Xu, Master
Emory University
Authors:
Yuhao Xu, Master - Emory University; Xiaoda Wang, PhD - Emory University; Jiaying Lu, PhD - Emory University School of Nursing's Center for Data Science; Sirui Ding, PhD - University of California San Francisco; Defu Cao, PhD - USC; Huaxiu Yao, PhD - UNC; Yan Liu, PhD - USC; Xiao Hu, PhD - Emory University; Carl Yang, Ph.D. - Emory University;
Poster Number: P143
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Artificial Intelligence, Data Mining, Large Language Models (LLMs), Machine Learning
Working Group: Student Working Group
Primary Track: Foundations
Electrocardiogram (ECG) analysis plays a vital role in the early detection, monitoring, and management of various cardiovascular conditions. While single-task models have achieved notable success in ECG interpretation, they fail to leverage the interrelated nature of various cardiac abnormalities. Conversely, developing a single model capable of extracting all relevant features for multiple tasks remains a significant challenge. Large-scale foundation models, though powerful, are not typically pretrained on ECG data, making full re-training or fine-tuning computationally expensive. To address these challenges, we propose EnECG, an ensemble-based framework that integrates multiple specialized foundation models, each excelling in different aspects of ECG interpretation. Instead of relying on a single model or single task, EnECG leverages the strengths of multiple specialized models to tackle a variety of ECG-based tasks. To mitigate the high computational cost of full re-training or fine-tuning, we introduce a lightweight adaptation strategy: attaching dedicated output layers to each foundation model and applying Low-Rank Adaptation (LoRA) only to these newly added parameters. Our experimental results demonstrate that by minimizing the scope of fine-tuning, EnECG significantly reduces computational and memory costs while maintaining the strong representational power of foundation models. This framework not only enhances feature extraction and predictive performance but also ensures practical efficiency for real-world clinical applications. The code is available at https://github.com/yuhaoxu99/EnECG.git.
Speaker:
Yuhao Xu, Master
Emory University
Yuhao Xu, Master
Emory University
Authors:
Yuhao Xu, Master - Emory University; Xiaoda Wang, PhD - Emory University; Jiaying Lu, PhD - Emory University School of Nursing's Center for Data Science; Sirui Ding, PhD - University of California San Francisco; Defu Cao, PhD - USC; Huaxiu Yao, PhD - UNC; Yan Liu, PhD - USC; Xiao Hu, PhD - Emory University; Carl Yang, Ph.D. - Emory University;
Yuhao
Xu,
Master - Emory University
Yuhao Xu, Master - Emory University
Yuhao Xu, Master - Emory University
Leveraging Wikipedia and a Large Language Model to Develop a Knowledge Resource Containing World Foods and Allergens
Poster Number: P144
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Artificial Intelligence, Large Language Models (LLMs), Public Health
Primary Track: Applications
Programmatic Theme: Consumer Health Informatics
This study aimed to develop a knowledge resource for matching global foods to common ingredients and potential allergens. It leveraged Wikipedia's 'Infobox food' template and OpenAI's ChatGPT 3.5 API to create a catalog of ingredients and allergens for over 6,000 foods. Frquency counts against our electronic health record (EHR) system were obtained to understand how often these might appear in health records. The worked revealed wheat as the most prevalent allergen. Issues with data consistency were noted. Use of Wikipedia and large language models (LLMs) showed potential but highlighted limitations and inconsistencies.
Speaker:
Simon Shavit, BA candidate
University of Michigan
Authors:
Simon Shavit, BA candidate - University of Michian; David Hanauer, MD - University of Michigan;
Poster Number: P144
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Artificial Intelligence, Large Language Models (LLMs), Public Health
Primary Track: Applications
Programmatic Theme: Consumer Health Informatics
This study aimed to develop a knowledge resource for matching global foods to common ingredients and potential allergens. It leveraged Wikipedia's 'Infobox food' template and OpenAI's ChatGPT 3.5 API to create a catalog of ingredients and allergens for over 6,000 foods. Frquency counts against our electronic health record (EHR) system were obtained to understand how often these might appear in health records. The worked revealed wheat as the most prevalent allergen. Issues with data consistency were noted. Use of Wikipedia and large language models (LLMs) showed potential but highlighted limitations and inconsistencies.
Speaker:
Simon Shavit, BA candidate
University of Michigan
Authors:
Simon Shavit, BA candidate - University of Michian; David Hanauer, MD - University of Michigan;
Simon
Shavit,
BA candidate - University of Michigan
Charting the Evolution of “AI” Mental Health Chatbots: From Rule-Based to Large Language Models
Poster Number: P145
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Artificial Intelligence, Large Language Models (LLMs), Evaluation, Fairness and elimination of bias, Health Equity
Primary Track: Applications
Programmatic Theme: Clinical Research Informatics
This systematic review (2020–2025) of 168 studies introduces two frameworks to evaluate AI mental health chatbots: a tripartite classification (rule-based, ML, LLM) and a three-tier evaluation system (Foundational Bench, Pilot Feasibility, Clinical Efficacy Testing). LLM-based chatbots surged to 44% of studies by 2024 but only 16% underwent rigorous clinical testing. Discrepancies between marketed claims and technical capabilities occurred in 39% of papers. The frameworks guide clinical validation to address global mental health workforce shortages.
Speaker:
Yining Hua, MSc
Harvard T.H. Chan School of Public Health
Authors:
Steve Siddals, MSc - King’s College London; Zilin Ma; Winna Xia, BS - Beth Israel Deaconess Medical Center; Christine Hau, BS - Beth Israel Deaconess Medical Center; Hongbin Na, MSc - University of Technology Sydney; Cyrus Ayubcha, PhD - Harvard Medical School; John Torous, MD - Harvard Medical School;
Poster Number: P145
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Artificial Intelligence, Large Language Models (LLMs), Evaluation, Fairness and elimination of bias, Health Equity
Primary Track: Applications
Programmatic Theme: Clinical Research Informatics
This systematic review (2020–2025) of 168 studies introduces two frameworks to evaluate AI mental health chatbots: a tripartite classification (rule-based, ML, LLM) and a three-tier evaluation system (Foundational Bench, Pilot Feasibility, Clinical Efficacy Testing). LLM-based chatbots surged to 44% of studies by 2024 but only 16% underwent rigorous clinical testing. Discrepancies between marketed claims and technical capabilities occurred in 39% of papers. The frameworks guide clinical validation to address global mental health workforce shortages.
Speaker:
Yining Hua, MSc
Harvard T.H. Chan School of Public Health
Authors:
Steve Siddals, MSc - King’s College London; Zilin Ma; Winna Xia, BS - Beth Israel Deaconess Medical Center; Christine Hau, BS - Beth Israel Deaconess Medical Center; Hongbin Na, MSc - University of Technology Sydney; Cyrus Ayubcha, PhD - Harvard Medical School; John Torous, MD - Harvard Medical School;
Yining
Hua,
MSc - Harvard T.H. Chan School of Public Health
LLM-Assisted Translation and Interpretation of Discharge Instructions for Spanish-speaking Cardiology Patients
Poster Number: P146
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Artificial Intelligence, Large Language Models (LLMs), Natural Language Processing, Diversity, Equity, Inclusion, and Accessibility, Health Equity, Patient Engagement and Preferences, Education and Training
Primary Track: Applications
Language barriers in discharge instructions contribute to worse health outcomes for Spanish-speaking patients. We used natural language processing metrics to evaluate Spanish translations of cardiology discharge documents generated by the large language model (LLM) GPT-4o. Results showed strong semantic fidelity, with LLM outputs comparable to professional translations. These findings support using LLMs to improve discharge communication and understanding, ultimately reducing health disparities. Future work includes human evaluation to assess readability, actionability, and cultural sensitivity.
Speaker:
Eduardo Perez-Guerrero, MD
Stanford University
Authors:
Asad Aali, MS - Stanford University; Isaac Bohart, MD - Stanford University; Sergio Perez, BS - University of Puerto Rico; Nicole Corso, BS - Stanford University; Sneha Jain, MD, MBA - Stanford University; Ramzi Dudum, MD, MPH - UCSF; Jason Hom, MD - Stanford University; Akshay Chaudhari, MD, PhD - Stanford University; Roxana Daneshjou, MD, PhD - Stanford University; Fatima Rodriguez, MD, MPH - Stanford University; Christine Santiago, MD, MPH - Stanford University;
Poster Number: P146
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Artificial Intelligence, Large Language Models (LLMs), Natural Language Processing, Diversity, Equity, Inclusion, and Accessibility, Health Equity, Patient Engagement and Preferences, Education and Training
Primary Track: Applications
Language barriers in discharge instructions contribute to worse health outcomes for Spanish-speaking patients. We used natural language processing metrics to evaluate Spanish translations of cardiology discharge documents generated by the large language model (LLM) GPT-4o. Results showed strong semantic fidelity, with LLM outputs comparable to professional translations. These findings support using LLMs to improve discharge communication and understanding, ultimately reducing health disparities. Future work includes human evaluation to assess readability, actionability, and cultural sensitivity.
Speaker:
Eduardo Perez-Guerrero, MD
Stanford University
Authors:
Asad Aali, MS - Stanford University; Isaac Bohart, MD - Stanford University; Sergio Perez, BS - University of Puerto Rico; Nicole Corso, BS - Stanford University; Sneha Jain, MD, MBA - Stanford University; Ramzi Dudum, MD, MPH - UCSF; Jason Hom, MD - Stanford University; Akshay Chaudhari, MD, PhD - Stanford University; Roxana Daneshjou, MD, PhD - Stanford University; Fatima Rodriguez, MD, MPH - Stanford University; Christine Santiago, MD, MPH - Stanford University;
Eduardo
Perez-Guerrero,
MD - Stanford University
Early Prediction of Neonatal Jaundice Using Clinical Data
Poster Number: P147
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Artificial Intelligence, Machine Learning, Pediatrics
Primary Track: Applications
Programmatic Theme: Clinical Research Informatics
Introduction:\
Jaundice is a common condition in newborns, affecting up to 60% of full-term infants and 80% of preterm infants. While most cases of symptoms are benign, severe or prolonged Jaundice can lead to complications and require hospitalization. Previous studies have used various machine learning methodologies to predict neonatal jaundice but did not consider the time of prediction. This research aims to study how soon it is possible to predict Jaundice with reasonable accuracy and to create machine learning models to make such predictions.
Methods:
We obtained data from MIMIC III our cohort was selected as first ICU neonates admission per patient, typically at birth between 2001 to 2012. Three dependent variables were created based on the total serum bilirubin: (1) more than 6 mg/dL in the first 8 to 24 hours of life; (2) 24–48 hours if TSB > 10 mg/dL; and (3) 48–72 hours if TSB > 13 mg/dL. The input variables comprised demographics, vitals, and labs. Random Forest, Logistic Regression, and Gradient Boosting were used. The effectiveness of models was assessed using AUC score.
Results
The 24–48 hours’ time window model achieved the highest AUC (0.72 for LR, 0.69 for RF, 0.68 for GB), followed closely by 48-72 hours with ACU (0.71 for LR, 0.68 for RF, 0.67 for GB). The 8–24 hours' time window had an AUC (0.66 for LR, 0.63 for RF, 0.57 for GB) indicating lower predictive power due to early-stage bilirubin metabolism.
Conclusions:
Our analysis identified 24-48 time as the best time window for predicting jaundice with reasonably high accuracy.
Our findings are consistent with the literature that the serum bilirubin level reaches a clinically detectable level by day 3 or 4. Therefore, his study helps predict it before the clinically detectable level, helping with improving clinical decision-making and neonatal outcomes.
Speaker:
ARWA ALZEER, phD Student
George Mason University
Authors:
ARWA ALZEER, phD Student - George Mason University; Janusz Wojtusiak, PhD - George Mason University;
Poster Number: P147
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Artificial Intelligence, Machine Learning, Pediatrics
Primary Track: Applications
Programmatic Theme: Clinical Research Informatics
Introduction:\
Jaundice is a common condition in newborns, affecting up to 60% of full-term infants and 80% of preterm infants. While most cases of symptoms are benign, severe or prolonged Jaundice can lead to complications and require hospitalization. Previous studies have used various machine learning methodologies to predict neonatal jaundice but did not consider the time of prediction. This research aims to study how soon it is possible to predict Jaundice with reasonable accuracy and to create machine learning models to make such predictions.
Methods:
We obtained data from MIMIC III our cohort was selected as first ICU neonates admission per patient, typically at birth between 2001 to 2012. Three dependent variables were created based on the total serum bilirubin: (1) more than 6 mg/dL in the first 8 to 24 hours of life; (2) 24–48 hours if TSB > 10 mg/dL; and (3) 48–72 hours if TSB > 13 mg/dL. The input variables comprised demographics, vitals, and labs. Random Forest, Logistic Regression, and Gradient Boosting were used. The effectiveness of models was assessed using AUC score.
Results
The 24–48 hours’ time window model achieved the highest AUC (0.72 for LR, 0.69 for RF, 0.68 for GB), followed closely by 48-72 hours with ACU (0.71 for LR, 0.68 for RF, 0.67 for GB). The 8–24 hours' time window had an AUC (0.66 for LR, 0.63 for RF, 0.57 for GB) indicating lower predictive power due to early-stage bilirubin metabolism.
Conclusions:
Our analysis identified 24-48 time as the best time window for predicting jaundice with reasonably high accuracy.
Our findings are consistent with the literature that the serum bilirubin level reaches a clinically detectable level by day 3 or 4. Therefore, his study helps predict it before the clinically detectable level, helping with improving clinical decision-making and neonatal outcomes.
Speaker:
ARWA ALZEER, phD Student
George Mason University
Authors:
ARWA ALZEER, phD Student - George Mason University; Janusz Wojtusiak, PhD - George Mason University;
ARWA
ALZEER,
phD Student - George Mason University
Predictive Utility of School Violence Risk Assessments on Prospective Gun Violence
Poster Number: P148
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Artificial Intelligence, Machine Learning, Real-World Evidence Generation, Pediatrics
Primary Track: Applications
Programmatic Theme: Clinical Research Informatics
While progress has been made in the study of crime-related risk factors and school-violence prevention programs, the current state-of-the-art still requires improvement. Our research focuses on evaluating the language used by students in schools to assess risk of self-violence and violence toward others. We developed a risk assessment protocol involving two sets of standardized interview questions including yes/no and open-ended questions: 1) BRACHA: Brief Rating of Aggression by Children and Adolescents (School Version); and 2) SSS: School Safety Scale. In this work, we examine the questions in the BRACHA and SSS that relate to gun ownership, gun violence, and weaponry to identify the predictive utility of these questions on prospective gun violence in schools. For two questions that relate to gun ownership and weaponry, most of the participants (46%) with negative attitudes (supportive of violence) were identified as high-risk individuals for school violence using the SSS assessment. Screening tools that incorporate targeted questions about behavioral environment around firearms, and access to firearms can help identify students at higher risk of engaging in or being victims of gun-related incidents.
Speaker:
Judith Dexheimer, PhD
Cincinnati Children's Hospital
Authors:
Lara Kanbar, PhD - CCHMC; Alexander Osborn, MS - Cincinnati Children's Hospital Medical Center; Andrew Cifuentes, BS - Cincinnati Children's Hospital Medical Center; Jennifer Combs, MS - Cincinnati Children's Hospital Medical Center; Michael Sorter, MD - Cincinnati Children's Hospital Medical Center; Drew Barzman, MD - Cincinnati Children's Hospital Medical Center; Judith Dexheimer, PhD - Cincinnati Children's Hospital;
Poster Number: P148
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Artificial Intelligence, Machine Learning, Real-World Evidence Generation, Pediatrics
Primary Track: Applications
Programmatic Theme: Clinical Research Informatics
While progress has been made in the study of crime-related risk factors and school-violence prevention programs, the current state-of-the-art still requires improvement. Our research focuses on evaluating the language used by students in schools to assess risk of self-violence and violence toward others. We developed a risk assessment protocol involving two sets of standardized interview questions including yes/no and open-ended questions: 1) BRACHA: Brief Rating of Aggression by Children and Adolescents (School Version); and 2) SSS: School Safety Scale. In this work, we examine the questions in the BRACHA and SSS that relate to gun ownership, gun violence, and weaponry to identify the predictive utility of these questions on prospective gun violence in schools. For two questions that relate to gun ownership and weaponry, most of the participants (46%) with negative attitudes (supportive of violence) were identified as high-risk individuals for school violence using the SSS assessment. Screening tools that incorporate targeted questions about behavioral environment around firearms, and access to firearms can help identify students at higher risk of engaging in or being victims of gun-related incidents.
Speaker:
Judith Dexheimer, PhD
Cincinnati Children's Hospital
Authors:
Lara Kanbar, PhD - CCHMC; Alexander Osborn, MS - Cincinnati Children's Hospital Medical Center; Andrew Cifuentes, BS - Cincinnati Children's Hospital Medical Center; Jennifer Combs, MS - Cincinnati Children's Hospital Medical Center; Michael Sorter, MD - Cincinnati Children's Hospital Medical Center; Drew Barzman, MD - Cincinnati Children's Hospital Medical Center; Judith Dexheimer, PhD - Cincinnati Children's Hospital;
Judith
Dexheimer,
PhD - Cincinnati Children's Hospital
How Can Artificial Intelligence Tools Assist Clinicians in Providing Patient-Centered Diabetes Care?
Poster Number: P149
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Artificial Intelligence, Natural Language Processing, Patient / Person Generated Health Data (Patient Reported Outcomes), Chronic Care Management
Primary Track: Applications
Programmatic Theme: Clinical Research Informatics
We explored AI tools to support clinicians in providing patient-centered care, analyzing 528,199 messages from 11,123 diabetes patients using natural language processing. Drafted AI tools were evaluated by five endocrinologists for perceived usefulness and risk. Effective tools included summarizing insurance changes and creating customizable patient guides. Findings suggest AI applications can enhance precision patient care for diabetes and beyond.
Speaker:
Jiyeong Kim, PhD
Stanford University
Authors:
Tina Hernandez-Boussard, PhD - Stanford University; Jonathan Chen, MD, PhD - Stanford University Hospital; Eleni Linos, MD, DrPh - Stanford School of Medicine;
Poster Number: P149
Presentation Time: 05:30 PM - 07:00 PM
Abstract Keywords: Artificial Intelligence, Natural Language Processing, Patient / Person Generated Health Data (Patient Reported Outcomes), Chronic Care Management
Primary Track: Applications
Programmatic Theme: Clinical Research Informatics
We explored AI tools to support clinicians in providing patient-centered care, analyzing 528,199 messages from 11,123 diabetes patients using natural language processing. Drafted AI tools were evaluated by five endocrinologists for perceived usefulness and risk. Effective tools included summarizing insurance changes and creating customizable patient guides. Findings suggest AI applications can enhance precision patient care for diabetes and beyond.
Speaker:
Jiyeong Kim, PhD
Stanford University
Authors:
Tina Hernandez-Boussard, PhD - Stanford University; Jonathan Chen, MD, PhD - Stanford University Hospital; Eleni Linos, MD, DrPh - Stanford School of Medicine;
Jiyeong
Kim,
PhD - Stanford University
Rapid Review of Models Assessing Suicide Risk from Patient Portal and Crisis Text Line Messages
Category
Poster - Student
Description
Custom CSS
double-click to edit, do not edit in source
11/16/2025 07:00 PM (Eastern Time (US & Canada))