Outcome-wide Analysis of Electronic Health Records Data for Identifying Sequelae in Behçet's Disease

Presentation Time: 08:00 AM - 08:12 AM

Abstract Keywords: Machine Learning, Bioinformatics, Precision Medicine, Chronic Care Management, Clinical Decision Support
Primary Track: Applications
Programmatic Theme: Clinical Informatics

Behçet’s disease (BD) is a rare, chronic inflammatory disorder often progressing for many years before being formally diagnosed. This study leverages the extensive EHR database within the Providence Saint Joseph Health system to conduct a high-dimensional analysis. The study aims to bridge the gap between disease progression and formal diagnosis by identifying significant associations with clinical outcomes. This approach enhances clinical decision support, reduces diagnostic delays, and contributes to precision medicine for rare diseases.

Speaker:
Aparajita Saha, PhD
University of Washington

Authors:
Aparajita Saha, PhD - University of Washington; Sevda Molani, PhD - Institutes for System Biology; Philip Mease, MD - Swedish Health Service; Jennifer Hadlock, MD - Institute for Systems Biology;

---

Silent Prospective Validation Reveals Performance Gap from Published Model of Hospital-Acquired Venous Thromboembolism

Presentation Time: 08:12 AM - 08:24 AM

Abstract Keywords: Data Transformation/ETL, Machine Learning, Clinical Decision Support, Artificial Intelligence
Primary Track: Applications
Programmatic Theme: Clinical Informatics

Translating Healthcare AI into practice requires multiple steps to achieve Action-Oriented AI. The fourth phase emphasizes prospective validation of risk algorithms in production settings prior to deployment. This phase might unearth performance gaps as it has in real-time suicide risk prediction. In this study, our multidisciplinary team conducted silent validation of a published risk model of Hospital Acquired Venous Thromboembolism (HA-VTE), a leading cause of death in hospitalized patients. Here, a published HA-VTE risk model was applied to Epic Clarity-sourced data for the same admissions as those used in the original published study. Model performance was compared using AUROC and decision curves, where higher net benefit infers better performance. The original features were also retrained by running non-regularized logistic regression on Clarity-sourced features for those same encounters on the original, published train-test split. From January 1, 2018, to May 31, 2022, we calculated scores for 127,020 patients across 194,876 admissions at VUMC. The original published algorithm applied to the VUMC Research Derivative had an AUROC of 0.89, identical to the original, and the same model applied to Clarity had an AUROC of 0.58. Retraining features on Clarity data improved the AUROC to 0.83. Decision Curves show improvement of the retrained model toward original published performance though net benefit curves do not overlap. This study demonstrates the importance of and a path to facilitate deployment of validated research models. Silent validation illustrates performance gaps bridged with model retraining as shown here and, by extension, model updating in future work.

Speaker:
Colin Walsh, MD MA
Department of Biomedical Informatics, Vanderbilt University

Authors:
Colin Walsh, MD MA - Department of Biomedical Informatics, Vanderbilt University; Michael Ripperger - Vanderbilt University Medical Center; Ketan Jadhav, BS - Vanderbilt University Medical Center; Benjamin French, PhD - Vanderbilt University Medical Center; Henry Delmonico, PhD - Vanderbilt University Medical Center; Benjamin Tillman, MD - Vanderbilt University Medical Center; Amanda Mixon, MD - Vanderbilt University Medical Center; Shannon Walker, MD - Vanderbilt University Medical Center; Laurie Novak, PhD, MHSA - Vanderbilt University Medical Center Dept of Biomedical Informatics; Megan Salwei, PhD - Vanderbilt University Medical Center; Peter Embi, MD - VUMC;

---

Advancing HIT Maturity in Long-Term Post-Acute Care (LTPAC)

Presentation Time: 08:24 AM - 08:36 AM

Abstract Keywords: Surveys and Needs Analysis, Nursing Informatics, Policy
Primary Track: Foundations
Programmatic Theme: Clinical Informatics

This paper explores advances in HIT maturity measurement in LTPAC. It highlights best practices for assessing HIT
capabilities, usage, and integration to enhance patient care in LTPAC. Using a Delphi methodology, a revised HIT
Maturity model was developed, incorporating expert consensus on key content areas and maturity stages. The study
underscores the critical role of HIT in improving patient outcomes and calls for integrating maturity metrics into
national quality reporting systems to support LTPAC HIT.

Speaker:
Gregory Alexander, PhD, RN, FAAN, FACMI, FIAHSI
Columbia University School of Nursing

Authors:
Isaac Longobardi, BA - Moving Forward Coalition; Terrence O'Malley, MD - Harvard University;

---

Reducing Excessive Continuous Pulse Oximetry and Promoting Patient Safety: An Electronic Health Record-Based Quality Improvement Effort

Presentation Time: 08:36 AM - 08:48 AM

Abstract Keywords: Clinical Decision Support, Healthcare Quality, Informatics Implementation
Primary Track: Applications
Programmatic Theme: Clinical Informatics

We implemented an electronic health record (EHR)-based quality improvement effort aimed to safely reduce hospital-wide continuous pulse oximetry (CPO) utilization through a revised CPO order panel, admission and discharge orderset, and new best practice advisory. An interrupted time series analysis demonstrated a sustained reduction of 20,843 monthly CPO hours following intervention rollout, with no increase in patient harm events. These findings demonstrate how EHR interventions can effectively reduce CPO utilization without increasing patient harm.

Speaker:
Justin Zhang, MD
University of California, San Francisco

Authors:
Nirav Bhakta, MD PhD - University of California, San Francisco; Madeline Chicas, MHA - University of California, San Francisco; Anoop Muniyappa, MD, MS - UCSF; Sarah Kohut, MHA - University of California, San Francisco; Raman Khanna, MD, MAS - University of California, San Francisco; Nerys Benfield, MD MPH - University of California, San Francisco; Aida Venado Estrada, MD MAS - University of California, San Francisco;

---

Differences in Physician EHR Use by Telemedicine Intensity: Evidence from Two Academic Medical Centers

Presentation Time: 08:48 AM - 09:00 AM

Abstract Keywords: Telemedicine, Workflow, Quantitative Methods, Documentation Burden
Primary Track: Applications
Programmatic Theme: Clinical Informatics

The COVID-19 pandemic accelerated telemedicine adoption, impacting physician workflows and EHR use. Using EHR audit logs from two academic medical centers, we examined how telemedicine intensity influenced EHR time and physician activities. Higher telemedicine use was associated with increased EHR time and reduced ordering activity, with institutional differences in information-seeking and communication patterns. Findings highlight the need for telemedicine-optimized EHR tools, workflow adaptations, and support systems to balance physician workload and virtual care efficiency.

Speaker:
Seunghwan Kim, MS
Washington University in St. Louis

Authors:
Seunghwan Kim, MS - Washington University in St. Louis; Robert Thombley, BS - University of California, San Francisco; Elise Eiden, MS - Washington University in St. Louis; Sunny Lou, MD, PhD - Washington University, St. Louis; Julia Adler-Milstein, PhD, FACMI - UCSF School of Medicine; Thomas Kannampallil, PhD - Washington University School of Medicine; A J Holmgren, PhD - University of California, San Francisco;

---

Factors Associated with Timely, Delayed, and Missed VTE Diagnoses: An Electronic Clinical Quality Measure-Based Study

Presentation Time: 09:00 AM - 09:12 AM

Abstract Keywords: Nursing Informatics, Healthcare Quality, Patient Safety, Machine Learning
Primary Track: Applications
Programmatic Theme: Clinical Informatics

We expanded and validated an electronic clinical quality measure (eCQM) leveraging electronic health records to detect delayed and missed venous thromboembolism diagnoses in primary and urgent care settings. Among 1,529 symptomatic patients, delayed or missed diagnoses occurred in 73%. Presentation with isolated cough and history of prior VTE significantly increased the risk of diagnostic delays. This validated eCQM provides a scalable approach to enhancing clinical decision support and improving patient outcomes.

Speaker:
Veysel Baris, Nurse
Dokuz Eylul University

Authors:
Alexandra Frost, Bachelor's of Science (BS) - Brigham and Women's Hospital; Ann Hurley, DNSc - BWH; Minakshi Shukla, MD - The University of Alabama at Birmingham, Birmingham, Alabama, USA; Shifatul A. Apurba, PhD Candidate - The University of Alabama at Birmingham, Birmingham, Alabama, USA; Patricia Busta-Flores, PhD Candidate - The University of Alabama at Birmingham, Birmingham, Alabama, USA; Craig A. Grimes, MD - Massachusetts General Hospital, Boston, Massachusetts, USA; Li Zhou, MD, PhD - Department of Medicine, Brigham and Women’s Hospital, Boston, Massachusetts, USA; Stuart R. Lipsitz, ScD - Department of Medicine, Brigham and Women’s Hospital, Boston, Massachusetts; Calvin Franz, PhD - Eastern Research Group, Inc., Concord, Massachusetts, USA; Erin C. O’Fallon, MD - Department of Medicine, Brigham and Women’s Hospital, Boston, Massachusetts, USA; David M. Shahian, MD - Department of Medicine, Brigham and Women’s Hospital, Boston, Massachusetts, USA; Hacho B. Bohossian, MD - Massachusetts General Hospital, Boston, Massachusetts, USA; Mona Z. Hinrichsen, MD - Massachusetts General Hospital, Boston, Massachusetts, USA; David W. Bates, MD, MSc - Department of Medicine, Brigham and Women’s Hospital, Boston, Massachusetts, USA; Jin Chen, PhD - University of Alabama at Birmingham; Patricia C. Dykes, PhD, RN - Department of Medicine, Brigham and Women’s Hospital, Boston, Massachusetts, USA;

---