An Integrated Platform for Early Warning and Forecasting of Health Emergencies: the TrustAlert Project

Presentation Time: 09:45 AM - 10:00 AM

Abstract Keywords: Biosurveillance, Deep Learning, Disease Models, Information Extraction, Large Language Models (LLMs), Natural Language Processing
Primary Track: Applications
Programmatic Theme: Public Health Informatics

Responding promptly to emergencies remains challenging due to an information gap between crisis onset and management. TrustAlert addresses this issue by providing early warnings and forecasting tools. Using an AI framework based on large language models, TrustAlert extracts information from diverse data sources and maps medical vulnerabilities within communities. Additionally, simulations conducted in a novel living lab assess response strategies. These collaborative efforts aim to offer timely insights to enhance community resilience in future crises.

Speaker(s):
Emanuele Koumantakis, MD
University of Turin

Author(s):
Emanuele Koumantakis, MD - Dept. of Clinical and Biological Sciences, University of Torino, Torino, Italy; Giuseppe Rizzo, PhD - LINKS Foundation, Torino, Italy; Marco Dragoni, PhD - Foundation Bruno Kessler, Trento, Italy; Marco Beccuti, PhD - Dept. of Computer Sciences, University of Torino, Torino, Italy; Paola Berchialla, PhD - Dept. of Clinical and Biological Sciences, University of Torino, Torino, Italy;

SmartAlert: Integrating Machine Learning and Alert Triggers into Live Electronic Medical Record Systems, Targeting Low-Yield Inpatient Lab Tests

Presentation Time: 10:00 AM - 10:15 AM

Abstract Keywords: Informatics Implementation, Machine Learning, Clinical Decision Support
Primary Track: Applications
Programmatic Theme: Clinical Informatics

This study explores integrating machine learning into electronic medical record systems to predict stability of inpatient lab tests. A 'smart alerts' system was developed and tested at Stanford Hospital. The system identifies stable lab results, advising clinicians on test ordering. Live deployment showed desired precision at good recall in predicting test result stability, with suggestions for system optimization identified. This approach may significantly decrease low-yield testing and enhance personalized clinical decision-making.

Speaker(s):
Yixing Jiang, PhD student
Stanford

Author(s):
Yixing Jiang, PhD student - Stanford; Stephen Ma, MD, PhD - Stanford University School of Medicine; April Liang, MD - Stanford University; Grace Kim, BS - Stanford; Aakash Acharya, NA - Stanford Health Care; Sreedevi Mony, NA - Stanford Health Care; Soumya Punnathanam, NA - Stanford Health Care; John Makeown, NA - Stanford Health Care; Jeya Jose, PhD - Stanford; Lisa Shieh, PhD - Stanford Health Care; Tho Pham; Andrew Y. Ng, PhD - Stanford; Jonathan Chen - Stanford University Hospital;

Association of Clinical Decision Support Burden with Primary Care Burnout and Healthcare Quality: Lessons from the Veterans Health Administration

Presentation Time: 10:15 AM - 10:30 AM

Abstract Keywords: Clinical Decision Support, Healthcare Quality, Real-World Evidence Generation, Evaluation
Primary Track: Applications
Programmatic Theme: Clinical Informatics

Clinical reminders (CR) provide clinical decision support in Veterans Health Administration (VHA). However, across VHA facilities there is wide variation in the number of CRs. The increased number of CRs may lead to provider burnout or lower quality of care. We examined the variation of CRs across VA facilities and the association with burnout and diabetes outcomes. We find significant variation of CRs across VA facilities, but no association with burnout or diabetes outcomes.

Speaker(s):
Joshua Rolnick, MD
Department of Veterans Affairs

Author(s):
Eric Gunnink, MS - Department of Veterans Affairs; Karin Nelson, MD, MPH - Department of Veterans Affairs/University of Washington; Gregory Strandberg, RN, BSN, MBA/TM - Alaska VA Health Care System; Mary Lynn Ayers, MD - Department of Veterans Affairs/University of Colorado; Scott Pawlikowski, MD - Department of Veterans Affairs; Jeffrey Balsam, PharmD - Department of Veterans Affairs; Ashok Reddy, MD, MS - Department of Veterans Affairs/University of Washington;

CliniPrompt: An Accessible Auto-Prompting Website for Clinical Applications

Presentation Time: 10:30 AM - 10:45 AM

Abstract Keywords: Natural Language Processing, Large Language Models (LLMs), User-centered Design Methods
Primary Track: Applications
Programmatic Theme: Clinical Informatics

CliniPrompt is an LLM web interface that is focused on the area of prompting within the medical domain. Specifically, CliniPrompt is a prompt engine interface that allows medical professionals to view and create optimal prompts without any knowledge of prompt engineering. During prompt creation, CliniPrompt takes in a naive prompt and automatically outputs an optimized prompt based on the latest, state-of-the-art prompt evaluation techniques.

Speaker(s):
Jason Wang, B.S. in Computer Science
University of Wisconsin, Madison

Author(s):
Jason Wang, B.S. in Computer Science - University of Wisconsin, Madison; John Caskey - University of Wisconsin-Madison; Matthew Churpek, MD, MPH, PhD - University of Wisconsin-Madison; Anoop Mayampurath, PhD - University of Wisconsin - Madison; Frank Liao, PhD - University of Wisconsin, Madison - UW Health; Brian Patterson, MD MPH - University of Wisconsin-Madison; Yanjun Gao, PhD - University of Wisconsin Madison; Majid Afshar, MD, MSCR - University of Wisconsin - Madison;

The effects of providers’ fatigue on clinical decision making in an Emergency Department

Presentation Time: 10:45 AM - 11:00 AM

Abstract Keywords: Clinical Decision Support, Healthcare Quality, Patient Safety, Qualitative Methods, Documentation Burden
Primary Track: Applications
Programmatic Theme: Clinical Informatics

Emergency Department providers experience fatigue, leading to impaired performance. We qualitatively examined the effects of fatigue on clinical decision-making in a pediatric emergency department. We leveraged dual-processing theory and interviewed thirty providers. The interview questions included narrative and intuitive decision-making, the effects of fatigue on decision-making and some clinical tasks. The fatigue related factors identified in this study should be considered in the development of CDSs; so that fatigue related vulnerabilities can be minimized.

Speaker(s):
Mustafa Ozkaynak, PhD
University of Colorado-Denver | Anschutz Medical Campus

Author(s):
Trystn Daley, RN - University Of Colorado | Anschutz Medical Campus; Cristian Sarabia, MPH; Amy Yu, MD - University of Colorado | Anschutz Medical Campus; Paul Cook, MD - University of Colorado | Anschutz Medical Campus; Sarah Schmidt, MD, MSHI - University of Colorado School of Medicine;