Comparing Approaches to Implementing Predictive Models in Electronic Health Records: Lessons Learned from Real-World Examples

Presentation Time: 08:00 AM - 08:20 AM

Abstract Keywords: EHR Implementation and Optimization, Artificial Intelligence/Machine Learning, Adaptive Clinical Decision Support
Primary Track: AI and Care Outcomes
Programmatic Theme: Emerging Technology and Technical Infrastructure

Integrating predictive models into real-world clinical systems remains a significant challenge despite advances in machine learning and artificial intelligence. We compare four distinct implementation approaches used at Vanderbilt University Medical Center, each tailored to address different clinical needs, scalability, and workflow requirements. Attendees will gain insights into the advantages and limitations of these strategies, helping them optimize predictive model integration in electronic health records to improve patient care and operational outcomes.
Speaker(s):
Allison McCoy, PhD, ACHIP, FACMI, FAMIA
Vanderbilt University Medical Center

Author(s):
Holly Ende, MD, MSACI - Vanderbilt University Medical Center; Edward Qian - Vanderbilt University Medical Center; Colin Walsh, MD, MA - Vanderbilt University Medical Center; Dandan Liu, PhD - Vanderbilt University Medical Center; Alan Storrow, MD - Vanderbilt University Medical Center; Adam Wright, PhD - Vanderbilt University Medical Center;

Desiderata for Clinical Informatics Platforms to Support Learning Health System

Presentation Time: 08:20 AM - 08:40 AM

Abstract Keywords: Learning Health System, Educating on Self Service Analytics, Clinical informatics organizational models, Driving Digital Equity
Primary Track: AI and Care Outcomes
Programmatic Theme: Emerging Technology and Technical Infrastructure

This presentation introduces a novel conceptual framework and analytic maturity model for Learning Health System (LHS) platforms, addressing key informatics and data science challenges in implementing these systems. We will showcases a real-world example of an LHS platform at Montefiore-Einstein that unifies clinical informatics, data science, and operational analytics, demonstrating how such platforms can learn from both local and networked data sources, and how such platforms can be incrementally matured and extended to support the learning cycles of an LHS.
The proposed framework and maturity model can guide academic medical centers and CTSIs in adopting and implementing LHS platforms, potentially accelerating the translation of research into practice and improving healthcare outcomes. By inviting collaboration through open-source development, this work aims to foster a community-driven approach to advancing LHS platforms, ultimately benefiting patients, informing health policy, and improving healthcare delivery across diverse populations.
Speaker(s):
Parsa Mirhaji, MD, PhD
Albert Einstein College of Medicine

Author(s):

A Framework for Developing Metrics for AI Monitoring

Presentation Time: 08:40 AM - 09:00 AM

Abstract Keywords: Artificial Intelligence/Machine Learning, Algorithmic bias and impacts on Health Equity, Data Governance
Primary Track: AI and Care Outcomes
Programmatic Theme: Organizational Challenges

Although several frameworks for AI trustworthiness and governance are available and many health systems are starting to consider their approaches to AI monitoring, most lack mature processes for establishing critical metrics for monitoring AI implementations in healthcare. Using a combination of literature reviews and consensus, we developed the UCSF IMPACC (Impact Monitoring Platform for AI in Clinical Care) metric development framework to guide the development of AI monitoring metrics. The framework adds to previous work by adding critical domains such as value and health care professional experience, and is being practically applied to achieve rapid-cycle but still robust AI monitoring infrastructure to ensure safe and responsible deployment in healthcare.
Speaker(s):
Julia Adler-Milstein, PhD
UCSF School of Medicine

Author(s):