CLIF Meets RL: A Hands-On Workshop on Training Reinforcement Learning Models with Common Longitudinal ICU Format (CLIF) Data

Presentation Type: Workshop - Instructional
Not recorded for AMIA Now

Presentation Time: 10:30 AM - 12:30 PM

Primary Track: Data Science/Artificial Intelligence


ICU teams make complex decisions every day, yet the data needed to study and improve these decisions are often recorded differently from hospital to hospital. This lack of consistency makes it difficult to compare outcomes, share findings, or build tools that can support clinicians in real time. The Common Longitudinal ICU Format (CLIF) was created to solve this problem by providing a simple, shared structure for representing key elements of ICU care—vital signs, treatments, ventilator settings, medications, and patient trajectories—across many hospitals. Today, CLIF supports multi-center studies involving more than 800,000 patients and allows health systems with very different EHR systems to contribute to the same research questions without sharing identifiable data.

This workshop introduces clinicians and researchers to CLIF and demonstrates how standardized ICU data can support new ways of understanding and improving care. Using examples from a recent multi-site study of ICU readmissions, we show how CLIF makes it possible to compare patient outcomes and care patterns across institutions. We then guide participants through a hands-on exercise using CLIF data, where they will explore how reinforcement learning (RL) learns from clinical decisions over time. Participants will learn how states, actions, and rewards (elements for MDPs) can be curated from CLIF and how an RL model can be trained to evaluate clinical decision strategies. By the end of the session, attendees will understand how CLIF supports reproducible ICU research and how AI methods like RL can be used to study and potentially improve critical care decisions.

Speaker(s):
Yikuan Li, Ph.D.
George Mason University

Kaveri Chhikara, Senior Data Scientist
University of Chicago

Saki Amagai, Student
Northwestern University

Author(s):
Kaveri Chhikara, Senior Data Scientist - University of Chicago; Saki Amagai, Student - Northwestern University; Yikuan Li, Ph.D. - George Mason University;

---