Modeling PICU Journeys Through the AI-Simulated Baymax Approach to Optimize Care Models

Document Type

Conference Proceeding

Publication Date

3-2026

Publication Title

Critical Care Medicine

Abstract

Introduction: Precision medicine tools (genomes, transcriptomes, metabolomes) are transforming acute PICU care and revealing origins of chronic disease. While these omic platforms generate rich molecular hypotheses, they lack context on how acute interventions map to long-term risks. The Corewell Health PICU leverages omic discoveries, but we need frameworks that weave in patient, caregiver, and clinician perspectives to guide hypothesis testing of the omics.

Methods: We developed Baymax (Behavioral and ArtificiallY-Modeled Analysis eXchange), a hybrid AI framework to model acute-to-chronic transitions. Using GPT-4o, Baymax simulated 100 journeys spanning pre-admission, acute intervention, and 1–5-year follow-up. Each journey seeds key clinical anchors into clinician forums, guideline repositories (ELSO, HLH-94, ACR MIS-C), and literature, while capturing patient, caregiver, and clinician voices via Reddit, Facebook, PatientsLikeMe, TeensHealth, MyHealthTeams, Stack Exchange, Sermo, Medscape Consult, and Slack/Discord channels. Logical pathflows integrate these components into coherent, human-centered narratives.

Results: In a pilot of 100 journeys reflecting ~1 month of a Level I PICU, we modeled 15 sepsis/infection, 20 respiratory (ARDS, RSV), 15 cardiac, 15 neurological, 10 trauma, and 10 other (metabolic, allergic) cases. Rapid, point-of-care testing drove favorable outcomes in sepsis and trauma stories. HLH, MIS-C, and Kawasaki cases benefited from prompt immunotherapies. Narratives underscored the need for long-term surveillance, including pulmonary function, echocardiography, neuropsychological assessments, and renal ultrasonography. Identified chronic conditions included asthma, HLH, cystic fibrosis, and epilepsy. Recurring themes for caregivers and clinicians were guilt, fear, anxiety, and helplessness.

Conclusions: Baymax enables discovery of acute-to-chronic disease transitions, informs caregiver support and clinician decision-making, and guides system-level optimizations. Adapting these methods across diverse geographic and socioeconomic contexts may yield tailored PICU guidance and more accurate hospital-level modeling.

Volume

54

Issue

3 Suppl

Comments

Society of Critical Care Medicine Critical Care Congress, March 22-24, 2026, Chicago, IL

DOI

10.1097/01.ccm.0001186252.37229.1b

ISSN

1530-0293

Link to Full Text

Share

COinS