Dr. Nathan Moore, medical director of the BJC Healthcare ACO
Photo: Dr. Nathan Moore
Multi-AI agents – multiple artificial intelligence agents working together in a shared environment – can be used to address persistent workflow challenges in clinical decision support, drawing from real-world experience to improve a clinical decision support system for end-of-life care planning.
At the 2026 HIMSS Global Health Conference & Exposition in March, Dr. Nathan Moore, medical director of the BJC Healthcare ACO, will demonstrate how large language model-enhanced, self-reflecting and tool-enhanced agents can automate high-risk patient selection, reduce manual workload and improve care coordination. The session is titled "Multi-Artificial Intelligence Agents for Enhancing Clinical Decision Support."
Moving beyond chatbot use cases
"The session will focus on how multi-agent, LLM-enhanced systems can transform clinical decision support, using end-of-life and advance care planning workflows as a concrete, real-world case," Moore explained.
"This is particularly important for HIMSS26 attendees because many organizations are struggling with how to move beyond simple chatbot use cases toward safely deploying AI that actually takes action in complex workflows – pulling data, triaging patients and nudging clinicians at the right time.
"By grounding the discussion in a mature clinical decision support program that targets patients at high risk of mortality and supports timely advance care planning conversations, the session speaks directly to current pressures around value-based care, workforce burnout and aligning care with patient preferences," he added.
The central example in the session will be BJC Healthcare's "learning reviewer agent," which augments an existing mortality risk model and advance care planning outreach workflow.
"Historically, a deep learning mortality model generated risk scores that then went through a highly manual human review process, where retired clinicians read charts and applied their clinical intuition to decide who was actually appropriate for advance care planning outreach – a process that was accurate but slow, costly and not continuously learning," Moore said.
"The learning reviewer agent builds on several years of labeled data – more than 35,000 patients with risk scores, clinical notes and prior reviewer decisions – and uses supervised learning plus live reinforcement from clinician feedback to mimic and then gradually enhance those reviewer decisions," he continued.
Continuous learning
There will be various takeaways for HIMSS26 attendees from this session. One key takeaway will be a concrete framework for "memory-augmenting" agentic clinical decision support systems so they continuously learn from their own deployment rather than remaining static point systems, Moore revealed.
"The session will walk through how to design agents that store and reuse institutional memory to refine future recommendations and better fit local practice patterns and workflows," he said.
"Attendees will leave with a practical mental model for structuring agent roles, feedback loops and evaluation stages – historical testing, 'silent mode' and post-deployment monitoring – they can adapt to their own use cases, whether in advance care planning, chronic disease management or other high-stakes decision support domains," he concluded.
Dr. Nathan Moore's session, "Multi-Artificial Intelligence Agents for Enhancing Clinical Decision Support," is scheduled for Tuesday, March 10, from 10:15-10:45 a.m. in Palazzo D Level 5 at HIMSS26 in Las Vegas.
Follow Bill's health IT coverage on LinkedIn: Bill Siwicki
Email him: bsiwicki@himss.org
Healthcare IT News is a HIMSS Media publication.
WATCH NOW: Managing government policy and regulation in healthcare AI adoption
