STAT AI: An Emergency Medicine Podcast · Ethan Abbott and DJ Apakama

EP15 Andrew Taylor, MD, MHS University of Virginia Emergency Medicine

·49 min·2 clips

The model learned from 1.8 million patients using positive unlabeled learning.

DJ Abhakama and Ethan Abbott open the episode by framing AI as something that should actually help emergency medicine, not just sound impressive on paper. A guest then walks through a live implementation of TriageGo, an AI triage product used across three emergency sites in the Yale New Haven Health System. The rollout was staggered, which the guest describes as a stepped-wedge design. That mattered because the tool required real implementation work, especially around nursing and getting end users comfortable with a product that changes their workflow. Triage in emergency medicine already balances acuity and resource allocation. TriageGo takes a more direct route by predicting downstream outcomes. Those outcomes include ICU admission within 24 hours, emergency surgery within 24 hours, and hospital admission, with admission to the hospital being the main outcome the model is built around. The guest notes that the product could have shown raw probabilities to the user. Instead, it translates those predictions back into the one-through-five score nurses already know. That translation is a major usability advantage, because it bridges model output and front-line decision-making. The conversation keeps returning to what it means to test an AI tool in an actual clinical setting rather than in synthetic examples or vignettes. The episode stays focused on implementation details because those details determine whether a tool fits nursing workflow, earns buy-in, and can be used in practice. The tone stays practical and curious rather than promotional, with the real question being how software behaves when it reaches the intake desk.

As heard by us

Real ED AI talk, grounded in workflow rather than hype.

The episode keeps its attention on emergency-department workflow instead of AI hype. Its strongest section follows a live implementation of TriageGo at Yale New Haven Health, with a stepped-wedge design across three emergency sites and nurses as the main users.

Read the full review in PlayNext →

Why you'd press play

If you want to see AI meet real triage, start here.

Read the full recommendation in PlayNext →

Listen to the show on

Apple Podcasts Spotify

Play clips from this episode More from STAT AI: An Emergency Medicine Podcast