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

EP16 Fei Wang, PhD, Weill Cornell Medical College

·58 min·3 clips

If you knew six hours ahead that the waiting room would hit 40 patients, what would you change?

DJ Abhakama and Ethan Abbott talk with a guest about AI in emergency medicine. The discussion begins with a straightforward question: if models are getting better at pattern recognition, what actually changes on a busy shift, and how do you make those outputs clear enough for clinicians who need to move fast? Visuals matter. The guest notes that numbers are not always intuitive, so the way predictions are presented can matter as much as the prediction itself when the audience is human and the stakes are clinical. Patient flow comes up early. The group talks through the earlier goal of finding broad movement patterns in how patients travel through care, then moves toward a more frequent predictive layer that could help plan resources before the department gets overwhelmed. That is the promise. The conversation keeps that promise grounded by asking how to make AI useful without adding work or turning it into yet another layer of complexity. Sepsis is the next turn. DJ and Ethan ask about research on clinical deterioration and sepsis trajectories, including the idea that sepsis is heterogeneous and that different subtypes may point to different progressions or treatment paths. A single baseline snapshot is not enough. The guest explains that much of the earlier work focused on a specific time point such as ICU admission, while the broader framing here is more interested in identifying subtypes and tracking what those patterns might imply for progression and treatment. Alert fatigue is the hard part. They press on decision support and ask how a system could surface the right information for clinicians without filling the screen with boxes that get ignored, and the answer stays practical rather than magical. That balance matters. The episode keeps coming back to the same implementation questions: what can actually be used, what fits the workflow, and how to make predictions understandable to humans. The tone stays curious. Instead of selling AI as a fix-all, the hosts and guest treat it as a tool that could help if it is tuned to the realities of emergency care and clinical communication. The result is clear. This is a conversation for listeners who want a clinician-facing look at predictive modeling, not a victory lap for technology.

As heard by us

A grounded look at AI in emergency medicine, with a focus on implementation and clinical stakes.

Episode 16 works best when AI is treated as a practical tool for emergency medicine, not a slogan. DJ Abhakama and Ethan Abbott keep the focus on what clinicians actually have to deal with, from patient-flow prediction and resource planning to sepsis subtypes and clinical…

Read the full review in PlayNext →

Why you'd press play

You hear how AI could help predict emergency department flow and support clinicians in emergency medicine.

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