How Is AI Monitoring Inpatient Risk and Reducing Documentation Burden?

In this section of The AI Oncology Revolution, Matthew Matasar, MD, pointed Arturo Loaiza-Bonilla, MD, MSEd, FACP, to a Robert Wood Johnson University Hospital study published in JAMA, which used an AI-embedded deterioration index within Epic to flag hospitalized patients at risk of clinical decline. Then, Loaiza-Bonilla shared his own experience with Epic’s patient-facing chatbot fielding a question about a PET scan result, prompting a broader discussion of where clinicians are regarding their comfort in letting AI communicate directly with patients. Matasar was candid that he’s not yet comfortable with the chatbot’s default voice in his own practice, even as he acknowledged AI has already quietly taken over routine documentation work like letters of medical necessity and return-to-work letters.

Loaiza-Bonilla is systemwide chief of hematology and oncology at St. Luke’s University Health Network and co-founder of Massive Bio. Matasar is chief of the Division of Blood Disorders at Rutgers Cancer Institute/Jack & Sheryl Morris Cancer Center and professor of medicine at Rutgers Robert Wood Johnson Medical School, where his clinical and research interests focus on lymphoma.

Transcript

Matasar: Outside of my own field within blood disorders, Robert Wood Johnson University Hospital recently published in JAMA a study using an AI-embedded model within Epic—not focused specifically on oncology, but on internal medicine—called a deterioration index. It draws from individual patient data points within Epic and applies a sophisticated…model to identify which patients are at risk for deteriorating, flag those patients, and initiate interventions to try to preempt further clinical decline. Against historical controls, that work demonstrated the intervention is extremely successful at improving and enhancing patient outcomes in the inpatient hospital setting.

Loaiza-Bonilla: Those are really good use cases we can demonstrate now. Just out of curiosity—I also have Epic in my health system, and there’s a little chatbot called Art, which is like “Arturo,” like me—sometimes patients send questions into their inbox, and personally, this has become really helpful. For example, there was a question from a patient asking, essentially, how their PET scan looked, and whether they should do something about their surgery. I knew the PET scan was there, but the chatbot was able to pull out the latest PET scan and give an initial answer, which I could then modify and send out. Have you seen any value in that, to make our lives easier, when responding to patients?

Matasar: I’ll admit that I’m not a fan of their default answer prompts. I’m very fussy about how I answer my patients. For better or worse, I feel like I have a specific voice when I’m communicating digitally with my patients, and I haven’t been comfortable using the voice of that system in their responses. I know colleagues look at it differently, and I think it’s going to get better. Certainly, we use AI for helping with documentation all the time––I haven’t written a letter of medical necessity in longer than I can tell you; that’s purely AI at this point. Return-to-work letters, too. Some of these daily chores that burden our teams, as much as they burden us, are so easily and fluidly done through AI approaches.