FEBRUARY 6, 2025
Originally published by our sister publication Anesthesiology News
CLEVELAND—Incorporating technological advancements into perioperative care may be daunting for some, but it’s a leap that may well improve the physician–patient relationship and offer clinicians time to focus on their patients in ways they have been unable to do for years. In a recent presentation, Heather L. Evans, MD, discussed the several potential benefits of tech-enabled perioperative care, and what that
Originally published by our sister publication Anesthesiology News
CLEVELAND—Incorporating technological advancements into perioperative care may be daunting for some, but it’s a leap that may well improve the physician–patient relationship and offer clinicians time to focus on their patients in ways they have been unable to do for years. In a recent presentation, Heather L. Evans, MD, discussed the several potential benefits of tech-enabled perioperative care, and what that might look like for surgeons in the future.
“I am not a data scientist,” began Evans, the chief of surgery at the Ralph H. Johnson VA Medical Center, in Charleston, S.C. “But I am a surgeon who’s been applying informatics to try and come up with solutions for problems that we’ve had for some time. And I’m very interested in the postoperative space, and what happens after patients go home.”
For Evans, one of the earliest forays into this arena occurred in 2016, when she and a team of researchers developed a prognostic model of surgical site infection (SSI) using daily clinical wound assessment (J Am Coll Surg 2016;223[2]:259-270.e2). The study found that while serial features provided moderate positive predictive value and high negative predictive value for early identification of SSIs, the addition of baseline risk factors did not improve identification.
“More importantly, we showed that even four days before the day of diagnosis there was a signal that these patient populations were diverging into those that would and would not have SSIs,” Evans explained at the 2024 annual meeting of the Society of American Gastrointestinal and Endoscopic Surgeons. Perhaps more importantly, the study demonstrated the power of using machine learning to help improve patient care, a realization that spurred the investigators to develop mPOWEr (mobile postoperative wound evaluator), a patient-centered app designed to track surgical wounds at home (www.mpowercare.org).
As effective as that tool has proven over the ensuing years, Evans said the exercise proved illustrative for other reasons.
“With mPOWEr we created the tool for those people that were having problems, not everyone,” she said. “But the truth is that all patients want to know if they’re going to be OK; all patients want you to contact them. But we have a workflow conflict: We don’t prioritize this type of communication because we simply don’t have time.”
The lack of such communication and interaction robs both provider and patient with the connection that fosters our collective humanity, the exact reason that many physicians go into practice in the first place. This is where artificial intelligence may play an important role.
As demonstrated in a 2024 paper, ambient AI scribes may relieve surgeons of the burden of clinical documentation, and other data entry during patient encounters apply machine learning to conversations to facilitate scribe-like capabilities in real time (NEJM Catal Innov Care Deliv 2024;5[3]. doi:10.1056/CAT.24.0074). Although still in the early stages of development, the AI scribes were found to be acceptable to both clinicians and their patients, and largely improved the experience for both parties.
“What the technology gave these doctors back was time,” Evans said. “Time looking the patient in the eye, time to answer questions, time to pick up on all those nonverbal cues that define the interaction.”
Despite challenges, Evans believes that AI and large language models can ultimately prove to be effective adjunctive tools in the physician’s armamentarium—tools that not only offer busy doctors the gift of time but also a window into improving patient care through a renewed focus on empathy and the physician–patient relationship.
Session co-moderator Daniel A. Hashimoto, MD, MTR, an assistant professor of surgery, computer and information science at the University of Pennsylvania, in Philadelphia, agreed that such models have the potential to improve perioperative care, but urged caution in their use.
“These technologies are allowing us to change the way we interact with data in ways that are much more fluid and natural than ever before,” Hashimoto said. “However, I think we have to keep in mind how this current generation of large language models works, which is based exclusively on statistical probability. So, they’re only as good as the data on which they’re trained. If we are using these models to solve problems around issues that are underrepresented, it raises the risk of more dangerous outcomes and misinterpretation of how we should be caring for patients.”
By Michael Vlessides
Evans and Hashimoto reported no relevant financial disclosures.
