AI May Improve Clinician Decision-Making Within ICUs

In collaboration with the University of Pittsburgh and UPMC, researchers from Carnegie Mellon University’s Human-Computer Interaction Institute (HCII) found that an artificial intelligence (AI)-based tool assisted intensive care unit (ICU) clinicians in providing treatment for sepsis.

According to the Centers for Disease Control and Prevention (CDC), about 1.7 million US adults develop sepsis annually, at least 350,000 of whom die in the hospital or wind up in hospice.

To gain insight into whether merging AI and treatment practices for this condition would be effective, researchers from the HCII at Carnegie Mellon University provided 24 ICU physicians from UPMC with access to a novel tool intended to assist with decision-making.

Trained on a set of over 18,000 patients who reached sepsis diagnostic standards during an ICU stay, the tool aims to assist clinicians in filtering and searching for patients in a data set. It is also designed to predict how their condition will progress and enable comparisons of model predictions with the treatment decision ultimately taken.

Known as the AI Clinician Explorer, the interactive clinical decision support (CDS) interface was developed by the research team in 2018 to assist with sepsis treatment.

"Clinicians are always entering a lot of data about the patients they see into these computer systems and electronic health records," said Venkatesh Sivaraman, a PhD student in the HCII and member of the research team, in a press release. "The idea is that maybe we can learn from some of that data so we can try to speed up some of their processes, make their lives a little bit easier and also maybe improve the consistency of care."

The application of the AI tool consisted of a think-aloud study encompassing 24 clinicians, all of whom worked in the ICU and had experience with sepsis treatment. During the study, they all used the AI Clinician Explorer interface to analyze treatment decisions for four separate patient cases.

From their research, the team found that clinicians primarily displayed four behaviors: ignore, rely, consider, and negotiate. The ignore and rely groups were opposites, with the former meaning that members did not make decisions by considering AI recommendations and the latter meaning that members took some or all of the AI notes into account. The consider group included members who processed AI recommendations and made decisions regarding their application from there. The negotiate group, where most participants fell, included practitioners who used certain aspects of AI recommendations in specific decisions.

Although the study's results surprised the research team, they noted that their efforts led to further insight into how to improve the AI tool. Rather than using AI to replace clinician efforts, researchers aim to determine clinician decision patterns that may have gone undetected in previous trials and determine necessary steps for the future.

"There are a lot of diseases, like sepsis, that might present very differently for each patient, and the best course of action might be different depending on that," said Sivaraman. "It's impossible for any one human to amass all that knowledge to know how to do things best in every situation. So maybe AI can nudge them in a direction they hadn't considered or help validate what they consider the best course of action."

Using AI to enhance clinical decision-making in the ICU is growing more common.

In October 2022, Penn State Health began working with CLEW Medical to deploy a virtual intensive care unit (vICU) powered by an AI cloud-based platform.

Largely driven by the goal of reducing physician burnout, Penn State Health's vICU program is supported by a cloud-based tele-ICU platform from CLEW, which supplies AI analytics capabilities to assist providers in clinical decision-making in an effort to lower costs and enhance the healthcare experience.