ICU clinicians turn to artificial intelligence to prioritize patients
Researchers said they have developed a novel tool that uses artificial intelligence to help ICU physicians ascertain if a patient does not require an intervention and intense monitoring — allowing them to focus instead on more critical or unstable patients.
“Many times, a critical care physician is faced with multiple things happening at once,” Itai Pessach, MD, PhD, of the pediatric ICU at Sheba Medical Center in Israel, told Healio Primary Care. “We put a lot of time and effort into trying to prioritize in terms of what is going on with the patient right now, how we can intervene earlier and treat complications before they happen.”
Pessach said the tool, provided by CLEW Medical, uses data from tens of thousands of previous cases to make predictions about patients in real time, letting physicians know up to 14 hours in advance if someone in the ICU requires two key medical interventions — intubation and BP medications.
“This has not been done before,” he said.
Pessach and colleagues tested CLEW’s performance in three patient cohorts, with a total of 53,071 hospital stays. According to the researchers, the model correctly identified that more than 60% of the patients did not need an intervention or advanced provider attention in the next 14 hours. The model had a sensitivity of 0.63 in predicting stable periods, with a precision of 0.993 (area under the curve = 0.83). This corresponded to 2,951 work shifts of 12 hours that were correctly identified as not needing an intervention during the following 14 hours, and 1,659 stays correctly predicted to require no significant intervention during the entire ICU stay.
“This [model helps] physicians understand what's going to happen hours before that something happens,” Gal Salomon, CLEW CEO and founder, told Healio Primary Care. “For the first time, ICU physicians can be proactive and improve the outcome of their patients.” – by Janel Miller
Pessach I, et al. Focusing advanced clinicians on the more critically ill patients using artificial intelligence. Presented at: Society of Critical Care Medicine’s Critical Care Congress; Feb. 16-19, 2020; Orlando.
Disclosures: Pessach is a consultant to CLEW. Salomon is CEO and founder of CLEW.