Emergency room patients who were flagged by an artificial intelligence algorithm for possibly having sepsis received antibiotics sooner and had better outcomes, according to a peer-reviewed study published in Critical Care Medicine.
Over five months in 2019, the study’s authors tracked nearly 600 patients who came into the emergency department. MetroHealth implemented an electronic health record-embedded early warning system for sepsis.
Patients 18 and older presenting to the emergency department were randomized to standard care for sepsis versus the pathway augmented by the early warning system.
The early warning system alerted both the physicians and pharmacists. This resulted in the patient who was flagged receiving antibiotics significantly faster than those patients whose alert was hidden, according to the study.
Collectively, those who received early antibiotics were measured to have more days alive and out of the hospital more than those in the standard care group.
“We showed that when providers had access to the early warning system, patients had better sepsis-related outcomes,” said Yasir Tarabichi, an assistant professor of medicine at the Case Western Reserve School of Medicine and the study’s principal investigator. “These patients got their antibiotics faster and had, on average, more days ‘alive and out of hospital’ than the group that had usual care. Taken together, the increase in survival rates and reduction in hospital stay improved with the implementation of the early warning system.”
“This study adds to the recent national discourse about sepsis early warning systems,” Tarabichi said. “Recent studies assessed how that score worked in isolation, which is not reflective of how it would actually be used in the real world. We envisioned the early warning system’s role as supportive to our health care team’s response to sepsis. Most importantly, we assessed the utility of the tool with the highest quality approach—a randomized controlled study. In fact, our work stands out as the first published randomized controlled evaluation of a model-based early warning system in the emergency room setting.”