A new AI system makes patients 20% less likely to die of sepsis, according to new research.
The AI technology is called Targeted Real-time Early Warning System (TREWS), and is a new sepsis early intervention tool. The technology uses machine-learning to look at medical records and clinical notes to identify patients who are at a high risk of developing life-threatening complications.
Sepsis can be easy to miss because the early symptoms, including fever and confusion, are so common amongst other conditions. The AI will correlate these symptoms with the other risk-factors and alert doctors to the risk factor of the patient. Identifying sepsis early could also prevent patients who do survive from developing issues such as multiple-organ failure.
“It is the first instance where AI is implemented at the bedside, used by thousands of providers, and where we’re seeing lives saved,” says Suchi Saria, founding research director of the Malone Center for Engineering in Healthcare at Johns Hopkins University, and lead author of the studies
Treating sepsis can also be difficult due to people’s immune systems acting differently, making a course of treatment harder to determine.
“One of the main reasons that finding effective therapeutics to treat sepsis has been so challenging is due to this variation among patients.” Scott Brakenridge, MD, first author of the study and trauma surgeon at the University of Washington said in a press release.
The AI tracks a patient from their arrival at hospital to discharge. Five hospitals were involved in the trial, and 4,000 clinicians used the AI to treat 590,000 patients. 179,931 past cases were also reviewed.
In 82% of sepsis cases, the AI was accurate almost 40% of the time. This is a huge improvement on past attempts, where electronic tools were used and caught less than half that many cases with between 2 to 5% accuracy. The TREWS AI on average caught sepsis six hours earlier than traditional methods.
There is also the hope that the AI will be able to help in medical settings with less developed care, who lack experts in specific fields.
Suchi Saria, Associate Professor of Machine Learning and Healthcare, from Johns Hopkins and lead author of the studies says:
“This is an extraordinary leap that will save thousands of sepsis patients annually. And the approach is now being applied to improve outcomes in other important problem areas beyond sepsis.”