Illustration: Aïda Amer/Axios
A new AI tool developed by Johns Hopkins University helped detect early signs of sepsis in real time, reducing the likelihood of death by 20%, according to new research published in Nature Medicine.
Why it matters: Sepsis caused by the body's extreme response to an infection can be easy to miss and accounts for the most in-hospital deaths each year in the United States.
Details: The Targeted Real-time Early Warning System monitored more than 590,000 participants at five hospitals via their electronic health records starting in 2018.
- Researchers evaluated 6,877 patients that the system flagged based on clinical notes and other patient data indicating a life-threatening condition.
- In sepsis cases that were confirmed within three hours, in-hospital mortality, organ failure and length of hospital stay all declined, especially in patients who were high-risk.
Yes, but: The AI had an accuracy rate of 38%, an accompanying study found.
- That was still an improvement over earlier sepsis-warning systems, which had accuracy rates of about 12%, the study says.
- "This is a breakthrough in many ways," Albert Wu, the study author, said in a news release. “Up to this point, most of these types of systems have guessed wrong much more often than they get it right. Those false alarms undermine confidence."
The bottom line: The new tool shows the possibility of harnessing machine learning to identify one of hospitals' biggest killers.