TREWS for Early Sepsis Warning
A machine-learning model alerts hospital teams when patient data points to possible sepsis.
About
The Targeted Real-time Early Warning System, TREWS, analyzes electronic patient data in Johns Hopkins hospitals. The model looks for patterns that may indicate sepsis and shows a warning to the treatment team. In a prospective, observational study, rapid assessment of such a warning was linked to better outcomes. This link does not by itself prove that the algorithm was the cause, and many of the warnings require human review.
AI factor
Sepsis arises from a combination of changes in vital signs, lab results, and medical history. Machine learning can continuously monitor this stream and flag a possible risk; clinicians assess the patient and decide on testing and treatment.
Sources
Inspiration sourced from AI For Good. Not an Arnaldus customer case.