Sepsis is a killer; the CDC reports that each year at least 1.7 million adults in the U.S. develop sepsis, and one in three patients who die in a hospital have sepsis. Assigning the cause of death to sepsis is tricky because it’s not a disease nor an infectious agent. Sepsis occurs when a person’s body responds to an unchecked infection. A patient admitted to a hospital for a cardiac event, cancer, or any number of conditions can develop sepsis and die quickly if the grave condition isn’t detected early. According to published guidelines, early sepsis identification and treatment within hours of the condition’s onset significantly improve patient outcomes. Without rapid treatment, the condition can result in organ failure and death.
We’ve written before about artificial intelligence algorithms developed to detect sepsis, including work at the University of California San Francisco and Turku PET Centre, Finland. Researchers from Augusta Health in Virginia published a study in Nursing Care Quality earlier this month. The study documents their implementation of an automated sepsis screening tool in a community hospital. The Augusta Health team tested various artificial intelligence models based on six variables. The test design used 10,792 hospital admission records, including 339 cases of sepsis, to train the algorithms.
The Augusta Health team’s final model correctly screened 85.7% of the records in the test batch. After their research, Augusta Health used the AI model to screen 100% of admitted patients to send reports. The screening system analyzes data for all hospital inpatients every hour, proactively looking for clinical data outside normal ranges. When the model detects a combination of variables that fits the sepsis screening profile, the system alerts the patient’s care team for immediate response.
Augusta Health began testing AI sepsis detection models in the hospital emergency department in 2016. The refined model now screens all patients. The next steps are to expand the implementation of the AI sepsis early detection model in other hospitals.
