Machine learning model designed to measure sepsis risk

An AI-based sepsis screening tool could better help physicians find patients most at risk of developing the illness after it outperformed other traditional screening methods, according to a study published in the Annals of Emergency Medicine.

Several healthcare organizations in the United States and United Kingdom have turned to AI solutions to address sepsis in patients and prevent fatal outcomes. Last November, Massachusetts researchers developed an AI system that predicted when critical drugs should be administered to sepsis patients.

For this study, a research team used machine learning to develop and test a sepsis screening tool—called the Risk of Sepsis Score (RoS)—with electronic health record data from emergency room (ER) patients at 49 different urban community hospitals. The data featured about 2.7 million records.

The screening method was analyzed one, three, six, 12 and 24 hours after receiving the data. Additionally, its performance was rated with various metrics—alert rate, area under the receiver operating characteristic curve (AUC), sensitivity, specificity and precision. According to the study, the RoS was the “most discriminant screening tool” when identifying most at-risk patients, and was significantly more sensitive and precise than the next top performing sepsis screening method.

“In this retrospective study, RoS was more timely and discriminant than benchmark screening tools...further study is needed to validate the RoS score at independent sites,” Ryan J. Delahanty, PhD, data scientist at Tenet Healthcare, et al. concluded.