In this study Dascena validates a sepsis prediction algorithm using Sepsis-3 definitions. The findings are published in JMIR Medical Informatics.
Despite many years of research and development, sepsis prediction remains a serious problem because sepsis manifestation can be unclear until late stages. InSight, a machine learning algorithm, was developed precisely to take on this challenge. The algorithm uses simple, readily available patient data from the electronic health record (EHR) to perform its calculations. InSight is shown to be superior in performance to the Simplified Acute Physiology Score (SAPS II) and Sepsis-Related Organ Failure Assessment (SOFA) scores calculated upon admission and comparable to other machine learning methods without the laboratory values that these methods require. InSight also outperforms quickSOFA (qSOFA) and Systemic Inflammatory Response Syndrome (SIRS) scores using similar vital sign measurements, demonstrating its ready application in EHR-integrated environments. InSight’s performance was shown to be resistant to significant missing data.