Computer Generated Predictive Model for Sepsis Catches Diagnosis 24 hrs before Clinicians Do

A recently published article titled A targeted real-time early warning score (TREWScore) for septic shock   in Science Translational Medicine by Henry et al team at Hopkins describes a machine learning interface which predicts development of sepsis in ICU patients using variables that are collected directly from the electronic health records. The model has a ROC of 0.82, which is outstanding. This is really impressive because the program generates data directly from Mary lost extract it from the electronic health records and then crunches that data and provides a dynamic alert to the intensivist. Howard Hospital will serve as the pilot site for the program in the real world, prospective settings.  If validated, any hospitals with electronic medical records can implement software for clinical practice.