Giannini HM, Ginestra JC, Chivers C, et al. - Using a retrospective cohort design, researchers performed this inquiry at a tertiary teaching hospital system in Philadelphia, PA, to construct as well as implement a machine learning algorithm for the prediction of severe sepsis and septic shock. They also focused on the influence on clinical practice and patient outcomes. Using electronic health record data of all non-ICU admissions, they developed and validated a random-forest classifier, which was used both silently and later with an alert to notify clinical teams of sepsis prediction. Sensitivity and specificity of 26% and 98%, respectively, and a positive predictive value of 29% and a positive likelihood ratio of 13 were yielded by the algorithm. Overall, the impending occurrence of severe sepsis and septic shock can be predicted by this machine learning algorithm, with low sensitivity but high specificity. A modest influence of algorithm-generated predictive alerts on clinical measures was also observed.