Prediction of Bacteraemia and of 30-day Mortality Among Patients with Suspected Infection using a CPN Model of Systemic Inflammation

Abstract Prediction of both the likelihood of bacteraemia and of death within 30 days allows for prudent decisions to be made regarding the diagnostic workup and therapy of patients with suspected sepsis. In this paper, we combine two predictive models and perform machine learning to tune the new model’s ability to predict both bacteraemia and 30-day mortality. The model was then validated on three independent datasets. There was no difference in the discriminatory ability of the model compared to each of the predecessors. For bacteraemia prediction, the new model had an AUC = 0.71 for the training data, and AUC = 0.73, 0.74 and 0.79 for the validation data. For mortality prediction, the model had an AUC = 0.81 for the training data and AUC = 0.76, 0.84 and 0.80 for the validation data.