Sepsis is a life-threatening organ dysfunction due to a dysregulated response to infection and is the leading cause of global mortality. Bloodstream infections (BSIs) are a frequent cause of sepsis. Symptoms of BSI are nonspecific, and published guidelines do not provide clear indications for obtaining blood cultures. This leads to a low diagnostic yield, with low true positive rate, and false positives lead to unnecessary antibiotic use, increased cost, and length of hospital stay. Using large physiological data from UVA ICUs, we identified 15 features associated with BSI.
We will use the data captured in the ICUs of multiple hospitals to develop stronger predictive models of BSI using deep learning. Our hypothesis is that our algorithms will lead to significantly better predictions of BSI than current methods. This GIDI award will bring together a robust team of clinical and computer science experts to determine predictive models of BSI by using big data and deep learning. The data generated from this grant will allow us to apply for independent funding to support this work through prospective validation and clinical trials.