Grant: Project Grant

Sepsis, is a life-threatening inappropriate response to infection that may lead to organ failure and death. Early treatment may reduce human suffering and staggering costs. The problem is there are no biomarkers in clinical use that may predict early, which patients are at risk for serious deterioration. In collaboration with the National Research Council of Canada, we developed a diagnostics point-of-care device called PREDICT-PowerBlade (PREDICT-PB). The PowerBlade uses microfluidic technology. The team used machine learning and AI to identify and test a new gene-based signature of sepsis deterioration and adapted it for detection on the PowerBlade. In patients with suspected sepsis, the PREDICT-PB is capable of predicting clinical deterioration 24-48 hrs after presentation with an overall accuracy of 93%. We benchmarked results using 'gold standard' techniques and performed analytical validation of the PREDICT-PB device using patient samples. PREDICT-PB performs gene extraction from 50 microliters of whole blood and links it to gene detection making results available within 3 hrs, compared to 1-3 days by a conventional lab. Here we propose the first prospective study to test the PREDICT-PB in the 'real-world' patient environment. We will recruit 100 patients from the emergency and the intensive care unit with suspected sepsis and 50 controls, to evaluate, compare and combine results from the PREDICT-PB with clinical and laboratory measures commonly used. We defined deteriorations as any of the following: ICU admission, initiation of respiratory support (non-invasive ventilatory support), endotracheal intubation, administration of vasopressors/inotropes, or death within the first 24 to 72 hours of presentation. We will study the ability of PREDICT-PB to make accurate predictions, determine if by adding other clinical data we can improve these predictions and the feasibility of using the device at the bedside. Success may fast-track deployment to the clinic.