A “Virtual Patient” Cohort and Mathematical Model of Glucose Dynamics in Critical Care

Stress hyperglycemia is common in critically ill patients and is strongly correlated with increased patient morbidity and mortality. Tight glucose control has been studied as a route to improving patient outcomes by attempting to maintain euglycemia in critical care patients. Unfortunately, insulin sensitivity variability associated with trauma or stress coupled with tight glucose control may also lead to significant hypoglycemia, which has been shown to be strongly correlated with patient mortality. To combat stress hyperglycemia, while taking care to avoid hypoglycemia, a number of systems including closed-loop control with continuous glucose monitoring have been proposed. We synthesize a mathematical model describing a virtual patient cohort, using clinical data, as a means to test these types of algorithms in silico. Virtual patients are primarily characterized by time-varying insulin sensitivity and pancreatic insulin secretion profiles and exhibit trajectories consistent with physiological and clinical expectations. Overall, two patient groups result: (i) good sensitivity to insulin and stable insulin sensitivity trajectories (presumably returned to a healthy state); and (ii) depressed, variable sensitivity.