Endogenous glucose production parameter estimation for intensive care patients

Stochastic TARgeted glucose control (STAR) is a model based tight glycemic control protocol using Intensive Control Insulin-Nutrition-Glucose (ICING) model for describing the human metabolic system. ICING model parameters - including Endogenous glucose production (EGP) - are fixed to a cohort-based constant values except the insulin sensitivity (SI) which represent the `whole body' metabolic state condition as a single parameter. Negative values of SI is physiologically not possible and are prevented during the parameter identification phase by a non-negative constraint. These cases when SI is forced to take a non-negative value potentially results in a poor blood glucose (BG) fitting in simulation of the BG trajectory of intensive care patients. In this paper, these situations are analyzed and a method is suggested for the estimation of the minimum value of the physiological EGP. Results show that using an elevated EGP was efficient at reducing BG fitting error and surpass the limitation caused by the non-negative SI constraint.