Zone Model Predictive Control and Moving Horizon Estimation for the Regulation of Blood Glucose in Critical Care Patients

Abstract Critically ill patients commonly suffer from stress hyperglycemia, or elevated glucose levels, following injury or disease. Hypoglycemia, or low glucose level, is a frequent and serious complication of treating hyperglycemia. In order to reduce the incidence of hyper- and hypoglycemia, a linear zone model-predictive controller with moving horizon state estimation and output regulation is developed. Critical care patient data from an observational study was used to construct virtual patients. Closed-loop control in these virtual patients, versus clinical standard of practice, results in a substantial increase in time spent in the target glucose zone and significant reductions in both hyperglycemia and hypoglycemia. Overall, the proposed controller significantly enhances targeted glucose control in critically ill patients in silico, which may translate to improved clinical decision making and patient outcomes in the clinic.