In silico assessment of a computerized model-based glycaemic control approach in a Belgian medical intensive care unit

Abstract

Glycaemic control can be used to enhance critically ill patient outcome. This paper presents the in-silico design of a computerized model-based controller for a Belgian medical intensive care unit (CHU of Liege, Belgium). In silico trials are used to assess the current clinical protocol efficiency and safety and to compare this protocol with the existing Stochastic Targeted (STAR) control approach. The objective of this research is to optimize a glycaemic controller for its future clinical implementation and clinical workflow requirements. Results suggest that the currently used, paper-based sliding scale protocol is too general to achieve safe and effective glycaemic control. The computerized model-based protocol STAR leads to better glycaemic outcomes associated with increased safety. In particular, time in target band is higher than 80% with STAR targeting 90-150 mg/dL and 100-160 mg/dL. Time in the desired 100-150 mg/dL band is improved using STAR, and BG < 80 mg/dL is reduced. Results suggest that control targeting 100-160 mg/dL is associated with increased time in band and increased safety.