An Auto-regressive Model for an Arterial Continuous Glucose Monitoring Sensor

Continuous glucose monitoring (CGM) devices have the potential to reduce nurse workload in the ICU while improving glycaemic control (GC). However, larger point accuracy errors that are inherent in these devices can lead to adverse consequences for GC performance and safety. There is a need to characterize this error, caused by drift and sensor noise, to evaluate the impact it would have on GC when using CGM devices instead of intermittent blood glucose (BG) measurements. This paper presents an auto-regressive model to model the drift and sensor noise, which can then be used to simulate further CGM sensor traces. Validation by comparison between the original clinical data and the simulated sensor data is used to provide qualitative and quantitative assurance of model accuracy. The auto-regressive model and simulated CGM sensor traces are found to represent the Glysure CGM sensor well, and capture all necessary sensor behaviors.