In Silico Control of Blood Glucose in Intensive Care Patients

Abstract

GlucoSet AS has developed an intravenous continuous sensor for blood glucose measurement in intensive care units and wishes to test their hypothesis that blood glucose control with continuous glucose monitoring offers an improvement compared to manual measurements.

The foundation of this thesis is the author's project report, which comprises e.g. a metabolic simulator and preliminary virtual patients; they were re-worked in order to better capture intensive care patients' unstable dynamics. This thesis (1) examines today's standard for blood glucose management, (2) implements blood glucose control algorithms, (3) simulates the control algorithms on a virtual cohort, and (4) compares said algorithms with the respect of a continuous sensor.

%The first step in this thesis was to examine today's standard for blood glucose management and then create a "virtual intensive care nurse". The second step was to create and implement a blood glucose control algorithm employing a continuous sensor, simulate both the algorithm with the continuous sensor and the virtual nurse, and compare the results with respect to the effects of a continuous sensor.

The blood glucose management of Arendal Hospital and St.Olav's Hospital were researched. A total of six blood glucose control algorithms were implemented: two with different manual measurement intervals; two with a continuous sensor; one exact but older protocol from Stavanger; one autonomous PID controller. The algorithms were run on a virtual cohort of patients with different scenarios: high nutrition, low nutrition, and sensor bias.

Arendal Hospital has a computer decision support system and a goal range of 7.5-10.0 mmol/L. Blood glucose measurements are taken at least every 2 hours. St.Olav's hospital lets the individual nurse decide the insulin dose and takes measurements at least every 4 hours. Their goal range is 6.0-10.0 mmol/L. Continuous glucose monitoring lowers the peak hyperglycemic episodes and subsequent time above the goal range, but does not improve glycemic control.