Blood Glucose Control of Critically Ill Patients with Time-varying Insulin Sensitivity Using Nonlinear Model Predictive Control

Abstract

In critically ill patients suffering from hyperglycemia it has been recently shown that mortality and morbidity can be reduced by keeping glycemia within 80-110 mg/dL (4.4-6.1 mmol/L). However, maintaining blood glucose (BG) levels within such range is difficult because of the time variability in insulin sensitivity through the patient recovery. In this study, we introduce a nonlinear model predictive control algorithm using a time-invariant model of glucose-insulin metabolism for the eventual development of a control system that regulates BG levels of critically ill patients. Simulation results using virtual patients with time-varying insulin sensitivity show that BG levels can be kept 58.3% within the 80-110 mg/dL range with only 1.4% of BG levels decreasing below 80 mg/dL (4.4 mmol/L), which demonstrates the safety of the present control algorithm.

Glycemic Control for Critically Ill Patients with Online Identification of Insulin Sensitivity

“…The initial value of the insulin sensitivity is set to 0.3 at admission to ICU from the identi cation result of insulin sensitivity [12] and the results in Section 2, and the other parameters are set to the values listed in Table 2.…”

“…We construct 30 patients by combining the ten different insulin sensitivity pro les with patients weighing 50, 60 and 70 kg (female, 48 years of age, 170 cm in height, no history of diabetes) due to less in uence of gender, age and height [12].…”

“…To solve such problems, glycemic control for ICU patients to normalize BG without causing hypoglycemia has been studied [5][6][7][8][9][10][11][12][13][14], and computer-based insulin infusion protocols [5][6][7][8][9] have been shown to provide more effective and safer BG control for ICU patients than paper-based protocols. The Glucosafe system [5] for tight glucose control in critical care is a model-based decision support system that utilizes BG trajectory prediction based on the Glucosafe model [15].…”

“…In our previous study [12], we developed a BG control system using nonlinear model predictive control based on ICUMM [17] with no time-varying parameters. Although our system achieved 1.4% of BG below 80 mg/ dL when applied to virtual patients with time-varying insulin sensitivity, it did not have suf cient performance to maintain BG within the desired range.…”

Glycemic Control for Critically Ill Patients Using Zone Model Predictive Control

“…Due to limitation of paper‐based insulin infusion protocols, such as lack of considering inter‐individual differences between patients, computer‐based protocols have been widely studied [5–17]. The Glucosafe system [5] for tight glucose control in critical care is a model‐based decision support system based on the prediction of BG trajectory using the Glucosafe model [18].…”