Generalizability of a Nonlinear Model-based Glycemic Controller

详细信息    查看全文

• 作者：Kent Stewart^* ; Christopher G. Pretty^* ; Felicity Thomas^* ; Geoffrey M. Shaw^** ; Thomas Desaive^** ; Balazs Benyo^*** ; Jozsef Homlok^*** ; Attila Illyes^**** ; Noeimi Szabo Nemedi^**** ; J. Geoffrey Chase^* ; ^{Geoff.chase@canterbury.ac.nz" class="auth_mail" title="E-mail the corresponding author}
• 关键词：Glycemic Control ; Intensive Care ; Hyperglycemia ; Model-based ; Personalized ; Nonlinear ; Stochastic ; Targeted
• 刊名：IFAC-PapersOnLine
• 出版年：2016
• 出版时间：2016
• 年：2016
• 卷：49
• 期：5
• 页码：212-217
• 全文大小：836 K

文摘

Critically ill patients exhibit a highly complex and dynamic metabolic state, making safe, effective management of hyperglycemia difficult. Clinical practices vary significantly, thus making glycemic control (GC) protocols difficult to generalize across units. This study examines the generalizability of a nonlinear, model-based GC protocol, STAR (Stochastic TARgeted), across two different intensive care units (ICUs) with very different practices and use of GC. Patient data from the ICUs at Christchurch Hospital, New Zealand (267 Patients) and Kalman Pandy Hospital, Hungary (47 Patients) are examined. Safety and performance are examined, where complete generalizability would be indicated by similar glycemic performance distributions using non-parametric statistics as appropriate. STAR spent over 86% of time in the target BG band of 4.4-8.0 mmol/L per-episode in both Christchurch and Gyula, with the BG (blood glucose) distributions being almost identical. STAR provided safe GC with very few patients experiencing mild hypoglycemia (< 5 patients, 1.5%). The nonlinear model-based STAR GC protocol delivered equivalent high performance and safety across patient types, time, clinical practice culture, and clinical resources.