Tight Glycemic Control in Intensive Care: From Engineering to Clinical Practice Change

Abstract

Tight glycemic control (TGC) is prevalent in critical care. Providing safe, effective TGC has proven very difficult to achieve with clinically derived protocols. The problem is exacerbated by extreme patient variability and the need to minimize clinical effort and burden. These ingredients make an ideal scenario for model-based methods to provide optimised solutions. This paper presents the development, clinically validated virtual trials optimisation, and initial clinical implementation of a stochastic targeted (STAR) TGC method and framework. It is compared to a prior successful, model-derived, less flexible and dynamic TGC protocol (SPRINT). The use of stochastic models to safely forecast a range of glucose outcomes over 1-3 hours ensures better performance, more dynamic use of the range of insulin and nutrition inputs and thus better glycemic performance and safety from hypoglycemia, the latter of which was reduced by 3.0x times. Hence, the paper presents an overall engineering approach to TGC from engineering models to clinical implementation and ongoing clinical practice change.