Model-based decision support relies on a series of mathematical models and methods to convert raw clinical data into actionable recommendations. High clinical burden associated with measurement, and clinically significant outcomes, make glycaemic control an area where considerable benefit is possible. However, few glycaemic control protocols have been successful in critical care, and fewer exist for outpatient management of diabetes. Challenges faced include high levels of uncertainty and noise, limited measurements, and risk of iatraogenic low blood glucose events. This thesis aims to develop a successful glycaemic control framework, STAR, beyond the critical care environment, and set the stage for an outpatient glycaemic control protocol that individuals with diabetes can use to inform their day-to-day glucose management decisions. To achieve this goal, appropriate models and methods are developed, and validated against both clinical and in-silico data.