In a method for controlling the administration of insulin to a patient (P) a target glucose level (T) is set and a controller (10) computes a recommended infusion rate (R) based on the target glucose level (T) and a measured glucose level (G) of a patient (P) for administering insulin to the patient (P). The controller (10) comprises a controller gain (K) for computing the recommended infusion rate (R). Herein, an insulin sensitivity (IS) of the patient (P) is determined by means of a mathematical model taking into account the measured glucose level (G) and an actual infusion rate (IR) of insulin administered to the patient (P) and, based on the insulin sensitivity (IS), the controller gain (K) of the controller (10) is determined. In this way a method for controlling the administration of insulin to a patient is provided which in a reliable, computationally efficient manner allows for maintaining a patient's blood glucose level at or around a desired target glucose level.
