Control of blood glucose concentration in type-I diabetic patients in presence of meal disturbances has attracted the attention of many researchers from past 40 years. The success of strategies proposed in the literature depends on the accuracy of the model used in the control frame-work. In this work, a data based model predictive control (MPC) algorithm is developed to control the blood glucose concentration in the Type-I diabetic patients in the presence of meal disturbances under patient-model mismatch. A state space model with augmented states representing integrating type of disturbances is used for the future predictions and then in optimizing the future insulin infusion rate. The process states along with the disturbances at each sampling instant are estimated using recursive form of Kalman filter. Appropriate physical and physiological constraints are incorporated in the objective function of MPC to ensure feasible operating regime. Simulation studies are performed on three distinct patient models using Simulink®. The input-output data required for model identification has been obtained from the perturbation studies on patient-1. The state-space model developed is used in the state estimation based linear MPC, which is employed on all three patients. Simulation results revealed that, the proposed control strategy is able to control the blood glucose concentration well within the acceptable limits in the presence of meal disturbances. It was also observed that performance of this strategy, even under large patient-model mismatch along with unmeasured disturbances, is quite encouraging.