Approximate entropy, a measure of regularity, can be used to analyze the electroencephalogram of patients in general anesthesia to discriminate between different states of consciousness. EEG burst suppression patterns reflect a state of deep anesthesia. Due to the instationary character of this EEG pattern approximate entropy values do not correctly classify the patient state. Possible solutions to this problem may be limited by the demand of computing power for entropy calculation and the reaction time following changes in patient state. Different approaches for an online monitoring application are examined.