The present disclosure pertains to a system and method for determining whether a subject is likely to be disturbed by therapy levels of stimulation provided to the subject during sleep sessions. The present system is configured to automatically identify sensitive users using electroencephalogram (EEG) information from a reference sleep session with or without stimulation. For reference sleep sessions without stimulation, the alpha activity in detected deep sleep is used to predict whether the subject is likely to be disturbed by therapy levels of stimulation. For reference sleep sessions with stimulation, the acute increase in EEG delta (e.g., 0.5-4 Hz) power and/or an arousability index are used to predict whether the subject is likely to be disturbed by therapy levels of stimulation.