Current approaches to the diagnosis of delirium are based upon the recognition of symptoms that emerge in the pathological state. As an alternative, we propose an approach to case identification for research purposes based on the recognition of significant changes in the cognitive or cerebral state of the individual patient. Categorical change can be defined using prediction intervals calculated from repeated measures on a population of medically stable subjects. Data from subjects enrolled in a prospective study of delirium in a long-term care population were utilized to calculate prediction intervals for the Mini-Mental Status Examination and for measures of the electroencephalographic background frequency as obtained with a two-channel microprocessor-based EEG device. Preliminary findings support the validity of this quantitative approach for defining changes in brain state. Future research should evaluate both cognitive and electrophysiological techniques for monitoring vulnerable patients.