This paper describes the development of a model for classifying the patient population into risk groups according to their medication error probability. The methodology utilizes patient demographic and health care data combined with medication error reports. Variable selection is based on a measure of the amount of information that they provide. Using data from a 360-bed hospital, the model identified patient characteristics related to the health care services received as being the main variables related to medication error probability and divided the patient population into three risk groups defined in terms of these characteristics. The model provides information valuable to risk management and quality assurance efforts in the areas of detection of risk exposure, performance monitoring and resource allocation.