The objective of this research was to devise a survey instrument specifically applicable to prepaid health care plans that could accurately predict whether patients would disenroll from their current plan because of dissatisfaction when given the opportunity to do so. A "prequestionnaire" was sent to all employees at a southwestern university whose employee benefit package included the option of selecting one of several health maintenance organizations (HMOs) as a source of health care. The prequestionnaire included 90 variables reported in the literature as related to patient satisfaction. The prequestionnaire was mailed two months before "open enrollment," the time at which subjects would have the opportunity, if desired, to change HMOs. After open enrollment, a "postquestionnaire" was sent to the same subjects, asking whether or not they did change plans during open enrollment. There were 2,365 respondents enrolled in HMOs who formed the study population. Of these, 189 (8.0%) changed HMOs during open enrollment. Discriminant function analysis was used to identify prequestionnaire variables which were predictive that subjects had changed plans; 10 variables were identified. They were combined into a survey instrument, which can be scored to predict an individual subject's probability of changing plans.