Nonresponse and methods of data collection could affect satisfaction measurement. The goal of this study was to estimate the impact of (1) nonresponse and (2) distribution method on evaluation of patient satisfaction in a mail-back study measuring patient opinion of medical and nursing care. The study was conducted in an adult hospital. Patients were pseudo-randomized according to the initial mode of questionnaire distribution (given at hospital or sent by mail). Three reminders were made at 1-week intervals to nonrespondents, regardless of the method of initial questionnaire distribution. Groups were distinguished according to the delay of response: initial (before any reminder), middle (after one or two mailed reminders), and late respondents (after mailed reminders plus telephone contact). The study included consecutively discharged patients to obtain 300 patients per arm. 482 patients returned the questionnaire (248 in the group receiving the questionnaire at the hospital and 234 in the other group). Groups were compared for satisfaction scores and delay of response. Early respondents were compared with middle and late respondents for patient characteristics, modality of hospital care, and satisfaction scores. Multivariate analyses were performed. Participation rate before any reminder was higher when the questionnaire was mailed than when it was given at the hospital (45% versus 39.7%, p = 0.03). The initial method of distribution did not influence patient satisfaction level. Satisfaction did not differ between respondents with or without reminders. Distributing questionnaires by mail may be preferred to distribution at discharge to optimize response rate. Reminders do not seem necessary to estimate satisfaction of overall potential respondents.