Résumé : The present study compared the performance of a multidimensional patient satisfaction questionnaire according to the timing of questionnaire administration. Comparisons were made in terms of: (a) the completeness and representativeness of the data set (number of missing questionnaires, number missing item responses, respondents' representativeness to the target population); (b) the questionnaire acceptability to respondents (time and difficulty to complete); (c) the questionnaire reliability; and (d) variability of scores. One hundred and ten consecutive breast cancer patients hospitalised for surgery were randomised between being sent the comprehensive assessment of satisfaction with care (CASC) at 2 weeks (T 2W) or at 3 months (T 3M) after hospital discharge. The time to complete the CASC was shorter at T 3M than at T 2W and the mean percentage of item omission was lower at T 3M (1.68) than at T 2W (3.82). However, the response rate was much higher at T 2W (87%) than at T 3M (66%), making item omission non-significant. At both times of questionnaire administration samples were equally biased towards patients having undergone a less invasive surgery. Moreover, the multi-item scales of the CASC demonstrated adequate internal consistency coefficients, except the general satisfaction scale at T 3M, and fairly symmetrical distribution of scores. Response rate should be considered in priority. This criteria favoured an administration of the CASC shortly after discharge. Besides in a cancer patient population care experience and perception may vary in a 6 weeks time lapse. The timing of assessment needs to be clearly specified in cancer patients satisfaction survey. © 2002 Elsevier Science Ireland Ltd. All rights reserved.