The creation of a paediatric surgical unit requires autoevaluation in order to: assess the quality of the results with respect to recognised international standards, answer the family's questions about the results obtained and adhere to criteria of accreditation Between January 2003 and December 2004, 201 consecutive patients, children (N= 164) or operated for adult congenital heart disease (N= 37) were treated. No patient was excluded. The RACHS-1 risk score, the ARISTOTLE scores of complexity and performance and the CUSUM and VLAD graphic analyses were applied to the study of hospital mortality. An original "variable performance-adjusted display" (VPAD) graphic analysis was performed to show up any possible variations of performance. Paediatric hospital survival was 97.56% (95% CI: 93.9 - 99.1). The paediatric complexity and performance scores were 6.79 +/- 0.22 and 6.62 respectively. In the absence of statistical significance in this field of autoevaluation, graphic analyses indicated the performance of our unit with no "learning" curves. Graphic scores and analyses allow assessment of the function of a paediatric cardiac surgical unit and the variations of complexity with respect to time, before the appearance of statistical significance. The ARISTOTLE complexity and performance scores and their adaptation in VPAD seem to be more reliable and discriminating than the RACHS-1 score.