Aim The assessment of the quality of intensive care medicine is mandatory in the modern healthcare system. In Italy, the GiViTI (Gruppo Italiano per la Valutazione degli Interventi in Terapia Intensiva) network is working in this field since 1991 and it now involves 295 out of the about 450 Italian intensive care units (ICU). In 2002 GiViTI launched a project for the continuous quality assessment and improvement that is now joined by 180 ICUs. Data collected in 2005 are analyzed and presented. Methods. All admitted patients were entered in a validated software, which performs a multitude of validity checks during the data entry. Data were further reviewed by the co-ordinating center; patients admitted in months with more than 10% of incomplete or inconsistent records in each ICU were excluded from the analysis. Each year, a multivariate logistic regression model is fitted to identify predictors of hospital mortality. Starting from the SAPS 2 and the 2004 GiViTI model predictions of hospital mortality, two calibration tables and curves are presented. Results. In 2005, 180 Italian ICUs collected data on 55 246 patients. After excluding those admitted in months with an unjustified lower recruitment rate or with less than 90% of complete and consistent data, we had 52 816 (95.6%) valid cases. Although the rough hospital mortality in 2005 was 1% higher than in 2004 (22.6% vs 21.5%), the adjusted mortality shows a statistically significant 4% reduction (obser-ved-to-expected ratio: 0.96; 95% CI: 0.94-0.97). Conclusion Italian ICUs in 2005 performed better than in 2004, at a parity of patient severity.