Peut-on mesurer la performance des services grâce à l'IGS II?

Rationale. - Risk-adjustment is one of the main tools for assessing clinical performance of intensive care units (ICUs). However, numerous factors, notably case-mix or pre-admission conditions, perturb the uniformity of fit of predictive models, thus the use of standardised ratios to compare ICUs. Objective. - To investigate and validate an updating of the simplified acute physiology score II (SAPS II). Methods. - Prospective, multicenter study, involving 26 ICUs located in Paris area participating in the Cub-Rea performance research project and more than 30 000 stays upon 2 years. Logistic regression analysis incorporating SAPS II, admission type, diagnostic information was used to estimate ICUs and hospital mortality for intensive care patients. Diagnostic information was successively summarised by main diagnosis then by outcome related groups of diseases elaborated by means of a classification tree program. Internal and external validation were performed. Results. - All new models had good and similar or better discrimination compared with the SAPS II area under Roc curve (0.92 = vs. 0.88 for SAPS II). Uniformity of fit was obtained by inclusion of diagnostic information. Conclusion. - The SAPS II can be improved by integration of diagnostic information, thus gaining uniformity of prediction across various conditions.