THE PERFORMANCE OF SAPS-II IN A COHORT OF PATIENTS ADMITTED TO 99 ITALIAN ICUS - RESULTS FROM GIVITL

Objective: To assess the validity of SAPS II (new Simplified Acute Phy siology Score) in a cohort of patients admitted to a large sample of I talian intensive care units (ICU). Design and setting: The ability of the SAPS II scoring system to predict the probability of hospital mort ality was assessed with calibration and discrimination measures obtain ed using published coefficients. A new logistic regression equation wa s then developed and further formal calibration and discrimination mea sures were estimated for the customized model. Patients: From the 2202 consecutive patients recruited during a 1-month period in 99 ICUs, a total of 1393 patients were included in this validation study. Results : When the parameters based on the standard model were applied, the ex pected probability of mortality did not fit those actually observed in the cohort (p < 0.001), although it showed satisfactory discriminatio n (area under the receiver operating characteristic curve = 0.80). Suc h lack of fit yields an overall under prediction of mortality (observe d/expected ratio = 1.14) that reflects a uniform pattern across a pres elected set of subgroups. Customization allowed new mortality estimate s to be calculated, with satisfactory calibration (p = 0.82) and a mor e uniform pattern across subgroups. Conclusions: SAPS II maintained it s validity in an independent sample of patients recruited in a large n etwork of Italian ICUs only after appropriate adaptation (first-level customization). Whether the determinants of this relatively poor perfo rmance are related to differences in unmeasured case-mix, methods of a pplication, or quality of care delivered is a matter for discussion th at cannot be solved with the data presently available. However, these findings suggest that caution is warranted before implementing the sta ndard SAPS II scoring system parameters outside formal research projec ts.