A comparison of illness severity scores in an intensive care unit

Severity scores for both illness and pro gnosis of patients’ mortality are widely used in intensive care units to approach items related with effectiveness of intensive health care. Aim: The aim of this prospective cohort study was to determine the applicability of two well-known scoring systems, APACHE II and SAPS II in an interdisciplinary intensive care unit (ICU) in Greece and to evaluate the two systems’ performance in predicting hospital mortality. MATERIAL-Method: Data enabling the calculation of APACHE II and SAPS II were collected for 342 patients admitted in the ICU (pathological-surgical) over a oneyear period (January 2001–January 2002). Standardized mortality ratios, discrimination (area under the ROC curve) and calibration (Hosmer-Lemeshow test and calibration graphs) were calculated for both systems. Results: Both systems significantly underestimated mortality showing similar quality of adaptation to data. The area under the ROC curve was 0.747 for APACHE II and 0.791 for SAPS II, showing satisfactory discriminative power for both systems. The systems correspondence between observed and predicted mortality according to Hosmer-Lemeshow control was also adequate (APACHE II: x2=9.37, P-value=0.49, SAPS II: x2=8.34, P-value=0.59). However, the calibration curves showed poor calibration, especially for SAPS II, with a tendency for the system to underestimate mortality in the middle levels of risk. Conclusions: The two systems’ performance was deemed similar. Customization of the systems is necessary in order to improve their performance when used in this particular intensive care setting.