Outcome Prediction in a Surgical ICU Using Automatically Calculated SAPS II Scores

The objective of this study was to establish a complete computerized calculation of the Simplified Acute Physiology Score (SAPS) II within 24 hours after admission to a surgical intensive care unit (ICU) based only on routine data recorded with a patient data management system (PDMS) without any additional manual data entry. Score calculation programs were developed using SQL scripts (Structured Query Language) to retrospectively compute the SAPS II scores of 524 patients who stayed in ICU for at least 24 hours between April 1, 1999 and March 31, 2000 out of the PDMS database. The main outcome measure was survival status at ICU discharge. Score evaluation was modified in registering missing data as being not pathological and using surrogates of the Glasgow Coma Scale (GCS). Computerized score calculation was possible for all investigated patients. The 459 (87.6%) survivors had a median SAPS II of 28 (interquartile range (IQR) 13) whereas the 65 (12.4%) deceased patients had a median score of 43 (IQR 16; P<0.001). Of the physiological variables for SAPS II score calculation, bilirubin was missing in 84%, followed by P a O 2 / FiO 2 ratio (34%), and neurological status (34%). Using neurological diagnoses and examinations as surrogates for the GCS, a pathological finding was seen in only 8.8% of all results. The discriminative power of the computerized SAPS II checked with a receiver operating characteristic (ROC) curve was 0.81 (95% confidence interval (CI): 0.74–0.87). The Hosmer-Lemeshow goodness-of-fit statistics showed good calibration (H=5.55, P=0.59, 7 degrees of freedom; C=5.55, P=0.68, 8 degrees of freedom). The technique used in this study for complete automatic data sampling of the SAPS II score seems to be suitable for predicting mortality rate during stay in a surgical ICU. The advantage of the described method is that no additional manual data recording is required for score calculation.