Comparison of APACHE II, TRISS, and a proposed 24-hour ICU point system for prediction of outcome in ICU trauma patients.

The APACHE II system for predicting outcomes in critically ill patients is now being used to evaluate quality of care for patients in surgical intensive care units, including trauma patients. The trauma data, however, on which the APACHE outcomes are based, were derived from only 364 ICU trauma patients. We compared the outcome predictions by APACHE II, TRISS, and a proposed 24-hour ICU point system in 1,000 ICU patients. [table: see text] p less than 0.025 by unpaired t test for predictive power of ICU point system versus APACHE II. Values of more than 15.5 represent poor agreement between the outcomes estimated from the model and the observed outcomes; a low value represents good agreement. The APACHE system significantly overestimated the risk of death in the lower ranges of predicted risk and underestimated the deaths in the higher ranges. Although TRISS was not developed for ICU trauma patients, it tended to perform better than APACHE II in our sample. The 24-hour ICU point system performed well, with accurate agreement between the outcomes estimated from the model and the observed outcomes.