evaluation of the performance assessment of APACHE III scoring system in patients admitted in the ICUs of TUMS with Sina designed APACHE scoring model

ABSTRACT

Performance assessment of intensive care units may not be possible unless there are objective criteria. This performance eventually determines patients' outcome and is the most reliable outcome measure for intensive care in hospital mortality. Scoring systems have been proposed for individual patient outcome prediction, for evaluating the performance of ICUs and for therapeutic trials. Despite a reasonable accuracy of scoring systems for predicting mortality, probability of defined ICU populations in developed countries, the healthcare delivery systems and patient characteristics in developing countries might influence the correlation between a given scoring instrument and the probability of death. Accordingly, intensive care physicians in developing countries should verify the accuracy of the available severity scoring systems in their specific settings by an objective validation process. This study tries to evaluate validity of APACHE III model in patients admitted to ICUs of Tehran University Medical Sciences intensive care units based on assessing calibration and discrimination. In the study, 1312 patients were analyzed. The mean age was 48.7 and median of length of stay at ICU and hospital was 4 and 9 days, respectively. Mean APACHE III score was 60 at first day of admission. Customized APACHE III [APACHE IIIc] model had poor calibration [P<0.0001], but good discrimination [area under ROC curve =0.88].In comparison, Sina-designed APACHE III [APACHE Ills] model had appropriate calibration [P value =0.4] and good discrimination [area under ROC curve =0.87]. Although APACHE Ills model sensitivity was about 70%, acceptable specificity [86.1%] plus corrected classification percent as%79, it is being introduced as an acceptable model in intensive care units of Tehran University of Medical Sciences; however extended use of this model needs more comprehensive studies