Comparison of the accuracy and correctness of mortality estimates for Intensive Care Unit patients in internal clinics of the Czech Republic using APACHE II, APACHE IV, SAPS 3 and MPMoIII models.

Aim To verify and compare the accuracies of mortality predictions in the Intensive Care Unit (ICU) of the Internal Clinic of Central Military Hospital in Prague, Czech Republic, using model APACHE II and the newer systems of the APACHE IV, SAPS 3 and MPMo III. Methods The data were collected retrospectively between 2011 and 2012, 1000 patients were evaluated. The assessment of the overall accuracy of the mortality predictions was performed using the standardized mortality ratio (SMR), and the calibration was assessed using the Lemeshow-Hosmer "goodness-of-fit" C statistic. Discrimination was evaluated using ROC curves based on calculations of the areas under the curve (AUCs). Results The APACHE II, SAPS 3, and MPMo III systems significantly overestimated the expected mortality, whereas the APACHE IV model led to correct estimations of the overall mortality. The discrimination capabilities of the models assessed according to the constructions of the ROC curves were evaluated as good, only the APACHE II was evaluated as satisfactory. The calibrations of all models were evaluated as unsatisfactory. Conclusion The best mortality estimation for the investigated population sample was provided by the APACHE IV system. The discrimination capabilities of all models for the studied population were satisfactory, but the calibration of all of the systems was unsatisfactory. The conclusions of our study are limited by the relatively small size of the investigated sample and the fact that this study was conducted at only a single site.