The Utility of Trauma Scoring Systems in Outcome Prediction of Critical Patients

Background: Mortality predictions calculated using scoring scales are often not accurate in populations other than those in which the scales were developed because of differences in case-mix. Scoring systems should be validated to be used in a different population. Methods: Probabilities of ICU death for critically ill trauma patients were calculated using the anatomic scores: Injury severity score ISS and new injury severity score NISS, the physiologic scores: Revised trauma score RTS and acute physiology and chronic health evaluation II APACHE II score and the combined scores: Trauma and injury severity score TRISS and a severity characterization of trauma ASCOT score and the modification of TRISS to TRNISS. The predicted mortality probabilities were compared to the observed mortality. The scores were assessed for discrimination using receiver operator characteristics area under the curve ROC AUC and for goodness-of-fit using Hosmer-Lemeshow test. Aim: To study the utility of different trauma systems in outcome prediction and validation of these scores to be used in the Egyptian Hospitals. The primary outcome under investigation is mortality, but other outcomes were also studied which are the ICU length of stay (LOS) and Glasgow outcome score (GOS). Results: 80 trauma patients admitted to ICU over 1 year could be analyzed. Non survivors during ICU stay were 29 patients (36.25%). In the study 91.25% of the victim’s injuries were caused by blunt trauma and only 8.75% were caused by penetrating trauma. RTS, APACHE II and NISS showed very good ROC AUC while ISS showed a good one (ROC AUC =0.841, 0.881, 0.847 and 0.779 respectively). The combined scores TRISS, ASCOT and TRNISS showed excellent discrimination (ROC AUC =0.902, 0.937 and 0.911 respectively). Regaring calibration RTS, TRISS and ASCOT were well calibrated in our study (HL satistics; p=0.878, 0.781 and 0.843 respectively) while ISS, NISS, APACHE II and TRNISS were poorly calibrated (HL satistics; p=0.152, 0.233, 0.102 and 0.140 respectively). Conclusion: Both TRISS and ASCOT performed better than the anatomic and physiologic scores as regard calibration and discrimination. The replacement of ISS with NISS in Correspondence to: Dr. Gamal Hamed, The Department of Critical Care, Faculty of Medicine, Cairo University. TRISS (TRNISS) resulted in better discrimination but poor calibration. We recommend the combination of APS and NISS to be applied on a larger scale and validated.