Role of Single Worst Injury vs Multiple Injuries in Predicting Pediatric Trauma Mortality

• Published in: Academic emergency medicine 2003-05, Vol.10 (5), p.475-475
• Main Author: Sullivan, T.
• Format: Article
• Language: English
• ISSN: 1069-6563; 1553-2712
• DOI: 10.1197/aemj.10.5.475

BACKGROUND: Recently in the adult trauma population it had been suggested that the worst injury predicts mortality best (Kilgo et al: J Trauma 2002,53:196). OBJECTIVE: To compare worst injury and combined injury scores in prediction of death using a large national pediatric trauma registry. METHODS: The patient population is the National Pediatric Trauma Registry Phase IV data comprising admissions from April of 1994 through Jan 2002. The dataset was randomly divided into a development set (for ICISS and ICISS1) and a test set. ICISS and ICISS1 were established using the development set. ISS, NISS, MaxAIS, ICISS, and ICISS1 were than calculated on each member of the test dataset. Max-AIS is worse anatomic injury in the six body parts. ICISS is ICD-9 injury severity score; ICISS1 is based on single worst probability of survival from patients ICD-9 diagnosis codes. Discrimination was assessed by the area under the Receiver Operator Characteristic Curve (ROC A(z)) (with 95% CI) and calibration assessed for ICISS and ICISS1 using the Lemeshow-Hosmer C-Statistic (L/H C-stat). RESULTS: There were 50199 patients in the cohort. Average age was 8.1 years. There were 1449 deaths (2.9%). NISS and ISS were identical in prediction of mortality. ICISS and ICISS1 had better prediction of mortality. The single worst injury scores (Max-AIS and ICISS1) were not better predictors of mortality then their combined injury counterparts. The table below shows the discrimination and calibration results. ROC Az 95% CI LH-C ISS 0.888 0.869-0.907 NISS 0.890 0.870-0.908 Max-AIS 0.882 0.863-0.901 ICISS 0.956 0.948-0.965 3374 ICISS1 0.954 0.945-0.963 4217 CONCLUSIONS: In this pediatric population single worst injury was not superior to scores which use the patient's multiple injuries in prediction of mortality.