Predictive factors of outcome in severely traumatized children.

UNLABELLED: To identify risk factors associated with death in traumatized children, we prospectively studied 507 consecutive patients (7+/-4 yr) admitted to a level I pediatric trauma center over a 3-yr period. Pediatric Trauma Score (PTS), Glasgow Coma Scale (GCS) score, and Injury Severity Score (ISS) were calculated. Age, injury mechanism, injury pattern, and initial critical care were recorded. Univariate and multivariate analyses were performed for potential risk factors associated with mortality. Receiver operating characteristic curves were used to determine threshold values of variables identified by univariate analysis. Most children suffered from blunt trauma (99.6%), and head trauma was noted in 85%. Median values (range) of GCS scores, PTS, and ISS were 10 (3-15), 7 (-4 to 12), and 16 (3-75), respectively. The mortality rate was 12%. Using multivariate analysis, death was significantly associated with an ISS > or = 25 (odds ratio [OR] 22.2, 95% confidence interval 2.8-174.9), GCS score < or = 7 (OR 4.77, 1.8-12.7), emergency blood transfusion > or = 20 mL/kg (OR 4.3, 2.1-9.1), and PTS < or = 4 (OR 3.7, 1.4-9.7). An ISS > or = 25, GCS score < or = 7, immediate blood transfusion > or = 20 mL/kg, and PTS < or = 4 were significant and independent risk factors of death in an homogenous population of severely injured children. The probability of traumatic death was therefore 0 (95% confidence interval 0-0.0135) in children with no one of these threshold values in the four predictive factors and 0.63 (95% confidence interval 0.47-0.76) in those children with all the threshold values. IMPLICATIONS: Methods used for evaluating outcome of trauma patients have essentially been derived from adult series, and attempts to apply them to children have usually been inaccurate. Univariate and multivariate analyses were performed to identify risk factors associated with death in severely traumatized children, and Receiver operating characteristic curves were used to determine threshold values.