BACKGROUND: The purpose of this study was to develop a triaging tool to predict pediatric in-hospital mortality from data available soon after emergency department (ED) presentation. METHODS: The study group consisted of patients of less than 18 years of age from the National Trauma Data Bank with a reported in-hospital mortality status. Variables analyzed were (1) patient demographics, (2) Glasgow Coma Scale (GCS) values, (3) ED vital signs, (4) injury mechanism, and (5) number of days from trauma until admission. Chi-square-assisted interaction detection (CHAID) profiled patient subgroups. The final cohort was randomly divided into 2 equal sets: a training set to subgroup patients and a testing set to validate the prediction accuracy. RESULTS: The cohort consisted of 224,628 patients with 2.29% in-hospital mortality. Sixteen of 19 potential variables were associated with increased risk of in-hospital mortality. The relative risk of dying was 61.7 times greater (95% confidence interval 57.5-66.1) when CHAID predicted mortality relative to when the model predicted survival (P<0.0001). The most powerful variables of the CHAID model were low total GCS scores and systolic blood pressure in the ED. The CHAID model had an improved relative risk and a better combination of sensitivity and positive predictive value compared with GCS and systolic blood pressure in predicting mortality. CONCLUSIONS: The risk of in-hospital mortality for injured children may be identified soon after arrival in the ED. This information may be used by frontline providers to appropriately triage patients to pediatric trauma centers quickly, to guide resuscitation, and for teaching purposes.
BACKGROUND: The purpose of this study was to develop a triaging tool to predict pediatric in-hospital mortality from data available soon after emergency department (ED) presentation. METHODS: The study group consisted of patients of less than 18 years of age from the National Trauma Data Bank with a reported in-hospital mortality status. Variables analyzed were (1) patient demographics, (2) Glasgow Coma Scale (GCS) values, (3) ED vital signs, (4) injury mechanism, and (5) number of days from trauma until admission. Chi-square-assisted interaction detection (CHAID) profiled patient subgroups. The final cohort was randomly divided into 2 equal sets: a training set to subgroup patients and a testing set to validate the prediction accuracy. RESULTS: The cohort consisted of 224,628 patients with 2.29% in-hospital mortality. Sixteen of 19 potential variables were associated with increased risk of in-hospital mortality. The relative risk of dying was 61.7 times greater (95% confidence interval 57.5-66.1) when CHAID predicted mortality relative to when the model predicted survival (P<0.0001). The most powerful variables of the CHAID model were low total GCS scores and systolic blood pressure in the ED. The CHAID model had an improved relative risk and a better combination of sensitivity and positive predictive value compared with GCS and systolic blood pressure in predicting mortality. CONCLUSIONS: The risk of in-hospital mortality for injured children may be identified soon after arrival in the ED. This information may be used by frontline providers to appropriately triage patients to pediatric trauma centers quickly, to guide resuscitation, and for teaching purposes.
Authors: Jody A Vogel; Nicole Seleno; Emily Hopkins; Christopher B Colwell; Craig Gravitz; Jason S Haukoos Journal: Am J Emerg Med Date: 2015-07-06 Impact factor: 2.469
Authors: N Penel; M V Glabbeke; S Mathoulin-Pelissier; I Judson; S Sleijfer; B Bui; P Schoffski; M Ouali; S Marreaud; V Brouste; A Duhamel; P Hohenberger; J-Y Blay Journal: Br J Cancer Date: 2011-04-19 Impact factor: 7.640