Prediction of mortality from catastrophic burns in children

Background: We previously developed a model to predict survival in massive paediatric burns (>80% total body surface area [TBSA]). This model included not only demographic variables, but also variables obtained throughout the hospital course. We aimed to prospectively validate our model for accuracy of outcome prediction. Methods: We admitted 33 paediatric burn patients with burns greater than 80% TBSA. We recorded age, burn size, inhalation injury, resuscitation, packed-cell volume at admission, base deficit, serum osmolarity, sepsis, inotropic support, platelet count, creatinine, and ventilator dependency. We entered these data into our previous models. Results: 20 male and 13 female children with mean age 7.6 (SD 1) years with TBSA burns of 88% (SD 1; full thickness 86% [SD 1]) were admitted. Mortality was 39.4% (13 of 30). When all variables were integrated into our final model, we predicted outcome with 97% accuracy. When we used a model based only on demographic characteristics of age, burn size, and presence of inhalation injury, outcome was correctly predicted in only 51% of patients. Conclusions: We show prospectively that mortality in severely burned children can be reliably estimated at a burn centre, and that outcome cannot be reliably predicted on the basis of demographic and injury characteristics alone. These data suggest that all severely burned children should be given a course of treatment before consideration of treatment futility.