Do standard burn mortality formulae work on a population of severely burned children and adults?

Accurate prediction of mortality following burns is useful as an audit tool, and for providing treatment plan and resource allocation criteria. Common burn formulae (Ryan Score, Abbreviated Burn Severity Index (ABSI), classic and revised Baux) have not been compared with the standard Acute Physiology and Chronic Health Evaluation II (APACHEII) or re-validated in a severely (≥20% total burn surface area) burned population. Furthermore, the revised Baux (R-Baux) has been externally validated thoroughly only once and the pediatric Baux (P-Baux) has yet to be. Using 522 severely burned patients, we show that burn formulae (ABSI, Baux, revised Baux) outperform APACHEII among adults (AUROC increase p<0.001 adults; p>0.5 children). The Ryan Score performs well especially among the most at-risk populations (estimated mortality [90% CI] original versus current study: 33% [26-41%] versus 30.18% [24.25-36.86%] for Ryan Score 2; 87% [78-93%] versus 66.48% [51.31-78.87%] for Ryan Score 3). The R-Baux shows accurate discrimination (AUROC 0.908 [0.869-0.947]) and is well-calibrated. However, the ABSI and P-Baux, although showing high measures of discrimination (AUROC 0.826 [0.737-0.916] and 0.848 [0.758-0.938]) in children), exceedingly overestimates mortality, indicating poor calibration. We highlight challenges in designing and employing scores that are applicable to a wide range of populations.