Assessment of mortality prediction models in a Ghanaian burn population

Purpose: Over 40 new or modified outcome prediction models have been developed for severe burns; with age, total burned surface area (TBSA) and inhalation area as major determinants of mortality. The objective of this study was to assess their applicability in a developing country. Procedures: Data were collected retrospectively of a consecutive series of 261 patients (2009-2011) admitted to a Burns Intensive Care. Five outcome prediction models based on admission criteria were evaluated: Bull grid, Abbreviated Burn Severity Index ABSI, Ryan-model, Belgian Outcome in Burn Injury - BOBI and revised Baux. Discriminative power and goodness-of-fit were assessed by receiver operating characteristic analyses (area under the curve - AUC) and Hosmer-Lemeshow tests. Findings: Median age was 10.5 years (IQR: 2.5-27 years), median TBSA 21% (IQR: 11-34%); 55.2% were male, 28 patients died (10.7%). Only 2 patients were intubated (0.8%). The AUC were between 77 and 86%. The ABSI model showed the best calibration (28.7 expected deaths). Ryan, BOBI and rBaux significantly underestimated mortality, whereas Bull showed an overestimation. Conclusion: This study on a young group of burn patients showed moderate to good discriminative power using all five prediction models. The expected number of deaths tended to be underestimated in the three most recent prediction models. (C) 2012 Elsevier Ltd and ISBI. All rights reserved.