A simple tool for mortality prediction in burns patients: APACHE III score and FTSA.

Prediction of outcome for patients with major thermal injury is important to inform clinical decision making, alleviate individual suffering and improve hospital resource allocation. Age and burn size are widely accepted as the two largest contributors of mortality amongst burns patients. The APACHE (Acute Physiology and Chronic Health Evaluation) III-j score, which incorporates patient age, is also useful for mortality prediction, of intensive care populations. Validation for the burns specific cohort is unclear. A retrospective cohort study was performed on patients admitted to the Intensive Care Unit (ICU) via the Victorian Adult Burns Service (VABS), to compare observed mortality with burns specific markers of illness severity and APACHE III-j score. Our primary aim was to develop a mortality prediction tool for the burns population. Between January 1, 2002 and December 31, 2008, 228 patients were admitted to the ICU at The Alfred with acute burns. The mean age was 45.6 years and 81% (n=184) were male. Patients had severe injuries: the average percent TBSA (total body surface area) was 28% (IQR 10-40) and percent FTSA (full thickness surface area) was 18% (IQR 10-25). 86% (n=197) had airway involvement. Overall mortality in the 7-year period was 12% (n=27). Non-survivors were older, had larger and deeper burns, a higher incidence of deliberate self-harm, higher APACHE III-j scores and spent less time in hospital (but similar time in ICU), compared with survivors. Independent risk factors for death were percent FTSA (OR 1.03, 95% CI 1.01-1.05, p=0.01) and APACHE III-j score (OR 1.04, 95% CI 1.02-1.07, p<0.001). Mortality prediction based on both of these variables in combination was more specific than either individual variable alone (AUROC 0.85, 95% CI 0.79-0.92). Likelihood of death for patients with severe thermal injury can be predicted with accuracy from APACHE III-j score and percent FTSA. Prospective validation of our model on different burn populations is necessary.