Ruslan Konstantinov Tachev

Burn mortality equations are developed in broad burn populations and then applied to severe burns, yet validation studies report discrimination far more often than calibration. Discrimination establishes only that a score ranks patients correctly within the population at hand; it is the predicted probability that is used for audit, benchmarking and discussion of prognosis. We externally validated the Burn Mortality Prediction (BUMP) and revised Baux equations as probability models, with Baux and the Abbreviated Burn Severity Index (ABSI) as ranking scores, in 172 consecutive adults with acute thermal burns of at least 20% total body surface area admitted to a tertiary burn centre in Bulgaria between 24 January 2019 and 18 July 2026. Published coefficients were applied without refitting. The cohort differed substantially from the BUMP development registry: median total body surface area was 27.5% against 7%, and in-hospital mortality was 37.2% (64 deaths) against 8.8%. Both equations nonetheless tracked observed risk. Mean predicted mortality was 32.7% for revised Baux and 32.3% for BUMP, observed-to-expected ratios were 1.14 (95% confidence interval, 0.88–1.45) and 1.15 (0.89–1.47), calibration slopes were 1.05 (0.73–1.38) and 1.03 (0.71–1.36), and Brier scores were 0.142 and 0.150. Areas under the receiver operating characteristic curve were 0.878 for ABSI, 0.870 for revised Baux, 0.859 for BUMP and 0.827 for Baux. The prespecified comparison of the six-input BUMP score with the three-variable revised Baux score gave a difference of −0.011 (−0.041 to 0.019), excluding any BUMP advantage larger than 0.019. In major burns, published equations transported with acceptable calibration and the more elaborate model conferred no discernible advantage.