Modified ABSI Versus Revised Baux for In-Hospital Burn Mortality at an Indonesian Burn Unit: A Paired Accuracy and Calibration Study

Abstract

Background: The Modified Abbreviated Burn Severity Index (mABSI) and the Revised Baux score (R-Baux) are bedside scores for predicting death after burn injury, but they have not been compared in the same patients in Sumatra.

Objective: To compare the discrimination, accuracy, agreement and calibration of the two scores for in-hospital mortality.

Methods: Paired retrospective diagnostic-accuracy study of 45 consecutive eligible burn admissions in 2025 at Dr. Mohammad Hoesin General Hospital, Palembang. Both scores were derived from admission data; the reference standard was in-hospital death. We estimated areas under the receiver operating characteristic curve (AUC), accuracy at Youden-optimal cutoffs, paired McNemar tests, agreement, logistic risk curves and calibration against the published mABSI risk categories. Paired analyses needing patient-level linkage are reported as ranges compatible with the observed tables.

Results: Twenty patients (44.4%) died. AUCs were similar: 0.811 (95% CI 0.690–0.932) for the mABSI, 0.808 (95% CI 0.679–0.937) for the R-Baux and 0.803 for burn size alone, with no paired difference (p ≥ 0.922; differences >0.15 excluded). The Youden-optimal cutoff was ≥6 for the mABSI (sensitivity 75.0%, specificity 72.0%) and ≥81 for the R-Baux (70.0% and 84.0%). Deaths were 2.36 times those expected from the published mABSI categories, and at scores of 4–5 mortality was 27.8% rather than 2%. Predicted mortality reached 50% at an mABSI of 6.3 and an R-Baux of about 85.

Conclusion: Both scores ranked mortality equally well but added nothing detectable to burn size, and the published categories underestimated local risk. Locally recalibrated estimates may suit triage but remain provisional pending multicenter validation.