Refining Baux and rBaux: The impact of full-thickness burn depth on in-hospital mortality prediction
Location
Moultrie, GA
Start Date
17-4-2026 12:00 PM
End Date
17-4-2026 1:00 PM
Description
Background
The Baux and revised Baux (rBaux) scores, combining age, total burned total body surface area (TBSA), and inhalation injury, are widely used to predict burn mortality. Clinically, deeper burns at a given TBSA are thought to carry higher risk, but it is unclear whether burn depth (proportion of 3rd-degree TBSA) meaningfully improves prediction beyond Baux/rBaux in a contemporary cohort.
Methods
We performed a retrospective single-center study of adult burn inpatients with available Baux/rBaux scores and documented 2nd- and 3rd-degree TBSA. The primary outcome was in-hospital mortality. We fit logistic regression models comparing Baux and rBaux to depth-augmented versions (adding proportion 3rd-degree TBSA or 3rd-degree TBSA per 10%) and used likelihood ratio tests (LRTs) and area under the ROC curve (AUC) to quantify incremental predictive value in the full cohort and in a high-risk subgroup (Baux >100). Secondary analyses assessed calibration and AUC across burn-depth strata and tested score-by-depth interaction. An exploratory analysis extended rBaux with burn etiology and depth.
Results
Patients who died had substantially deeper burns than survivors (median proportion 3rd-degree ≈0.80 vs 0.00). In the primary models, adding burn depth significantly improved prediction beyond Baux and rBaux. In the full cohort, AUC increased from 0.73 to 0.84 for Baux and from 0.75 to 0.85 for rBaux when proportion 3rd-degree was added, with LRT p-values < 10⁻⁵ for all depth-augmented models. Each additional 10% of 3rd-degree TBSA was associated with higher mortality (OR≈1.6 per 10%, 95% CI roughly 1.2–2.3) after adjusting for Baux. Similar AUC gains were seen in the Baux >100 subgroup (e.g., Baux AUC 0.74→0.84; rBaux 0.77→0.85). In high-Baux patients, Baux/rBaux over-predicted mortality when ≤25% of the burn was 3rd-degree (observed ≈19%, predicted ≈38–41%) and under-predicted when >75% was 3rd-degree (observed ≈78%, predicted ≈59–60%), while discrimination remained moderate across depth strata and no significant score-by-depth interaction was detected.
In exploratory extensions of rBaux, adding burn etiology increased AUC from 0.77 to 0.85 (LRT p≈2×10⁻⁴), and further adding proportion 3rd-degree increased AUC to 0.88 (LRT p≈0.02); in this fully adjusted model, higher burn depth remained independently associated with mortality (OR≈6.9 for prop_3rd, 95% CI ≈1.4–41).
Conclusions
In this single-center burn cohort, burn depth adds substantial predictive value for in-hospital mortality beyond Baux and rBaux, improving discrimination and explaining miscalibration at extremes of depth, especially among patients with Baux >100. Exploratory multivariable modeling suggests that incorporating both burn mechanism and depth into extended scores may further enhance risk stratification, but these findings require external validation in larger datasets.
Embargo Period
5-27-2026
Included in
Refining Baux and rBaux: The impact of full-thickness burn depth on in-hospital mortality prediction
Moultrie, GA
Background
The Baux and revised Baux (rBaux) scores, combining age, total burned total body surface area (TBSA), and inhalation injury, are widely used to predict burn mortality. Clinically, deeper burns at a given TBSA are thought to carry higher risk, but it is unclear whether burn depth (proportion of 3rd-degree TBSA) meaningfully improves prediction beyond Baux/rBaux in a contemporary cohort.
Methods
We performed a retrospective single-center study of adult burn inpatients with available Baux/rBaux scores and documented 2nd- and 3rd-degree TBSA. The primary outcome was in-hospital mortality. We fit logistic regression models comparing Baux and rBaux to depth-augmented versions (adding proportion 3rd-degree TBSA or 3rd-degree TBSA per 10%) and used likelihood ratio tests (LRTs) and area under the ROC curve (AUC) to quantify incremental predictive value in the full cohort and in a high-risk subgroup (Baux >100). Secondary analyses assessed calibration and AUC across burn-depth strata and tested score-by-depth interaction. An exploratory analysis extended rBaux with burn etiology and depth.
Results
Patients who died had substantially deeper burns than survivors (median proportion 3rd-degree ≈0.80 vs 0.00). In the primary models, adding burn depth significantly improved prediction beyond Baux and rBaux. In the full cohort, AUC increased from 0.73 to 0.84 for Baux and from 0.75 to 0.85 for rBaux when proportion 3rd-degree was added, with LRT p-values < 10⁻⁵ for all depth-augmented models. Each additional 10% of 3rd-degree TBSA was associated with higher mortality (OR≈1.6 per 10%, 95% CI roughly 1.2–2.3) after adjusting for Baux. Similar AUC gains were seen in the Baux >100 subgroup (e.g., Baux AUC 0.74→0.84; rBaux 0.77→0.85). In high-Baux patients, Baux/rBaux over-predicted mortality when ≤25% of the burn was 3rd-degree (observed ≈19%, predicted ≈38–41%) and under-predicted when >75% was 3rd-degree (observed ≈78%, predicted ≈59–60%), while discrimination remained moderate across depth strata and no significant score-by-depth interaction was detected.
In exploratory extensions of rBaux, adding burn etiology increased AUC from 0.77 to 0.85 (LRT p≈2×10⁻⁴), and further adding proportion 3rd-degree increased AUC to 0.88 (LRT p≈0.02); in this fully adjusted model, higher burn depth remained independently associated with mortality (OR≈6.9 for prop_3rd, 95% CI ≈1.4–41).
Conclusions
In this single-center burn cohort, burn depth adds substantial predictive value for in-hospital mortality beyond Baux and rBaux, improving discrimination and explaining miscalibration at extremes of depth, especially among patients with Baux >100. Exploratory multivariable modeling suggests that incorporating both burn mechanism and depth into extended scores may further enhance risk stratification, but these findings require external validation in larger datasets.