Recalibrating the revised Baux score: Admission and 48-hour landmark survival models in contemporary older adult patients with burn injury.

Abstract

Mortality prediction in burn care commonly relies on the revised Baux (rBaux) score, yet its calibration in contemporary cohorts of older adults is uncertain. Because older patients exhibit heterogeneous physiologic reserve and responses to injury, early markers of organ dysfunction may improve prognostication beyond age and burn size. We aimed to assess the discrimination and calibration of rBaux in older adults and to develop simple admission and early landmark models using routinely available physiologic variables. We conducted a single-center retrospective cohort study of patients aged ≥ 65 years with acute burn injury admitted to a tertiary burn center between 2018 and 2023. rBaux discrimination was assessed using ROC analysis and calibration via observed-versus-expected (O:E) mortality. Multivariable logistic regression models were developed at admission and at a 48-hour landmark, prioritizing clinically interpretable predictors and comparing AUCs with rBaux. Among 109 patients, 50 (45.9%) died in hospital. rBaux underpredicted mortality (O:E 1.30, p = 0.022) with acceptable discrimination (AUC 0.839). Two admission models incorporating age, burn extent, inhalation injury, and admission creatinine showed excellent corrected discrimination (AUC 0.904 and 0.897) and outperformed rBaux (p < 0.02). Among 48-hour survivors (n = 89), mortality was 34.8%; landmark models incorporating mechanical ventilation, age, admission creatinine, and burn extent achieved high corrected discrimination (AUC 0.919 and 0.910) and outperformed rBaux (AUC 0.786; both p ≈ 0.001). The revised Baux score demonstrates clinically meaningful calibration drift in contemporary older adult burn patients, with systematic underestimation of mortality. Simple admission models incorporating burn severity, inhalation injury, and admission creatinine improved prognostic performance compared with rBaux, while 48-hour landmark models may support reassessment after initial resuscitation.