Prognostic Value of Mortality Scoring Systems in Patients With Severe Burns: Identifying Key Predictors of Mortality and Comparative Analysis Between Survivors and Non-Survivors

Introduction

Burn injuries pose a significant clinical and economic challenge to modern healthcare systems worldwide. Although they represent one of the most common unintentional injuries, recent decades have seen a global reduction in burn incidence, a decrease in the severity of burns, and a reduction in the length of hospital stays, which is associated with advances in medical care and increased availability of specialized treatments. Despite this progress, burn-related mortality remains substantial [1,2], with an estimated 180 000 deaths annually and mortality rate in burn centers commonly ranging from 10% to 20% [3,4].

While prevention is the most effective strategy in burn care, once a severe injury occurs, early identification of patients at high risk of death becomes a clinical priority. Numerous prognostic tools are used in everyday practice, including the Baux score and revised Baux score, Abbreviated Burn Severity Index (ABSI), Belgium Outcome Burn Injury (BOBI), and Burn Mortality Prediction (BUMP). Most of these tools were developed based on patient data from diverse populations, with varying demographic and epidemiological profiles. Although age and total body surface area (TBSA) are widely recognized as fundamental determinants of survival, they remain insufficient as isolated predictors. Consequently, burn centers worldwide continue to seek an optimal and comprehensive prognostic assessment tool, underscoring the need for further research [5,6].

In Central and Eastern Europe, including , Poland, epidemiological data on severe burns, particularly among patients requiring intensive care, remain limited. Moreover, few regional studies have directly compared commonly used prognostic scales within a single cohort, creating a gap in evidence relevant for local clinical practice and healthcare planning. Therefore, the aim of this study was to perform a retrospective analysis of patients hospitalized in the intensive care unit (ICU) of the East Center of Burns Treatment and Reconstructive Surgery from 2018 to 2022. The objectives were to evaluate the effectiveness of TBSA, ABSI, revised BAUX, BUMP, BOBI scores in predicting mortality, and to identify key prognostic factors, providing data that may contribute to optimizing treatment strategies and improving the prevention of severe burns worldwide.

Material and Methods

STUDY DESIGN:

We conducted a retrospective analysis of medical records of burn patients admitted to the burn ICU. The study covered a 5-year period, from January 1, 2018, to December 31, 2022. The East Center of Burns Treatment and Reconstructive Surgery, the largest burn center in , Poland, served as the study site. During this timeframe, treatment standards and clinical protocols at our institution remained consistent, and no systematic changes or additional factors were introduced that could have acted as confounders.

STUDY POPULATION:

The total of 144 patients were included in the final analysis. During the study period, all adult burn patients admitted to the center were screened. Patients not requiring ICU admission, those under 18 years of age, and those with incomplete data were excluded. The final cohort consisted exclusively of ICU-treated patients, representing the most severe burn cases.

DATA COLLECTION:

Data were extracted retrospectively from electronic and paper medical records using a structured data extraction form developed specifically for this study. The form was created based on variables commonly reported in burn-related prognostic research and aligned with recommendations for observational retrospective studies. To ensure reliability, the extraction form was pilot-tested and revised to improve clarity and consistency. Data extraction was conducted independently by 2 trained researchers, and discrepancies were resolved through consensus to minimize the risk of information bias. Missing data were handled using a complete-case analysis approach. The proportion of missing data for individual variables was assessed prior to analysis and was below 10% for all variables included in the final models. Variables with incomplete records were not imputed due to the retrospective nature of the study. Inhalation injury and burn depth were assessed based on clinical documentation and bronchoscopy findings where available, which may introduce interobserver variability. The overall process was designed and reported in accordance with the STROBE guidelines for retrospective observational studies [7].

VARIABLES AND PROGNOSTIC SCORES:

The following variables were collected: age, sex, presence of third-degree burns, inhalation injury, selected medical procedures performed during the ICU stay, basic vital signs on admission, results of laboratory tests, and length of ICU stay. Burn depth was assessed in our center clinically by experienced burn surgeons based on standard criteria, including visual inspection and intraoperative findings where applicable.

The following prognostic scoring systems were calculated retrospectively based on clinical data. The TBSA was estimated using the Lund Browder Chart and expressed as percentage of total body surface area burned. The Baux score was calculated as the sum of age and TBSA, and the revised Baux score additionally included inhalation injury (+17 points if present). The ABSI incorporates age, sex, presence of inhalation injury, and burn depth, producing a cumulative score associated with mortality risk. The BOBI is based on categorized age, TBSA, and inhalation injury. The BUMP is a multivariable model incorporating demographic, clinical, and laboratory parameters. All scores were calculated according to their original published definitions.

To evaluate and compare the prognostic performance of individual clinical variables and established composite scoring systems, 2 complementary analytical approaches were applied. First, the baseline model included age, TBSA, and inhalation injury as core predictors, as these variables are consistently recognized as key determinants of mortality in burn patients. Second, the predictive performance of commonly used burn scoring systems (ABSI, revised BAUX, BUMP, and BOBI) was assessed individually. Selected pairwise comparisons were performed focusing on clinically relevant models.

Variables considered to be downstream consequences of clinical deterioration, such as renal replacement therapy, were not included in the multivariable model, to avoid overadjustment and potential collider bias.

ETHICS STATEMENT:

The study was conducted in accordance with the principles of the Declaration of Helsinki. The study received approval from the Institutional Ethics Committee of the Independent Public District Hospital in Leczna (ref. number: 03/WCLO/2023). Due to the retrospective nature of the study and the use of data derived from routine clinical procedures, all of which were fully anonymized and did not permit patient identification, the requirement for informed consent was waived by the institutional ethics committee.

STATISTICAL ANALYSIS:

Data are presented as number (n) and percentage (%). Data with a normal distribution are expressed as mean and standard deviation (SD). Non-normal data are expressed as median and interquartile range (IQR). Data distribution was assessed using the Kolmogorov-Smirnov test and the Lilliefors test. Baseline data were compared using the chi-square test or Fisher exact test for categorical variables, and the t test or Mann-Whitney U test for continuous variables, depending on data distribution.

To address potential multicollinearity between composite scoring systems and their constituent variables, 2 separate multivariable logistic regression approaches were applied. First, a baseline clinical model was constructed including independent variables identified as significant in univariate analysis, including age, TBSA, inhalation injury, and intubation-related variables. Variables were selected based on clinical relevance and to avoid overfitting, rather than including all variables significant in univariate analysis. Second, separate logistic regression models were developed for each prognostic scoring system (ABSI, revised Baux, BUMP, and BOBI), with each score included as a single predictor variable.

The discriminative ability of both the baseline clinical model and individual prognostic scores was assessed using receiver operating characteristic (ROC) curves and the area under the ROC curve (AUROC). In addition to discrimination, model calibration was evaluated using the Hosmer-Lemeshow goodness-of-fit test and calibration plots. ROC curves were generated using MedCalc Statistical Software based on the study dataset. No external images or templates were used.

Pairwise comparisons between AUROC values were performed using the DeLong test to determine whether differences between models were statistically significant. Optimal cut-off values were determined using the Youden index. Sensitivity and specificity were calculated accordingly.

Data were statistically analyzed using IBM SPSS for Windows, version 25.0 (IBM Corp, Armonk, NY, USA) and MedCalc Statistical Software version 20.218 (MedCalc Software Ltd, Ostend, Belgium). A 2-sided P value of <0.05 was considered statistically significant.

Results

PATIENT CHARACTERISTICS:

In the 5-year period from 2018 to 2022, a total of 144 burn patients were admitted to the ICU of the East Center of Burns Treatment and Reconstructive Surgery, of whom 62 died (43.1%). The group was 75% male with an average age of 52 years. Most patients presented with third-degree burns (93.1%), and inhalation injury was identified in 41.0% of cases. The median length of stay was 26 days.

UNIVARIATE ANALYSIS:

The univariate comparison of characteristics between surviving and non-surviving patients is summarized in Table 1. Patients who did not survive were older (59.2 years vs 46.5 years), presented more frequently with respiratory tract burn injury (59.7% vs 26.8%), underwent endotracheal intubation on-site more often (54.8% vs 37.8%), and were more likely to require renal replacement therapy (53.2% vs 22.0%). Non-survivors also had higher median TBSA (42.3% vs 35.3%) and ABSI scores (8.0 vs 7.0). Similarly, higher mean revised BAUX (103.9 vs 82.4), BUMP (−1.0 vs −2.3), and BOBI (4.8 vs 2.8) scores were observed among non-survivors.

Laboratory parameters assessed in the univariate analysis showed significant differences in creatinine levels on day 2 after admission (0.9 vs 0.8) and sodium levels on day 10 (145.0 vs 140.6). A significant change in creatinine levels between admission and day 2 was also noted among non-survivors (5.6 vs −14.9). No additional significant differences were identified. Detailed data is presented in Table 2.

MULTIVARIABLE AND ROC ANALYSIS:

In the baseline clinical multivariable logistic regression model, age, TBSA, and inhalation injury remained independently associated with mortality. Increasing age and TBSA were associated with a higher risk of death, while the presence of inhalation injury significantly increased mortality risk. Intubation variables showed elevated odds ratios; however, their statistical significance was attenuated after adjustment for other clinical variables (Table 3). The Hosmer-Lemeshow goodness-of-fit test did not indicate poor model fit (χ2=14.11, df=8, P=0.079).

The baseline clinical model demonstrated excellent discriminative ability, with an AUROC of 0.87 (95% CI, 0.81–0.93). In separate logistic regression models, each prognostic scoring system was significantly associated with mortality. The revised BAUX score demonstrated the highest discriminative ability (AUROC 0.86; 95% CI, 0.80–0.92), followed by BUMP (AUROC 0.83; 95% CI, 0.76–0.89) and BOBI (AUROC 0.83; 95% CI, 0.76–0.89). The ABSI score showed moderate performance (AUROC 0.78; 95% CI, 0.70–0.85), whereas TBSA and creatinine at 48 hours demonstrated lower predictive value (AUROC 0.68 and 0.63, respectively). Pairwise comparisons of AUROC values using the DeLong test showed no statistically significant difference between BOBI and revised BAUX scores (P=0.185), nor between BOBI and BUMP (P=0.998). However, the revised BAUX score performed significantly better than ABSI (P<0.001) and TBSA (P<0.001). When compared with the baseline clinical model, no statistically significant differences were observed for the revised BAUX score (P=0.526), while the baseline model demonstrated superior performance compared with ABSI (P=0.0009). Findings are presented in Figure 1. For clarity, AUROC values are also presented in Table 4.

These findings indicate that both the baseline clinical model and selected composite scores, particularly the revised BAUX, BOBI, and BUMP scores, provide strong and comparable predictive performance.

Discussion

STUDY LIMITATIONS:

The limitations of our study include its retrospective design and the fact that it was conducted in single center, which may introduce selection bias and limit the generalizability of our findings. The study population represents a highly selected cohort of critically ill patients, which can limit the generalizability of the findings to less severe burn populations. The relatively small sample size may increase the risk of overfitting, particularly in multivariable analyses. Additionally, our database lacks complete information on patient comorbidities, restricting the ability to incorporate this variable into the analysis. Prospective studies could provide more robust and accurate data regarding the influence of therapeutic interventions or comorbidities on mortality outcomes.

Moreover, the regional character of the cohort limits the broader novelty of the findings, as the results may not fully reflect differences in burn care practices across other health systems or geographic settings. Importantly, while BOBI, ABSI, and revised BAUX scores all incorporate similar core variables, differences in predictive performance may arise from variations in weighting schemes and categorization thresholds. The slightly higher performance observed for specific models in this study may reflect better calibration to the characteristics of the analyzed population rather than a fundamentally superior model structure. Alternatively, given the lack of statistically significant differences between the highest-performing scores, these variations may also represent random fluctuations related to sample size.

Conclusions

Our findings confirm that both baseline clinical variables and composite prognostic scores provide meaningful support in assessing mortality risk in patients with severe burns. The baseline clinical model, based on age, TBSA, and inhalation injury, demonstrated excellent predictive performance comparable to the best-performing scoring systems. Among the evaluated tools, the revised BAUX score achieved the highest discriminatory ability. However, differences between revised BAUX, BOBI, and BUMP scores were not statistically significant, suggesting comparable clinical utility.

These findings indicate that simple clinical models may serve as effective alternatives for mortality prediction, particularly when implemented as calculators or nomograms. At the same time, established scoring systems remain valuable due to their simplicity and ease of bedside application. Further refinement of mortality prediction methods and better integration of coordinated multidisciplinary burn care remain essential. Multicenter prospective studies are needed to validate these observations and determine their applicability across different healthcare settings.