A 10-year mono-center study on patients with burns ≥70% TBSA: prediction model construction and multicenter validation - retrospective cohort

Abstract

Background: Burn injuries with ≥70% total body surface area (TBSA) are especially acute and life-threatening, leading to severe complications and terrible prognosis, while a powerful model for the prediction of overall survival (OS) is lacking. The objective of this study is to identify prognostic factors for the OS of patients with burn injury ≥70% TBSA and construct and validate a feasible predictive model.

Materials and methods: Patients diagnosed with burns ≥70% TBSA admitted and treated between 2010 and 2020 in our hospital were included. A cohort of the patients from the Kunshan explosion were assigned as the validation set. The χ2 test and K-M survival analysis were conducted to identify potential predictors for OS. Then, multivariate Cox regression analysis was performed to identify the independent factors. Afterward, we constructed a nomogram to predict OS probability. Finally, the Kunshan cohort was applied as an external validation set.

Results: Sex, the percentage of third-degree and fourth-degree burns as well as organ dysfunction were identified as significant independent factors. A nomogram only based on the factors of the individuals was built and evidenced to have promising predictive accuracy, accordance, and discrimination by both internal and external validation.

Conclusions: This study recognized significant influencing factors for the OS of patients with burns ≥70% TBSA. Furthermore, our nomogram proved to be an effective tool for doctors to quickly evaluate patients' outcomes and make appropriate clinical decisions at an early stage of treatment.

Graphical abstract

Figure 1

A detailed flowchart of our study. The blue part was our selection criteria, and altogether 144 patients who were admitted to our hospital from 2010 to 2020 and diagnosed with burn injuries ≥70% TBSA were included. Following, the green part shows that six parts of patients’ information were extracted for the following analysis (comorbidities, complications, burn area and depth, admission information, demographic factors, and OS). Then, the orange concisely shows our statistical analysis processes. Finally, we performed external validation by using the Kunshan cohort to further verify our previous results and the prediction model.

Figure 2

K-M survival analysis identifies the relationship between the variables and OS probability. (A) Patients in 61–92 age group had the worst prognosis, while the 19–40 age group had the best (P<0.001). (B) Patients caused by chemicals had the highest survival probability, while patients burned by scald injuries and hot solid had the lowest (P=0.016). (C) Different survival probabilities of different TBSA groups (70–75, 76–80, 81–85, 85–90, 91–100) (P<0.001). (D) A high percentage of third-degree and fourth-degree burns contributed to the worse patients’ prognosis (P<0.001). (E) Patients with complications had worse prognosis (P<0.001). (F) Infectious shock severely deteriorated the outcomes of patients with burn injuries ≥70% TBSA (P<0.001). (G) Patients with sepsis demonstrated significantly low survival probability in the K-M survival analysis (P<0.001). (H) The survival probability of patients with organ dysfunction significantly decreased (P<0.001). (I) GI bleeding also served as an indicator of poor survival (P=0.002). GI, gastrointestinal; TBSA, total body surface area.

Figure 3

Explore the role of organ dysfunction in influencing patients’ OS through K-M survival analysis and subgroup analysis. (A) Patients with cardiac dysfunction had significantly shorter OS (P<0.001). (B) Liver dysfunction severely deteriorated the outcomes of patients (P=0.002). (C) Patients with GI dysfunction were significantly associated with a lower survival probability (P<0.001). (D) Patients diagnosed with pulmonary dysfunction had a significantly higher morbidity (P<0.001). (E) Patients with renal dysfunction had significantly shorter OS (P<0.001). (F) Patients in the higher age category were more likely to suffer organ dysfunction (P=0.006). (G) Patients in the higher TBSA category significantly shortened their’ OS (P=0.040). (H) Patients with organ dysfunction were more likely to be associated with sepsis (P<0.001). (I) Patients diagnosed with organ dysfunction were also more likely to suffer infectious shock (P<0.001). OS, overall survival; TBSA, total body surface area.

Figure 4

Construction and validation of a nomogram for predicting the survival probability of patients with burn injury ≥70% TBSA. (A) The nomogram included demographic information and the significant independent factors, which were capable of predicting 1-week, 4-week, 8-week, and 12-week OS probability. (B) The residual plot showed our multivariate Cox regression model’s residual distribution. (C) The calibration of the nomogram suggested that the nomogram was well-calibrated. (D) DCA of the nomogram. When the threshold probability was higher than 0.10, we could obtain high net benefits of 4 weeks OS and 8 weeks OS, while the prediction for 1-week OS yielded no net benefits. (E) The ROC curves of the nomogram for 1-week (AUC=0.854), 4-week (AUC=0.958), 8-week (AUC=0.947), and 12-week (AUC=0.947). AUC, area under the curve; DCA, decision curve analysis; OS, overall survival; ROC, receiver operating characteristic; TBSA, total body surface area.

Figure 5

K-M survival analysis of the Kunshan cohort and external validation of the predictive model. (A) Patients with a high percentage of third-degree and fourth-degree burns had a significantly worse prognosis (P<0.001). (B) Patients with organ dysfunction were significantly less likely to survive (P<0.001). (C) The calibration curve suggested the perfect accuracy of the model in predicting the outcomes of the patients in the Kunshan cohort. (D) The ROC curves with AUC of 0.636, 0.749, 0.88, and 0.912, respectively, in predicting the survival probability at 1-week, 4-week, 8-week, and 12-week indicated acceptable discrimination. AUC, area under the curve; ROC, receiver operating characteristic.

Figure 6

Schematic diagram demonstrating the most significant patient factors and treatment factors for the OS of patients with ≥70% TBSA. (A) The factors influencing OS of MBI patients can be simply categorized into patient factors (e.g. age, sex, cause, burn area, and burn depth), treatment factors (e.g. prehospital emergency measures, early fluid resuscitation, life support, and monitoring, injury and infection control, and surgical treatment), and other factors (e.g. insurance, finances, and medical resource). These factors are undoubtedly very complex and difficult to evaluate systematically. In this study, we constructed a nomogram including five patient factors, which proved to be an effective instrument for doctors to quickly evaluate the outcomes of the patients and make proper clinical decisions, although we did not incorporate all OS-influencing variables. The three factors circled in red (organ dysfunction, and percentage of third-degree and fourth-degree burns) were independent factors. While age, sex, and cause, circled in black, were also potential factors determining the outcome of the patients with burn injury ≥70% TBSA (left). Schematic diagram of widely acknowledged treatment factors and other factors influencing the OS of MBI patients (right). (B) The five specific kinds of organ dysfunction were displayed, including cardiac, pulmonary, GI, kidney, and liver dysfunction. The pathophysiological responses and severe symptoms once the organ dysfunction happened were depicted. GI, gastrointestinal; MBI, Major burn injury; OS, overall survival; TBSA, total body surface area.