[Construction and validation of a predictive model for the risk of ARDS in severely burned patients]

Objective: To construct and validate a nomogram model for predicting the risk of acute respiratory distress syndrome (ARDS) in severely burned patients. Methods: This study was a retrospective cohort study. From January 2015 to August 2025, 372 severely burned patients who met the inclusion criteria were admitted to the Fourth Medical Center of Chinese PLA General Hospital, including 279 males and 93 females, aged 18 to 78 years. According to whether ARDS occurred within 7 days post-injury, patients were divided into ARDS group and non-ARDS group. Stratified random sampling was used to allocate patients into training set and validation set in a ratio of 7∶3. In training set, there were 35 patients in ARDS group and 227 patients in non-ARDS group; in validation set, there were 14 patients in ARDS group and 96 patients in non-ARDS group. General clinical data of the patients were collected, including total burn area, modified Baux score, burn index, partial-thickness burn area, full-thickness burn area, severity of inhalation injury, whether mechanical ventilation or high-flow oxygen therapy performed upon admission, number of days after tracheostomy, length of hospital stay, duration in critical illness, and mortality rate, as well as data related to post-burn ARDS development within 24 hours of admission, including white blood cell count, neutrophil count, thrombin time, and levels of hemoglobin, creatinine, chloride, procalcitonin, albumin, and fibrinogen (FIB). Variables with statistically significant differences between two groups in training set were screened, and characteristic variables were further selected using least absolute shrinkage and selection operator (LASSO) regression combined with a weight adjustment strategy. Multivariate logistic regression analysis was performed to identify independent predictors, based on which a nomogram model was constructed. The performance of the model was validated in both training set and validation set. Results: In the general clinical data of training set, patients in ARDS group had significantly larger total burn area and full-thickness burn area (P<0.05), significantly higher modified Baux score, burn index, proportion of patients receiving mechanical ventilation or high-flow oxygen therapy upon admission, and mortality rate (P<0.05), significantly more severe degree of inhalation injury (P<0.05), significantly greater number of days after tracheostomy, length of hospital stay, and duration in critical illness (P<0.05), and significantly smaller partial-thickness burn area (P<0.05) than those in non-ARDS group. With the data related to post-burn ARDS development, patients in ARDS group had significantly higher white blood cell count, neutrophil count, and levels of hemoglobin, creatinine, chloride, and procalcitonin (P<0.05), significantly longer thrombin time (P<0.05), and significantly lower levels of albumin and FIB (P<0.05) than those in non-ARDS group within 24 hours of admission. LASSO regression analysis showed that modified Baux score, neutrophil count, albumin level, and FIB level were the characteristic variables for the development of ARDS in 262 severely burned patients in training set. Multivariate logistic regression analysis revealed that modified Baux score, neutrophil count, albumin level, and FIB level were independent predictors of ARDS in 262 severely burned patients in training set (with ORs of 1.058, 1.147, 0.752, and 0.615, respectively, 95%CIs of 1.045-1.072, 1.092-1.206, 0.694-0.814, and 0.500-0.757, respectively, P values all <0.05). A nomogram model for predicting the risk of ARDS in 262 severely burned patients in training set was constructed based on the above four independent predictors. Receiver operating characteristic curve analysis showed that the model had an area under the curve (AUC) of 0.960 (with 95%CI of 0.934-0.986) in training set and an AUC of 0.914 (with 95%CI of 0.837-0.991) in validation set. The calibration curve showed that the risk of ARDS predicted by the model was in high agreement with the actual observed values. Clinical decision curve analysis showed that within the threshold probability range of 17% to 78%, the clinical net benefit obtained by applying this model for clinical decision-making was significantly higher than that of the strategies of intervening in all patients or intervening in none. Conclusions: The constructed nomogram model incorporating the modified Baux score, neutrophil count, albumin level, and FIB level demonstrates good predictive performance for the risk of ARDS in severely burned patients and holds potential value for clinical decision-making.