Construction and validation of a predictive model for mortality risk in patients with Acinetobacter baumannii-induced septicaemia

Abstract

Purpose To develop and validate a predictive model for the risk of death in patients with Acinetobacter baumannii (A. baumannii) sepsis for clinical decision-making and patient management.Methods Demographic and clinical data related to patients with A. baumannii sepsis admitted to the Guangdong Second Traditional Chinese Medicine Hospital for 13 consecutive years from January 2011 to December 2023 were collected. The 160 patients admitted from January 2011–December 2020 composed the training cohort, and the 46 patients admitted from January 2021–December 2023 composed the validation cohort according to the time of admission. LASSO regression analysis and multivariate Cox regression were used to determine the independent risk factors for death in patients with A. baumannii sepsis, and a nomogram was constructed based on the results of multivariate Cox regression. The predictive model was evaluated using the area under the curve (AUC) of the subject's work characteristics (ROC) curve, decision curve analysis (DCA), and standard curves for discrimination, accuracy, and calibration.Results Comorbid septic shock, an elevated neutrophil/lymphocyte ratio (NLR), low hemoglobin levels, and low platelet counts were found to be independent risk factors for death in patients with A. baumannii sepsis. With the models constructed from these four variables, the AUCs of the ROC curves of the test and validation cohorts for the prognostic scenarios at 7, 14, and 28 days were not less than 0.850, and the AUCs of the ROC curves of the risk-of-death prediction model were the highest for both groups at 7 days, at 0.907 and 0.886, respectively. The two sets of calibration curves show that the calibration curves oscillate around a 45° diagonal line at 7, 14, and 28 days, and there is a good correlation between the actual risk and the predicted risk, with a high degree of calibration. The clinical decision curve shows that the model has a strong discriminatory ability when the probability is between 10% and 70%, and the net benefit is greater.Conclusion The variables for constructing the model are convenient and easily available, and the proposed model has good predictive value for the risk of death in patients with A. baumannii sepsis and can be widely used.