A Preliminary Analysis of Serum Iron's Role in Predicting Sepsis-associated Encephalopathy

Abstract

Abstract Sepsis is a life-threatening condition characterized by organ dysfunction resulting from an uncontrolled response to infection, with the nervous system being particularly vulnerable. Iron is an essential trace element in the human body and is closely associated with sepsis and neurological diseases. The MIMIC-IV database was utilized for a retrospective cohort study involving 936 patients in the intensive care unit who experienced severe adverse events (SAE). These patients were randomly assigned to either a training or validation cohort. Independent risk factors for SAE were identified through LASSO logistic regression. Subsequently, a nomogram was developed incorporating these factors to predict the occurrence of SAE in sepsis patients. The efficacy of the nomogram was evaluated using several statistical measures, including the AUC, calibration curve, Hosmer-Lemeshow test, IDI, NRI, DCA. Furthermore, in order to delve deeper into the correlation between serum iron and the occurrence of SAE, both univariate and multivariate logistic regression analyses were conducted. The analysis revealed that out of the 936 patients, there were a total of 649 cases of SAE. Additionally, the implementation of LASSO regression analysis identified several independent risk factors for SAE, namely mean arterial pressure, respiratory rate, type of microorganism, serum iron levels, elective surgery, SASPIII score, and OASIS score. Moreover, the performance evaluation of the developed nomogram, based on metrics such as AUC, NRI, IDI, and DCA, demonstrated superior results compared to the conventional combination of SOFA and delirium. Moreover, the satisfactory calibration of the nomogram was confirmed by the calibration curve and results of the Hosmer-Lemeshow test. Our nomogram scoring system exhibited superior NRI and IDI values compared to conventional diagnostic methods. The DCA curves demonstrated favorable clinical utility for the nomogram. Multivariate logistic regression analysis revealed that serum iron remained an independent predictor of SAE. Specifically, lower serum iron levels were associated with a higher risk of SAE (OR = 0.997, 95% CI = 0.993-1.000). The findings of our study clearly indicate that serum iron levels significantly influence the diagnosis of SAE.