Development of an early warning model for sepsis-associated encephalopathy using LASSO-Boruta-Logistic regression

This study aimed to develop an early warning model for sepsis-associated encephalopathy (SAE) to facilitate early clinical identification and intervention.

Clinical data from 631 patients diagnosed with sepsis and admitted to Hebei General Hospital between September 2023 and June 2025 were retrospectively analyzed. Participants were categorized into SAE and non-SAE groups and randomly assigned to a training set and a testing set in a ratio of 7:3. Feature selection was performed using least absolute shrinkage and selection operator (LASSO) and Boruta algorithms. Univariate analysis was conducted separately only to describe baseline characteristics. Five variables, including mechanical ventilation, procalcitonin-to-albumin ratio (PAR), triglyceride-glucose index (TyG index), number of infectious pathogens, and Sequential Organ Failure Assessment (SOFA) score, were identified and incorporated into a multivariable logistic regression model to construct a nomogram. Model performance was assessed in terms of discrimination using the area under the receiver operating characteristic curve (AUC), calibration using calibration curves, and clinical utility using decision curve analysis.

Mechanical ventilation, PAR, TyG index, number of infectious pathogens, and SOFA score were independently associated with SAE. A nomogram incorporating these predictors demonstrated strong discriminative ability, with AUC values of 0.897 in the training set and 0.880 in the testing set. Decision curve analysis indicated consistent clinical net benefit across both datasets.

The early warning model for SAE, based on mechanical ventilation, PAR, TyG index, number of infectious pathogens, and SOFA score, demonstrates high accuracy and clinical applicability. This model may assist in the early identification of SAE and facilitate timely clinical decision-making.