Analysis of Risk Factors and Development of a Predictive Model for Delirium in ICU Patients: Construction,...

To analyze the influencing factors contributing to the occurrence of delirium in patients within the Intensive Care Unit (ICU) and to construct a prediction model for delirium in critically ill patients, subsequently verifying its predictive value.

A prospective study was conducted involving 641 patients admitted between January 2023 and June 2024. A simple random sampling method was employed to develop the predictive model, with a validation set comprising 193 patients, thus creating a training set of 448 patients. Delirium was assessed using the Confusion Assessment Method for the ICU (CAM-ICU). The baseline data of the two patient groups in the training and validation sets were compared. Logistic regression analysis was utilized to identify independent risk factors influencing the onset of delirium. The R programming language was employed to establish a column-line graph model for predicting delirium occurrence in ICU patients. The Bootstrap method facilitated model validation, while calibration curves and Receiver Operating Characteristic (ROC) curves were utilized to evaluate the model's discriminatory ability and predictive efficacy. Finally, the prediction model was validated using the validation set.

In the training cohort, the incidence of delirium among patients was 35.71%. Logistic regression analysis revealed that the Glasgow Coma Scale (GCS) score (OR=0.421, 95% CI: 0.355-0.501, P<0.001), blood urea nitrogen (BUN) (OR=1.169, 95% CI: 1.014-1.348, P=0.031), emergency surgery (OR=2.735, 95% CI: 1.42-5.268, P=0.003), use of sedative medications (OR=3.816, 95% CI: 1.968-7.397, P<0.001), and postoperative status following major cardiovascular surgery (OR=2.124, 95% CI: 1.205-3.745, P=0.009) were identified as independent risk factors for delirium in the ICU.

The predictive model developed in this study for the occurrence of delirium in ICU patients has been validated, demonstrating high predictive efficacy and offering significant clinical early warning guidance.