A predictive model for sepsis risk in patients with non-traumatic cerebral hemorrhage based on the MIMIC-IV database

Patients with non-traumatic cerebral hemorrhage admitted to the intensive care unit (ICU) are known to be at high risk for developing sepsis. However, limited research exists to quantify this risk. Therefore, this study aimed to develop a reliable predictive model to assess the risk of sepsis in ICU patients with non-traumatic cerebral hemorrhage. We extracted data on patients admitted to the ICU with non-traumatic cerebral hemorrhage from the Medical Information Mart for Intensive Care IV (MIMIC IV) database. Afterward, the patients were then randomized in a 7:3 ratio into a training set (N = 1,365) and a validation set (N = 585). Least Absolute Shrinkage and Selection Operator (LASSO) regression and stepwise logistic regression were employed to screen variables within the training set. The final logistic regression model was constructed using the identified key predictors. Finally, the model’s performance was evaluated using decision curves, calibration curves, and receiver operating characteristic (ROC) curves. A total of 1,950 patients were included in the study. The training and validation sets comprised 1,365 and 585 patients, respectively. The training set analysis revealed nine crucial predictors for secondary sepsis in ICU patients with non-traumatic cerebral hemorrhage. These factors included liver disease, acidosis, anemia, thrombocytopenia, urinary tract infection, invasive mechanical ventilation, Glasgow Coma Scale (GCS) scores, leukocyte counts, and blood calcium levels. These factors were incorporated into the final model. The area under the ROC curve (AUC) was 0.821 for the training set and 0.845 for the validation set, indicating the model’s high accuracy in predicting sepsis. Calibration curves demonstrated good agreement between the model’s predictions and actual outcomes. Furthermore, the decision curve analysis indicated that the model offers favorable clinical utility. This study successfully developed a dynamic nomogram model for predicting the risk of secondary sepsis in ICU patients with non-traumatic cerebral hemorrhage. The model is expected to provide valuable predictive information to facilitate timely interventions by healthcare professionals.