Nomogram establishment for gastrointestinal bleeding prediction in ICU patients with traumatic brain injury based on the MIMIC-IV database

Ziming Huang ^1* Hengfa Ge ¹ Ying Sun ²

• ¹ Huai'an First People's Hospital, Huai'an, China
• ² Jiangsu Food & Pharmaceutical Science College, Huai'an, China

Objectives: This study aimed to develop a robust nomogram for predicting the occurrence of gastrointestinal bleeding (GIB) in patients with traumatic brain injury (TBI) during their ICU stay, thereby facilitating the optimization of intervention strategies and enabling personalized treatment approaches. Methods: Patient data were extracted from the publicly available MIMIC-IV (Medical Information Mart for Intensive Care IV) database. In this retrospective cohort study, a total of 2,774 patients with traumatic brain injury (TBI) were included. A 7:3 ratio was applied to allocate patients into the training and validation cohorts. A LASSO logistic regression model was constructed using the training set to identify potential predictors of gastrointestinal bleeding (GIB). The selected features were subsequently utilized to develop a nomogram model. The performance of the nomogram was evaluated using receiver operating characteristic (ROC) curves, calibration curves, and decision curve analysis (DCA). Result: A nomogram model comprising six variables-gender, blood urea nitrogen (BUN), Shock Index (SI), albumin, SOFA score, and diabetes mellitus-was developed. These variables were identified as independent risk factors for gastrointestinal bleeding (GIB) in patients with traumatic brain injury (TBI) (P < 0.05). The area under the receiver operating characteristic curve (AUC) for the derivation cohort and validation cohort was 0.8541 (95% CI: 0.833 to 0.911) and 0.8381 (95% CI: 0.752 to 0.863), respectively. The calibration curve demonstrated good agreement between the predicted probabilities and actual observations, while decision curve analysis (DCA) highlighted the clinical utility of the predictive model.This study developed a predictive model for GIB in patients with TBI, which may assist clinicians in early identification of high-risk patients and help mitigate the burden of GIB in susceptible populations.