An externally validated prognostic model for critically ill patients with traumatic brain injury

Patients with traumatic brain injury (TBI) who are admitted to the intensive care unit often exhibit critical conditions; thus, early prediction of in-hospital mortality is crucial. In this study, we aimed to develop a reliable and easily promotable model for predicting the in-hospital mortality of critically ill patients with TBI using easily accessible indicators and validate the model using external data.

Patient data from the Medical Information Mart for Intensive Care-IV 2.2 database were used as training and internal validation sets to establish and internally validate the prognostic model. Data from the Affiliated Dongyang Hospital of Wenzhou Medical University were used for external validation. The Boruta algorithm was used for the initial feature selection, followed by univariate and multivariate logistic regression analyses to identify the final independent predictors. The predictive performance was evaluated using a receiver operating characteristic curve, calibration curve, clinical practicality decision curve analysis, and clinical impact curve.

This study included 3225 patients (training set: 2042; internal validation set: 874; and external validation set: 309). Ten variables were selected for inclusion in the nomogram model: age, mechanical ventilation usage, vasoactive agent usage, intracerebral hemorrhage, temperature, respiration rate, white blood cell count, platelet count, red blood cell distribution width, and glucose. The nomogram demonstrated good predictive performance in both the internal and external validation sets.

We developed an externally validated nomogram that exhibited good discrimination, calibration, and clinical utility for predicting in-hospital mortality in critically ill patients with TBI.