A Predictive Model of In-Hospital Death for Traumatic Head and Neck Injuries: A Study Based on the Mimic-IV Database

Head and neck injuries (HNIs) can lead to lethal outcomes, necessitating early prediction of severity in clinical practice.

This study aimed to identify the predictive factors and develop a predictive model for in-hospital death (IHD) among patients with traumatic HNIs.

We used patient data from the Medical Information Mart for Intensive Care IV database, categorizing 6740 subjects into in-hospital nondeath and IHD groups. The analyses included demographic details, clinical profiles, comorbidities, and laboratory findings. Significant variables from univariate analysis were subjected to multivariate logistic regression analysis to create a predictive model. The model was visualized using a nomogram and its performance was assessed using a receiver operating characteristic curve, decision curve analysis, and a calibration plot. Accuracy, sensitivity, specificity, positive predictive value, negative predictive value, and Youden's index were calculated to evaluate the effectiveness of the model. The validation data were used to verify the efficiency of the model.

Multivariate logistic regressions identified 23 predictors of IHD due to traumatic HNIs. The predictive model had an area under the curve (AUC) of 0.899 and a Youden index of 0.662 at a cut-off of 0.08. The predictive model demonstrated good calibration, and the decision curve yielded an acceptable net benefit. The area under the curve was 0.959 when predicting the validation data.

The identified predictors and constructed model offer new insights for identifying patients at high risk of IHD due to traumatic HNIs, indicating the need for intensive care and treatment.