Development and Validation of a Prediction Model for 28-Day Mortality Based on D-dimer to Platelet Ratio in Moderate-to-Severe Traumatic Brain Injury: A Retrospective Cohort Study

Background: Moderate-to-severe traumatic brain injury (msTBI) carries high mortality, yet accurate early prognostication remains challenging. Post-injury coagulopathy, involving elevated fibrinolysis (D-dimer) and platelet consumption (platelet count), is a key driver of poor outcomes. The D-dimer to platelet ratio (DPR) has shown prognostic value in other critical illnesses, but its role in predicting 28-day mortality after msTBI is unexplored. This study aimed to evaluate the predictive value of DPR for 28-day all-cause mortality in msTBI patients and develop a DPR-based prediction model. Methods: This retrospective, single-center cohort study included 319 patients with moderate-to-severe traumatic brain injury (GCS 3-12). Patients were divided into training and validation cohorts (7:3 ratio) via stratified sampling. The primary endpoint was 28-day all-cause mortality. Admission data, including D-dimer to platelet ratio (DPR), were collected. A prediction model was developed in the training cohort using Boruta feature selection and multivariable logistic regression. Model performance was evaluated by discrimination (Area Under the Curve, AUC), calibration, and clinical utility (Decision Curve Analysis). The model was validated internally (bootstrapping) and externally. Results: In the multivariable logistic regression, Glasgow Coma Scale (GCS) (OR: 0.59, 95%CI: 0.45–0.74), admission blood glucose (ABG) (OR: 1.15, 95%CI: 1.02–1.30), and log10-DPR (OR: 2.20, 95%CI: 1.11–4.54) were identified as independent risk factors for 28-day mortality. A prediction model (the DPR-B model) incorporating these three variables was developed. The model demonstrated excellent discrimination in the training (n=223) cohort (AUC: 0.870, 95%CI: 0.809–0.931) and the validation (n=96) cohort (AUC: 0.871, 95%CI: 0.792–0.950). Calibration curves showed excellent agreement, and decision curve analysis confirmed high net clinical benefit, outperforming GCS alone. Conclusions: Admission D-dimer to platelet ratio (DPR) is a novel, independent predictor for 28-day mortality in patients with moderate-to-severe traumatic brain injury. The new DPR-B model, incorporating GCS, admission blood glucose, and DPR, is a simple, robust tool. It provides excellent predictive performance and clinical utility, aiding clinicians in early risk stratification and decision-making for this high-risk population.