Clinical decision-making on the prognosis prediction of traumatic brain injury

Prognosis and prediction of traumatic brain injury (TBI) play an important role in clinical decision-making which focuses on the influence of specific disease-related characteristics on prognosis. Given the limitation of previous predictors in patients with TBI, new biomarkers and models must be explored and developed. The predictive factors of TBI prognosis starts from different population characteristics, patient differences and study endpoints related to TBI, finds or establishing univariate relationships, as well as analysing and adjusting relationship of variables related to outcome, and explores the relationship between specific disease characteristics and prognosis. A TBI prediction model of prognosis is built on the basis of multidimensional predictive factors. The performance of this model is explained, estimated, evaluated and verified. Determining proper prognostic markers of TBI not only is useful in building a more accurate TBI prediction model, but also is essential for risk stratification and clinical decision-making after TBI diagnosis. In addition, the building of a TBI prediction model can achieve more accurate and practical comparison among different models, as well as contribute to risk stratification among different patients with TBI. The limitations of research on TBI prognosis prediction are mainly related to the lack of prognostic information, the exploration of nonlinear effects and the evaluation of heterogeneity among studies. This paper reviews the current methods and principles of prognosis prediction of patients with TBI to provide a reference for future investigations on TBI prognosis prediction. We emphasize the importance of patient selection, factor screening and research design in studying predictive factors. In establishing a prediction model, we highlight the necessity of model development and the principle of internal and external validation.