Abstract:

Objective To develop and validate prognostic models for mortality within 30 d after traumatic brain injury and for 6-month unfavorable prognosis after traumatic brain injury. Methods The clinical data of 1 016 patients with moderate to severe traumatic brain injury were collected to develop prognostic models. The relationship between admission-related risk factors and prognosis was systematically analysed, different prognostic models were established with Logistic regression analysis, and the performance of these models was assessed with goodness-of-fit test and C statistic (area under the receiver operating characteristic curve). These models were internally and externally validated, the ultimate prognostic model was determined to serve as the tool of prognostic evaluation for patients with traumatic brain injury. Results Logistic regression analysis revealed that old age, loss of pupillary light reflex, decreased motor Glasgow Coma Score, abnormal CT features and abnormal routine laboratory findings were independent risk factors for unfavorable prognosis in patients with traumatic brain injury. The prognostic model based on the risk factors of admission had favorable performance (P>0.05 for goodness-of-fit test, 0.709-0.882 for C statistic). No overoptimism was revealed by internal validation, and the external validity was proved to be better by external validation (P>0.05 for goodness-of-fit test, 0.844-0.922 for C statistic). Conclusion The established model, which is convenient to manage, can timely and accurately predict the prognosis of patients with traumatic brain injury, and can help in decision-making in clinics.

Key words: prognostic model, prediction, traumatic brain injury, validation, prognosis