RESUMO
Objective To explore the awakening probabilistic prediction models of coma patients with traumatic brain injury on admission and six months after treatment,and develop and apply the software of the models.Methods Clinical data of 190 coma patients with traumatic brain injury,admitted to our hospital from September 2010 to October 2012,were analyzed retrospectively.Potential predictive factors at admission and after awakening were analyzed by binary Logistic regression analysis; based on these factors,the awakening probabilistic prediction models of coma patients with traumatic brain injury were established; C++ language was used to write the computer software that could predict the awakening probability of 103 patients with traumatic brain injury.Results Multinomial Logistic regression analysis showed that 6 factors,including age,pupillary light reflex,movement Glasgow coma scale (mGCS) scores,morphology changes of mesencephalon surrounding cisterna,eye opening time after treatment,and percentages of ischemic brain volume in CT images,were independent factors to predict the awakening probability of coma patients with traumatic brain injury.Model A and B owned high performance (C statistics of models:0.955 and 0.975; accept rate of models:90.5％ and 94.0％).The established software based on models was easy to use with reliable results (the accept rate of 103 patients were 87.3％ and 93.2％).Conclusion The established models can timely and accurately predict the awakening probability of coma patients with traumatic brain injury; the software named sober probabilistic prediction for coma patients with traumatic brain injury can help in decision-making in clinics.