Objective To investigate the relationship of clinical manifestations with prognosis of patients with diffuse axonal injury (DAI) and select the indices that can accurately evaluate the prognosis so as to create a predictive model of outcome of DAI patients. Methods Clinical data of 98 DAI patients were retrospectively analyzed to identify the relationship between the indices and outcomes through Cochran-Mantel-Haenszel x2 test or Student-Newman-Keuls one way variance test via SAS 8.1.Then,the indices was analyzed through multivariant cumulative stepwise way so as to establish the Logistic regression model. Results (1)The indices including Glasgow Coma Score (GCS) on admission,combined with other type of cerebral injuries,unstable vital signs,abnormal pupil size and reactivity,decerebrate rigidity,imaging with characteristic DAI signs,paralysis of the limbs and length of coma were correlated with the outcome,while the indices including gender,age,injury mechanism,pathologic signs and decorticate rigidity were insignificantly associated with the outcome.(2) The logistic regression model showed that GCS on admission,paralysis degree of the limbs,abnormal pupil size and reactivity,and combined with cerebral injuries were the optimum predictors of outcome of the DAI patients. Conclusions The logistic regression model in the study is a reliable tool in predicting the prognosis of DAI patients.The patients wit higher GCS on admission may have better outcome.The patients with light degree of paralysis,normal pupil size and reaction,and combined with none cerebral injuries may have sound prognosis. Key words: Diffuse axonal injury; Prognosis; Regression analysis