Predicting long-term clinical outcome for patients with traumatic brain injury (TBI) at the beginning of rehabilitation provides essential information for counseling of the family and priority-setting for the limited resources in intensive rehabilitation. The objective of this study is to work out the probability of the one-year outcome at the beginning of rehabilitation. Sixty-eight patients with moderate-to-severe TBI and known one-year outcome were employed for outcome prediction using the logistic regression model. A large number of prospectively collected data at admission (age, Glasgow Coma Scale [GCS] Score, papillary response), during intensive care unit (ICU) management (duration of coma, intracranial pressure [ICP] and its progress) and at the beginning of rehabilitation (baseline Functional Independence Measure [FIM], Neuro-behavioral Cognitive Status Examination [NCSE] and Functional Movement Assessment [FMA]) were available for preliminary screening by univariate analysis. Six prognostic factors (age, GCS, duration of coma, baseline FIM, NCSC and FMA) were utilized for the final logistic regression model. Age, GCS and baseline FIM at the beginning of rehabilitation have been found to be independent predictors for one-year outcome. The accuracy of prediction for a good Glasgow Outcome Score is 68% and an outcome for disability (either moderate or severe) is 83%. Validation of this model using a new set of data is required.