Study design: Retrospective, descriptive study. Objectives: To describe patients' length of stay (LOS), functional status and discharge setting after rehabilitation and how degree of impairment (complete/incomplete paraplegia/tetraplegia) impacts on these outcomes. To compare actual LOS with estimated LOS. Estimated LOS was based on an Australian model, the Australian National Sub-acute and Non-acute Patient Classification System (AN-SNAP), which classifies patients using admission Functional Independence Measure (FIMt) scores. To further describe outcomes for each AN-SNAP class by degree of impairment. Setting: Spinal Injuries Unit of major Metropolitan hospital in Brisbane, Australia. Methods: Retrospective chart review of 167 patients with traumatic spinal cord injury (SCI). Main outcome measures were rehabilitation LOS, discharge FIMt scores and discharge setting. Injury measures were degree of impairment, acute LOS and rehabilitation admission FIMt scores. Standard demographic measures were also collected. Results: The median rehabilitation LOS was 83 days and mean discharge FIMt scores 102 for all patients. These differed by impairment (incomplete paraplegia LOS 43, FIM 117; complete paraplegia LOS 96, FIM 109; incomplete tetraplegia LOS 64, FIM 100; complete tetraplegia LOS 206, FIM 78). Patients discharged to the community (noncare facility) ranged from 93% with incomplete paraplegia to 73% with complete tetraplegia. For patients in the three AN-SNAP classes with the lowest FIM scores, the actual LOS was up to twice the estimated LOS. A large variability in discharge outcomes was found within individual AN-SNAP classes, despite similar FIMt scores on admission. Conclusions: Rehabilitation outcomes differed substantially by impairment. The variability in outcomes for patients within the same AN-SNAP class questions the ability of this system to accurately predict LOS, and therefore cost of rehabilitation services, for patients with traumatic SCI in Australia.