Early Functional Recovery Trajectories After Severe Traumatic Brain Injury: A Secondary Analysis of the TBIMS National Database

<b>Background:</b> Survivors of severe traumatic brain injury (TBI) show highly heterogeneous recovery but early functional trajectories across rehabilitation, and their relationship with 1-year outcomes remain poorly characterized. <b>Methods:</b> We performed a secondary analysis of the Traumatic Brain Injury Model Systems (TBIMS) National Database. Adults with severe TBI (Glasgow Coma Scale ≤ 8, post-traumatic amnesia (PTA) > 7 days, or neurosurgical intervention) who received inpatient rehabilitation with Functional Independence Measure (FIM) admission and discharge totals and 1-year Glasgow Outcome Scale-Extended (GOS-E) were included (n = 9438). K-means clustering on FIM admission, FIM discharge, and 1-year GOS-E identified early functional recovery trajectories. Multinomial logistic regression related trajectory class to age, sex, PTA duration, rehabilitation length of stay (LOS), and total LOS. Logistic regression examined associations between trajectory class and 1-year home residence and rehospitalization, adjusted for age and sex. <b>Results:</b> Three trajectories emerged: low-functioning/limited improvers (28.0%), substantial improvers (33.7%), and high-functioning (38.3%). The low-functioning trajectory showed lower FIM scores at admission and discharge and worse 1-year GOS-E than the other trajectories. In adjusted models, older age, longer PTA, and longer LOS were associated with less favorable trajectories overall, particularly reducing the likelihood of belonging to the high-functioning trajectory. Substantial improvers and high-functioning patients had higher odds of living at home and lower odds of rehospitalization at 1 year than low-functioning patients. <b>Conclusions:</b> Simple routine measures can yield interpretable recovery trajectories after severe TBI that may support prognosis, discharge planning, and follow-up; these trajectories should be interpreted as population-level probabilities rather than deterministic predictions for individual patients.